grid? :2

'3

 

 

 

vt
.1.
" r‘u .S'F‘G‘I
‘J‘hnfis
"‘ .‘3‘

1m,
‘
m“

.2»
.“ ‘ ’3!

Qt

h w. b. .. 3'5):

fimfi’r“”
53':

sh.
”"‘;.~‘~

xv . his)
"LL

‘;~‘.': Q3” ”23% , . ta 9' 3 U 7'
t '1 1"V V ‘l . U. . h ‘ .. '1' .7 ‘ A 2.4. Xq ,‘(!I'.;

k“ w‘l' ““a‘q‘ ' w"; 3 '3' 3 2:51" ‘ ' ' " I" ‘ ‘t . ' Iv {1
5. hyit.l‘“& ' 20;, y It 3' Q I - ‘%":“w '4 a .II,‘ .1 '1 {gr “2
. "8;. ~ '5 t- a." , 3”va Z- 3' g . w I
’i F}. ”4-:

“i
:‘1

: . a t- ‘4;
3.: C , _ - . ‘1‘ 3:,
\ _ 1‘>_~z
. ._ _ §
‘ 3x862». {v‘ :0 5:;

' ' '-.~.. . J.
_‘

l I’M}
1.; 3; .
{$3.5
! 1!, “0%. . ‘
C . / ‘1'
a! . . :5?

J'vv’f“ (1:5? I.“ ‘,'
H‘? 2"?!

$54. 3
3332a}:
- ‘ A -x.

4 ‘5."

f
'f“.\"
tth’h’c"

v * 13‘?”
WSW
(.1?! (A:
gréy‘. 3 ”W
$41.)-
I

C
. 1‘
)3”;

,

 

 

 

 

NIVERSITIY LIBRAR

III...“ I’ ll’ IIIIIII l 1 04 10 5‘9

_3 1293 00564 8310

‘—-—*- I WLIflnI-ARY
Michigan State
University

 

 

 

 

 

 

 

 

 

 

                                                 

 

 

 

This is to certify that the

thesis entitled

BEHAVIORAL AND SYSTEMS DYNAMICS SIMULATION MODELS:
SOME THEORY AND APPLICATIONS
TO A MID- -MICHIGAN DAIRY OPERATION

presented by

Thomas Robert Hebert

has been accepted towards fulfillment
of the requirements for

MS ____ degree in Ag . Econ .

é Sajor professor

Wows/711

0-7639 MS U is an Affirmative Action/Equal Opportunity Institution

 

 

MSU

LIBRARIES
.——.

‘v

 

 

RETURNING MATERIALS:
Place in book drop to
remove this checkout from
your record. FINES will
be charged if book is
returned after the date
stamped below.

 

 

 

 

 

. BEHAVIORAL AND SYSTEMS DYNAMICS SIMULATION MODELS:
SOME THEORY AND APPLICATIONS TO A MID-MICHIGAN DAIRY OPERATION

By

Thomas Robert Hebert

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF SCIENCE

Department of Agricultural Economics

1988

 

 

(‘3

n)

r.

0“

ABSTRACT

BEHAVIORAL AND SYSTEMS DYNAMICS SIMULATION MODELS:
SOME THEORY AND APPLICATIONS TO A MID-MICHIGAN DAIRY OPERATION

By

Thomas Robert Hebert

The primary questions asked concern the possible use of
behavioral theory to create systems simulation models of
farm businesses. Such models would be useful for management
problems and for sector level policy analysis. A systems
dynamics simulation model of a Mid-Michigan dairy operation
was undertaken.

The modelling effort and the literature indicate that
simulation models can represent much of a firm's business
behavior. The potential for using these models for sector
level policy analysis was found to be much less promising.
Although detailed behavioral models of a single-owner farm
operation are possible, their use for sector level market
analysis may require more resources than are practical.

Some potential and practical usefulness was found in some of
the system dynamics models.

The herd management portion of the dairy model was used
to examine different management issues. The results indicate
that such models can be useful for the examination of on-
farm management problems where behavior needs to be
represented in ways not possible with the assumptions of

nee-classical theory.

ACKNOWLEDGEMENTS

Dr. Sherrill Nott, a member of my thesis committee,
generously contributed his knowledge and constructive
insights throughout the research. Dr. Ralph Levine, another
committee member, spent more time with me than I could have
rightfully hoped for. He tolerated my many periods of
despair over the failings of the research, and helped me see
the way out of these failings or find practical
alternatives. It was the willingness of my thesis
supervisor, Dr. J. Roy Black, to tolerate and ask unusual
questions that gave me the opportunity to undertake this
particular effort. Both his breadth of curiosity and
dedication to finding answers continually gave me insights
into what good and exciting research can be. I am grateful
to all three for their time and effort.

My thanks to the many members of the Agricultural
Economics faculty at Michigan State University who explored
and discussed ideas and issues with me, and in general made
MSU an exciting experience. Thanks also to my friends and
fellow students at MSU, who studied hard and relaxed with
fervor, and who helped me to remember that our work is only
a part of life.

Deep thanks must also go to my father and the other
members of my family, who have been there throughout my
education and this research. I hope they know how much I
appreciate their support. In particular, I thank my sister
Jeanne, who by opening her home to me let me share in many
of the important parts of life that should not be missed,
even when there is work to be done.

And finally, my warmest thanks to Phyllis Landick,
without whose help this effort would never have been
possible. Her commitment to loving and respecting all of
that which makes me human taught me about loving and
respecting myself.

ii

BACKGROUND AND THE AREAS OF INVESTIGATION

Behavioral and System Dynamics Approaches to

TABLE OF CONTENTS

CHAP ER ONE

System Dynamics and Behavioral Theory
The Scope and Design of this Research

HAPTER TWO
ION P OCE SES AND HE FIRM

Subjective Expected Utility and Bounded

Applications to the Theory of the Firm

Other Applications of Behavioral Theory to

T

RE
A

I GL OWN R

RM

heory of the Single Owner- -Operator Firm

Operational Goals and Functional Areas

CH P O R
ULATION MODEL OF A MID-MICHIGAN

DAIRY FARM

Feedback with Respect to the Primary

The Entire Dairy Operation

1.1 Introduction
1.2
Modeling
1.3
1.4
B VIO AL HEOR ES OF E
2.1 Introduction
2.2
Rationality
2.3
2.4 Standard Operating Procedures
2.5 Summary of Behavioral View
T
V O T
3.1 Introduction
3.2 A T
3.2.1 Primary Goals
3. 2. 2
3.3
Small Firms
-3.4 Summary of the Small Firm Theory
A BEH V O SYSTEMS
4.1 Introduction
4.2 Primary Goals
4.2.1 Other Characteristics
4.2.2
Goals
4.3 Subsystems and Operational Goals
4.4 Causal
4.4.1 Introduction
4.4.2
4.4.3
4.4.4 Revenues

Cash and Expected Cash Flow

iii

Loop Diagrams of the Dairy Farm Model

U1

10

11
13
15
17

19
2O
20

23

26
28

30
31
32

33
37
4O
4O
41
42
44

mm

0‘

O‘O‘O‘O‘O‘O‘O‘O‘O‘

\IN

WNH

\DQNO‘UTkWNl-J

NH

4.4.5 Variable Costs .
4.4.6 Capital, Debt and Assets
Summary . .

H T FIV
OM HEO Y D APPL CATIONS O
BEHAVIORA I ULA ON MODE

Introduction

Judging the Validity of Simulation Models

Some Literature on Behavioral Simulation and

a General Theory of Markets .

The Literature of Behavioral Simulation

Models . . . . . . . . . . . . . .

5.4.1 Cyert and March; A Duopoly Model

5.4.2 Balderston and Hoggatt; A
Decentralized Market System .

5.4.3 Summary of the Implications of these

Studies
5.4.4 Meadows; A Dynamic Commodity Cycle
Model . . . . . . . . . . . . . . .
5.4.5 Lyneis; Corporate Planning and Policy
Design . . . . . . . . . . . . . .
Conclusions: Simulation and Sector Analysis

Answers to Research Questions

A T IX
IflE HERD MANAGEMENT SIMULATION MODEL

Introduction

Model Overview

Calves

Heifers

Young Cows

Cows

Desired Milkers

Cash Flow

Detailed Description of the Herd Management
Sub- Model

.10 Summary

W
O 0 NA EMENT OD
Introduction
The Herd Dynamics Without Purchases of
Replacements

7.2.1 Run IA; Baseline For No Purchases

iv

47
49
51

56
57

61

62
62

65

74

79

83

89
94

96
97
98
99
100
100
101
104

105
120

122

126
126

7.2.2 Run IB; Altering Calf Death Losses

. . . . . . . . . . . . . . . . 129
7.2.3 Run IC; Altering the Cull Rate
Standard For Cows . . . . . . 132
7.2.4 Discussion of Runs IB and IC . . . . 134
7.2.5 Run ID; Adequate Stocks of
Replacements . . . . . 135
7.2.6 Model Viability and Validity . . . . . 137
7.3 SOP on Constant Purchases of Replacements . . 137
7.3.1 Run IIA, Constrained Stock, No
Purchases . . . . . . . . . . . . . . 138
7.3.2 Run IIB, Constant Purchases SOP . 140
7.3.3 Run 116, Constant Purchases SOP,
Expand the Herd . . . . . . . . 142
7.3.4 Run IID, Constant Purchases, Expand
Young Milkers and Heifers . . . . . . . 147
7.3.5 Run IIE, Constant Purchases, Expansion
only in Heifers . . . . 150
7.4 SOP on Variable Purchases of Replacements . . 152
7.4.1 Run IIIA, Baseline for Variable
Purchases . . . . . . . . . . . . . . . 152
7.4.2 Run IIIB, Increasing the Purchasing
Delay to Four Months . . . . . . . . 154
7.4.3 Run IIIC, Variable Purchases,
Expansion in Month 60 . . . 157
7.5 Linking the Desired Milker Goal to Feedback . 160
7.5.1 Run IVA, B, and C, Analysis of
Averaging Times for Feedback . . 163
7.5.2 Run IVD, Linked Goal, Constant SOP on
Purchases, Expand with Heifers . . . . . 166
7.5.3 Run IVE, Linked Goal, Variable
Purchases SOP, Expand with Young Cows . 168
7.6 Summary of Simulation Results . . . . . . . . 171
7.7 Conclusions . . . . . . . . . . . . . . . . . 172
CHAPTER EIGHT
M AND O U ION ' AVIOR L M ND
H IS U AXIMIZATION
8 1 The Primary Research Questions . . . . . . 174
8.2 A Return to the Issue of Maximization . . . 175
ENDNOIES . . . . . . . . . . . . . . . . . . . . . . . 180
AP NDI O E‘ Y T M Y C E TI S ND FLOW
DIAGRAM§_, . . . . . . . . . . . . . . . . . . . . 183
P N I TWO‘ A S M Y MI G A OF THE HERD
G M 0 WI FIN TION O VARI LE . 186
BIBLIOGRAPHY . . . . . . . . . . . . . . . . . . . . . 191

Table 4.

1:

LISI OE TABLES

A Summary of the Farm's Subsystems and the

Associated Goals

Table 4.

2:

The number of variables, by type, in the

dairy farm simulation model, as compared to
Meadow's model.

Table 7.
Table 7.
Table 7.
Table 7.
Table 7.

IIA

Table 7.

la:
1b:
1c:
1d:
2a:

2b:

No Purchases, Baseline, Run IA

Altering Calf Death Losses, Run IB . .
Altering Cull Rate Standard Cows, Run IC
Adequate Stocks of Replacements, Run ID
No Purchases and Constrained Stock, Run

Constant Purchases SOP, Constrained Stock,

Run IIB.

Table 7.

2c:

Constant Purchases, Expansion in Month 60

(Exogenous Shock), Run IIC

Table 7.

Heifers

Table 7.

Heifers
Table 7.
Table 7.

2d:

2e:

3a:
3b:

1118

Table 7.

3c:

IIIC

Table 7.
6.
Table 7.
on
Table 7.
on

4a:

Constant Purchases, Expand Young Cows and

in Month 60; Run IID

Constant Purchases, Expand Herd With Only

Only in Month 60, Run IIE

Baseline for Variable Purchases, Run IIIA

Altered Time Constant from Baseline, Run

Baseline with Expansion in Month 60, Run

Analysis of Averaging Times for Feedback:

4 and 2 months, Runs IVA, IVB and IVC

4b:

Endogenous Desired Milkers, Constant SOP

Purchases, Expand with Heifers

4c:

Endogenous Desired Milkers, Variable SOP

Purchases of Young Cows to 90; Run IVE

vi

39

53
127
130
132
135
138
140
144
147

150
153

153

158

163

166

169

L S F CURE

Figure 4.1: Short~Term Cash Position, Accounts Payable
Loop. . . . . . . . . . . . . . .

Figure 4.2: Generation of Revenues Loop.

Figure 4.3: Variable Costs Loop. .

Figure 4.4: Capital, Debt and Assets Loop.

Figure 6.1: Partial Causal Loop Diagram of Herd
Submodel; With and Without an Endogenous Desired
Milker Goal. . . . . . . . . . . . . . . .

Figures 7.1a-b: Baseline For No Purchases, Run IA. .

Figures 7.2a-b: Baseline with Calf Death Loss Altered,
Run IB. . . . . . . . . . . . . . . . . . . . .

Figures 7. 3a- b: Baseline with Cull Rate Standard for
Cows Altered, Run IC. . . .

Figures 7. 4a- b: Baseline with Adequate Stocks of
Replacements, Run ID.

Figures 7. 5a- b: No Purchases and Constrained Stock,
Run IIA. .

Figures 7.6a-b: Constrained Stock, No Purchases, Run
IIB.

Figures 7.7a-b: Constant Purchases, Expansion in Month
60 (Exogenous Shock), Run IIC. . . . . . . .

Figures 7.7c-d: Constant Purchases, Expansion in Month
60 (Exogenous Shock), Run IIC. . . . . . .

Figures 7.8a-b: Constant Purchases, Expand the Herd
with Heifers and Young Cows, Run IID.

Figures 7.8c-d: Constant Purchases, Expand the Herd
with Heifers and Young Cows, Run IID.

Figures 7.9a-b: Constant Purchases, Expand Herd With
Only Heifers Only in Month 60, Run IIE

Figures 7.10a-c: Baseline for Variable Purchases, Run
IIIA. . . . . . . . . . . . . . . . . . .

Figures Ila-c: Altered Time Constant from Baseline,
Run 1113

Figures 7.12a-c: Variable Purchases with Expansion in
Month 60, Run IIIC. .

Figure 7.13: Causal Loop Diagram Linking Performance
to the Desired Milker Goal.

Figures 7.14a- c: The Effect of Differing Averaging
Times for Feedback: 6, 4 and 2 months, Runs IVA,
IVB and IVC. . . . . . . . . . . . . . . . . .

Figures 7.15a-b: Endogenous Desired Milkers, Constant
SOP on Purchases, Expand with Heifers . . . .

Figures 7.16a-c: Endogenous Desired Milkers, Variable
SOP on Purchases of Young Cows to 90; Run IVE

Figure A1: The three classes of variables in a systems
model . . . . . . . . . . . . . . . . .

IFigure A2: System dynamics diagram of the herd
management simulation model. .

vii

43
45
48
50

102

128

131

134

136

139

141

145

146

148

149

151

155

156

159

161

165

167

170

182

183

CH T R ONE
A GRO ND N T EAS NV STIG IO

1.1 Introduction

Two assumptions lie at the heart of most theories of
market economies. The first is that people make choices
that are intended to have favorable impacts on their own
interests or utility. The second is that people are in
general rational, and that they use this capacity when they
choose courses of action. Neo-classical theory of aggregate
economic behavior extends these assumptions into a third;
that people act as if they choose courses of actions that
lead to a maximum level of utility.

The third assumption is crucial because it leads
directly to much of the theory's explanatory properties,
particularly if we treat the level of monetary returns to a
course of action as a proxy for utility. All that is then
needed to explain and predict behavior in a competitive and
riskless economy is the mathematical function defining the
transformation of inputs to outputs as well as input and
output prices. Essentially, these three assumptions permit
nee-classical theory to be broadly used as an instrument
for prescribing policy actions that affect particular
markets or other highly aggregative sectors of the economy.1

Although the first two assumptions appear to be
txniformly accepted by most economists, the third crucial
aissumption does not enjoy the same degree of assent. In
IJarticular, Herbert Simon and others in the “Carnegie"

1

2
school emphasize the significant limitations of human
cognitive capacity and have developed a theory of economic
behavior that does not rely either explicitly or implicitly

on the assumption of maximization.2

The Carnegie school's
term for their perspective on human rationality and the
resulting implications is "behavioral theory". An obvious
limitation of behavioral theory from the viewpoint of many
economists is that without the assumption of maximizing
behavior, theories of market economies immediately lose much
of the valuable market-wide analytical capacity associated
with neo-classical theory.

The principle question being explored in this research
follows from the cognitive, behavioral approach, asking:

Can the behavioral theory of the firm, with its
emphasis on non-maximizing behavior and the use by firms of
standard operating procedures to pursue goals, be used to
meaningfully represent the decisions and actions of a
particular producer of agricultural commodities?

There are numerous, potentially useful applications of
behavioral theory if it can help explain and predict the
behavior of individual firms. For instance, applications
to farm management problems should be possible since
behavioral theory emphasizes decision making processes in
risky and uncertain environments. This and other

aipplications of behavioral theory to micro-level problems

are conceivable .

3

The farm management application deals with behavior and
problems at the level of the individual or the firm.
Theories of the firm also play another important role in
economics. As stated above, the micro-assumptions of firm
behavior in neo-classical economics lead directly to a
theory of resource allocation governed by prices in
competitive markets, a powerful tool for analysis of markets
and sectors composed of markets. A behavioral theory of
micro-behavior would be of added value if it could be
similarly used as a means for sector or market analysis.

Some researchers believe that such a new theory is
needed. The micro-assumptions of neo-classical theory are
criticized as having little empirical base, a problem which
is magnified by the impossibility of directly testing some
of the assumptions.3 Leibenstein and others ask that if the
micro-assumptions of neo-classical theory are not observable
and testable, and if in fact the assumptions bear little
resemblance to the behavior of firms in the real world,
will there not be serious shortcomings to the theory of all
firms in the aggregate.4

Behavioral theory, which is based on empirical
observation and more clearly testable assumptions, would be
of significant help beyond firm level problem solving if
the theory could be used to generate a new theory of
markets, or at least used as a tool for sector level policy

analysis. If there is a need for an empirically based

behavic
problem
fact re

be exte

sector I

1.2 Beh
TheI

been cre

aPPfoach

4
behavioral theory which could be applied to sector level
problems, then we need-to ask if behavioral theory can in
fact represent the behavior of firms and if the results can

be extended to more aggregative analysis.

The second principle research question follows from

this need:
If behavioral theories prove useful for explaining and
predicting behavior of individual firms, is there a

potential for developing it as a tool for performing
sector level policy analysis?

1.2 Behavioral and System Dynamics Approaches to Modeling

The behavioral simulation models of firms that have
been created by researchers associated with the Carnegie
approach bear a resemblance to expert systems, utilizing
simulation models of the decision processes within firms
given the behavioral constraints from the theory. The
complexity of these models generally necessitates that
computer simulation must be used to explore their
implications and functioning.

While this is a valid method for exploring the
implications of behavioral theory, it is also costly; the
design and construction of such models is extremely time
intensive. A method which permitted a more aggregative yet
behavioral view of the firm would be of value if it could
reduce the time and effort needed to construct meaningful

models. One such potential method is that used in system

dynamic
Forrest
Techno]
IE
dynamic
learned
incorpo
procedu
the con
above,
followi1
Car
agricult
theory 1
Th'

fOCUS 0f

1.3 Sys
As
the P0:
as defir
schocl’
Simulatj
System

enough c

 

behavioe

inter65f

5
dynamics, an approach which was initially developed by Jay
Forrester and others at the Massachusetts Institute of
Technology.

The modeling language techniques used by the system
dynamics researchers, available in the relatively easily
learned packages DYNAMO and STELLA appear to be suited for
incorporating expert opinions about actual operating
procedures and non-maximizing behavior into farm models. In
the context of the two primary research questions discussed
above, it would be useful to discover the answer to the
following question:

Can system dynamics be of use to researchers in
agricultural economics attempting to incorporate behavioral
theory into firm level and sector level analysis?

This question addresses the third and final area of

focus of this research.

1.3 System Dynamics and Behavioral Theory

As posed above, the three research questions involve
the potentially confusing combination of behavioral theory,
as defined by Simon and the other members of the Carnegie
school, with the theory of system dynamics. System dynamics
simulation models focus on aggregative processes within the
system in question. The models attempt to capture just
enough of the relevant structures of the system and the
behaviors of its participants to simulate the problem of

interest. There is a significant difference between this

approaci
of the I
Al
thesis
method *
model c
discre:
asnects

model n

p
.

As
occUl’I‘e
IEsult
itself
1151118 5
inVesti
level 1
litera

I

(SOP: S

OPQrat

 

 

 

 

6
approach and the explicit simulation of decision processes
of the Carnegie behavioral models.

Although the theoretical perspective used in this
thesis is behavioral in the Carnegie sense, the modeling
method used is from system dynamics. As a result, the
model created will have behavioral characteristics (i.e.,
discrete decision processes) as well as aggregative systems
aspects. The model is then in a sense neither a behavioral

model nor a system dynamics model.

1.4 The Scope and Design of this Research

As this research progressed, significant changes
occurred in its scope and design. The changes occurred as a
result of the effort to model a dairy operation, which in
itself provided a significant education on the potential for
using system dynamics and behavioral simulation models to
investigate firm-level behavior and the associated sector
level results. In addition, certain important pieces of
literature were not consulted at the start of the research.

The research initially was to be organized as follows:

1) Develop the basic standard operating procedures
(SOP's) governing the operation of a hypothetical dairy

operation;

2) Assemble all of the necessary production and cost
data;

3) Use the system dynamics software to construct a be-
havioral systems simulation model of the dairy farm using
the SOP’s and cost and production data;

4)
questio:
the res.
experts

5)
sector ;

l

of bet.

Fo'

model.

had to l

 

 

place,
discove
model 3
“P Var:

Tb
back tc
Experi¢
better
more i‘
Primar

A
princi
of the
comple
SuCceS
Progre

7
4) Explore the model through simulation, asking the
question "does it represent a real operation" by comparing
the results to available data and the expectations of
experts;

5) Consider the implications of these results for dairy
sector policy analysis.

Following this design, the research began with the use
of behavioral theory to design and assemble the simulation
model. Well into the assembly phase, the modeling process
had to be halted for even though most of the model was in
place, its complexity made the numerous bugs difficult to
discover and resolve. With the hope of coming back to the
model with a clean mind, attention was turned to writing
up various portions of the completed research.

The attempt to do the write-up, which involved going
back to the literature, led to the discovery that the
experience of building the model allowed more informed and
better questions to be asked. The better questions led to
more instructive literature and partial answers to the
primary research questions discussed above.

As a result, the nature of this research changed,
principally with respect to the best uses of the components
of the dairy simulation model that had been constructed and
completed. This thesis is organized largely according to the
successive stages of learning which occurred as the research
progressed so as to emphasize the effects of accumulating

knowledge on the research's final shape.5

Be?

their 1

 

examine

of the
agricult
i

particu

charactu

applied

 

adaptat
Three,
develop
Four_ CI
°Perati
c°mP1et
Ch
behavio
Chapter
ProceSS
the 11:
sy5tems
one, t‘
of the

3

8

Behavioral theory, as developed by Cyert and March in
their seminal book, A BEHAVIORAL THEORY OF THE FIRM [9], is
examined first in Chapter Two. Their book presents a theory
of the multi-person firm or business organization. Although
agricultural operations in general and dairy farms in
particular do have important multi-person firm
characteristics, the decision was made in this research to
focus on a dairy operation that has a single owner and
manager making the decisions. As a result, Cyert and
March's theory needed some reformulation so that it could be
applied to this situation. This single owner-operator
adaptation of behavioral theory is developed in Chapter
Three. This theory is then used to verbally and pictorially
develop a behavioral model of a dairy farm in Chapter
Four. Chapter Four concludes with a summary of the
operational problems of assembling and debugging the
complete dairy farm simulation model.

Chapter Five is a summary of relevant portions of
behavioral and system dynamics simulation literature. This
chapter coincides with the interruption to the modeling
process referred to above and represents a summary of what
the literature and the experience of building a behavioral,
systems model contributes to answering research questions
one, two and three. Conclusions are drawn about the ability
of the behavioral and system dynamics models to represent

real operations and their applicability to sector analysis.

61
applies
more ;
knowle:
simula:
than t3
manage:
capabil

Ch

conclus

laximi:

 

9

Given the conclusions of Chapter Five, different
applications of the system dynamics methodology are seen as
more practical. Chapter Six and Seven use this new
knowledge to develop a problem orientated herd management
simulation model. Although significantly smaller in scope
than the behavioral model of the dairy farm, the herd
management model demonstrates some of the problem solving
capabilities of the system dynamics methodology.

Chapter Eight presents a summary of the major
conclusions of this research and revisits the issue of

maximization in behavioral models and economic theory.

CODCEIE

pproac

 

 

is the
reflect

0

P“)

firm

that he

CHAPTER TWO

H VI RA EOR ES OF DECISION ROCES ND TH FIRM
2.1 Introduction

Although the neo-classical theory of the firm has been
either the dominant or at least one of the major voices in
theoretical and applied economic analysis for the past four
decades, there has been a concurrent debate concerning the
effectiveness and validity of this theory. The major
concern for researchers connected with the Carnegie
approach (including Simon, Cyert, March, Cohen and others)
is their belief that the standard view of the firm does not
reflect the actual goals, behavior and cognitive abilities
of firm decision makers. These researchers also believe
that neo-classical theory ignores the internal structure of
the firm and its potential effects on decision-making6.
They make a case for narrowing the focus of study to
decision making processes within the firm. Leibenstein
calls this general class of theory “micro-micro” theory. He
emphasizes building a theory based on the actual behavior
of the individual firm member. Leibenstein believes there
is a potential for creating a behavioral, positivist theory
of the firm, and in turn, of markets as well [18].

While Leibenstein's micro-micro theory considers the
minutiae of individual behavior in multi-person firms, the
Carnegie approach to firm theory draws back from the

individual processes. This chapter focuses on human

10

 

 

behavic
outlinil
some cc

placed

bypothe

sinula:

 

 

ll
behavior from the perspective of the Carnegie school,
outlining some of the basic concepts of this approach, with
some contribution of related work by Heiner. Emphasis is
placed on the theoretical foundations for the use of
hypothesized firm standard operating procedures (SOPs) in

simulation models.

2.2 Subjective Expected Utility and Bounded Rationality

According to Simon, the two primary propositions of
neo-classical firm theory are that the firm seeks to
maximize profits (or the discounted stream of future
profits) and it operates with perfect knowledge (or
knowledge of the probability of future states of the world)
[9, pages 8—9]. Simon subsumes these and related
propositions under one title, subjective expected utility
(SEU) [28, page 12].

SEU assumes the four following conditions:

1) The decision maker has a well defined utility
function which cardinally defines preferences concerning

particular future events.

2) The decision maker has a well defined set of
alternative actions to chose from.

3) The decision maker can assign a joint probability
distribution to each possible future state of the world.

4) The decision maker chooses the strategy which will
maximize the expected value of this utility function.

Simon's believes that SEU is in fact never applied in the
real world, ”with or without the largest computers" [28,

page 14]. He cites the behavioral research on human

subject;
situati:
behaviO'
page 17
the fat“

reasoni:

— I — —
—
. —
—

problem:
Simon's

Bo:
wOuld (
every c
decisio:
°f 1“fit
of hUma:

 

Bo;

seek sc

that Vii

conditlc

 

12
subjects of Tversky [30] which concludes that even in game
situations significantly simpler than real life, "human
behavior departs widely from the prescriptions of SEU" [28,
page 17]. Simon's conclusion is that "humans have neither
the facts, nor the consistent structures of values nor the
reasoning power at their disposal that would be required to
apply SEU principles" [28, page 17].

The limited cognitive capacities of human beings means
that we live in a world in which we are able to detect only
a relatively small number of variables or considerations
that impact on us. As a result we are forced to relate to
problems as if they were contained within separable spaces.
Simon's term for this constraint is bounded rationality.

Bounded rationality permits us to make decisions that
would otherwise be impossible if all of the implications of
every decision had to be simultaneously considered in the
decision calculus. Most of the decisions we make are not
of infinite time horizon, do not involve the entire range
of human values or preferences, and are not interconnected
with every other problem within our environment.

Bounded rationality implies that instead of trying to
seek solutions which are optimal, actions will be sought
that will result in performance that is good enough for the
conditions at hand. Simon calls this behavior "satisficing"

[27].

 

Ih
perform
goals,
not nak
local
is on ti
that we
that we

that at]

 

2.3 A?‘
Cy.
theory
FIR! :9
Unit ‘1
of Seve
aspira:
indepet

3!

13

The theory implies that firms will seek a level of
performance which comes as close as possible to their
goals, rather than an optimal level. Behavioral theory is
not making a distinction between decisions that lead to a
local rather than a global maximum. The focus of the theory
is on the actual processes used by firms to find solutions
that work within a certain range of performance. Solutions
that work in a local range are different from solutions

that are optimal for the range.

2.3 Applications to the Theory of the Firm

Cyert and March extend and expand Simon's work into a
theory of firms in their book, BEHAVIORAL THEORY OF THE
FIRM [9]. The firm in this theory is not viewed as a single
unit with a single goal. The firm is seen as being made up
of several different sub-units, each with its own
aspirational goal and its decisions being made relatively
independently of the other sub-units. The goals of the firm
are multiple and changing, but they are seen as falling
into the categories of inventory, production, profit, and
sales.

Each sub-unit observes the performance of the firm in
relationship to its own goal and treats perceived obstacles
to achieving its goal as a problem to be solved. Examples
of such problems are inadequate sales or market share,

inadequate profits or inappropriate inventories or

product

on the

 

by nanyl
given b
implies
discove:

is seart

decisio:

 

Cye
determi;
Iakers
results
I‘Ules u.
the has

Th
behavio
Potenti
leibehs
love C1
8031 de
the°ry.
”Ork 0n

s U
ub'SOa

14
production levels. Performance relative to goals is judged
on the basis of satisficing.

Simon has also developed a theory of the process used
by many organisms as they search for solutions to problems,
given bounded rationality. As applied to firms, the theory
implies that solutions to problems are searched for and
discovered sequentially. The order in which the environment
is searched determines the solution discovered and the
decisions made ([9, page 10], [28, page 23]).

Cyert and March consider feedback to be an important
determinant of firm behavior. Feedback informs decision
makers about the results of past decisions, with poor
results indicating that corrections need to be made. The
rules used to guide these decisions are also adjusted on
the basis of feedback from past experience.

The focus of Cyert and March's theory is on the
behavior of sub-units of the firm, with each unit
potentially being composed of a large number of people.
Leibenstein believes that this approach to firms does not
move close enough to the individual and the problems of
goal determination to represent a complete micro-micro
theory. In particular, a more complete theory would mean
work on the development of goals from within the context of
sub-goals and "a theory of conflict resolution of such
sub-goals” [18, page 483]. Certainly, there could be

beneficial uses of such a theory, but there are also

likely

2.4 S:
Cy
procedu
genera:
experie
23-29].
process
(rules
Confron

28-29],

a SEque
e“Vito:
C0ndit:
a COurs

Provide

comPllc

another

an aCCe
Series

satisfa
solutloJ

othEr’ (I

 

15
likely benefits to theories of the firm whose minimum level

of aggregation is that used by Cyert and March.

2.4 Standard Operating Procedures

Cyert and March emphasize the use of standard operating
procedures within firms. Both humans and firm sub-units
generate rules of behavior to use in response to commonly
experienced situations ([9, pages 100-101], [28, pages
23-29]. Simon emphasizes the combined workings of search
processes involving analysis and rule governed response
(rules in Simon's example being intuition) when people are
confronted with novel yet familiar situations [28, page
28-29].

Bounded rationality, in Simon's search theory, leads to
a sequential examination of the conditions in the immediate
environment surrounding the organization. Not all
conditions or possible course of action are examined, rather
a course of action is often selected if it is found to
provide an acceptable solution. If the problem is
complicated, this decision will need to be followed by
another, which also initiates another sequential search for
an acceptable solution. The final solution, the result of a
series of searches each concluded with a decision and a
satisfactory solution, is only one of many possible final
solutions. Each could be radically different from the

other, depending on the order in which the environment was

 

searche

Th
|
and di'fE
given b
to rene
proble:
current
reduce
courses
Ienembe
Procedu
Organiz

the [as

 

Again,
accepta
optimal
necesSa

He
and ti
COQStIE
Can be:
V111 be

the tin

indivr.

the d

16
searched.

This process of searching for solutions is a demanding
and difficult activity. Simon also holds that this fact,
given bounded rationality, leads organizations (organisms)
to remember the series of decisions made in response to past
problems and utilize this series in the attempt to solve
current similar problems. In this way, organizations can
reduce the time and effort spent in searching for acceptable
courses of action. These series of courses of action
remembered by an organization are it's standard operating
procedures (SOP's). SOP's serve as the memory of
organizations, storing the knowledge gained by experiencing
the results of past searches for solutions to problems.
Again, the fact that past solutions to problems were found
acceptable on the basis of adequate performance rather than
optimal implies that the SOP's used today will not
necessarily be an approximation of maximal rules.

Heiner presents a theory of how imperfect information
and the imperfect ability to use it together work to
constrain the size of the set of actions economic actors
can beneficially chose from (the range of activities that
will be permitted in the SOP's). His own conclusion is that
the theory, (termed reliability theory) "... implies the
necessity of using rules and procedures to constrain
individual agent's decisions and information spaces, and

the dynamic flow of information between them" [14, page 2].

He is
operati

He

 

 

neo.cla
incorpo
potenti

neo-cla

 

various

 

 

Organiz.

necessi

 

 

agents

2.5 su,
Bel
°f Per
firm d
Stated
th
(iifficu
empiric

11

17
He is essentially making a case for the use of standard
operating procedures.

Heiner's theory is stated in terms familiar to most
neo-classical decision theorists, and is an attempt to
incorporate the effects of imperfect information and the
potential for imperfect decisions into standard
neo-classical firm theory. He applies this theory to
various problems, for instance; the nature and functions of
organizational structure, given such errors [14], and the
necessity of adjustment delays in the actions of economic

agents [15].

2.5 Summary of Behavioral View

Behavioral theories of the firm were developed because
of perceived inadequacies in the neo-classical approach to
firm decision making. The issue for many researchers as
stated by Cyert and Hendrick, seems to be that even though
"... the problem (of revising economic theory) is clearly
difficult,... we wonder whether economics can remain an
empirical science and continue to ignore the actual
decision-making process of real firms", ([8, pages

398-412], as cited by Leibenstein [18, page 482]).

The behavioral school takes issue with neo-classicism
on at least two points:

1) Profits are not the sole goal of firms. Goals are,
rather, multiple and inter-related. A better depiction of

the 0;
these

2
cogniti
goals. I
aspira:
limita:
perform

CT
standa:
permit
experie
decisi:
The Vor
decisic
Support

declSlC

 

18

the operations of firms requires a better specification of
these goals;

2) Individuals and firms do not have the information or
cognitive capacity to permit day to day maximization of
goals. Human bounded rationality leads to a pursuit of
aspirational goals, goals that are good enough given the
limitations of the decision makers and the needed
performance.

One of the results of bounded rationality is the use of
standard operating procedures by decision makers. SOP's
permit the combination of knowledge gained from past
experiences with the use of analytical search processes in
decision maker's attempts to find solutions to problems.
The work of Heiner, which incorporates informational and
decision making error to neo-classical decision theory,

supports the contention that SOP's will dominate the

decision making environment.

operatic
number
focuses
in firm
and the
perform
Cyert a
l singl
of indi
Th
relativ
3“ ind
Emil-0:
for the
80315.
dEVelopl
Hid‘Hic,
{Echniq

°f a h“.

 

CH T T E
A AV A HE R H SM LL IN N RM

3.1 Introduction

Cyert and March's behavioral model of the multi-person
firm. is not directly applicable to the large number of farm
operations with only one manager and a relatively small
number of hired employees. The multi-person firm theory
focuses on the specialized sub-units generally observable
in firms, the processes by which sub-units form their goals
and the impacts of these goals on firm behavior and
performance. Yet, even with the multi-person orientation,
Cyert and March's theory should be applicable to firms with
a single owner and manager since it is grounded in models
of individual cognitive capacity.

This chapter attempts to use behavioral theory to model
relatively small single operator farms by emphasizing how
an individual's cognitive limitations in a complex
environment might lead to the development of both the goals
for the farm and the means of managing it to achieve these
goals. This theory is applied in the next chapter to the
development of goals and management techniques for a
Mid-Michigan dairy farm. These goals and management
techniques are then linked in a behavioral simulation model

of a hypothetical dairy operation.

19

 

a tea:
functic
for eat
charac:
used tc
be made

busines

 

exister
Within
behavic
the mar
taking

Tl
As de'

°Perat

20

3.2 A Theory of the Single Owner-Operator Firm
3.2.1 Primary Goals

Business operations, whether large or small, exist for
a reason, to fulfill a particular function or set of
functions. Although the particular reason can be different
for each owner or operator of each firm, one means of
characterizing a particular firm is according to the means
used to fulfill this function. The characterization cannot
be made independent of the environment in which the
business operates, yet each business' particular reason for
existence determines in part the operation's orientation
within its environment. One way to help determine the
behavior of a particular firm would then be to ascertain
the manager and or the owner's primary goals, or reasons for
taking part in the business activity.

These primary goals do not need to be complex to serve,
as determinants of how a particular farmer would orient his
operation to the world. Examples of these goals might be:

- I want to be a farmer, with a high degree of
self—reliance and independence.

- I want to continue in the profession and lifestyle
that my father and mother followed.

- I want my farm to be an example to my community of
modern techniques and high efficiency so that I can be a
community leader.

- I want to have enough income to fulfill the material
needs of my family.

- I just want to generate enough income that I can keep
my operation going for as long as possible.

Althoug

 

they 1
differe

them.

operati
goals.

rafere:

 

accepts
Profits
attain
order C
include
the Cc
predic:
Stekini
Tl
Specia:
therefe
°f to.
firm.
are the

dilers.

21
Although each of those goals are similar in many aspects
they imply a different set of needs to be fulfilled and
different ways of operating in the farm economy to achieve
them.

There is little theoretical difference between the
application of goal analysis to small single operator firms
and the multi-person situations investigated by Cyert and
March. Cyert and March hold that in a large, multi-person
operation different members of the firm have different
goals. These goals cannot be adequately specified by
reference to some individual profit function if one
accepts: 1, that people value things other than just
profits; 2, that there are means other than profits to
attain these valued things; and, 3, that people will to
order their lives to achieve them. If utility functions
include more than just profits, a more complete reference to
the content of these functions is necessary to adequately
predict the behavior of a person as a result of goal
seeking activities.

The multi-person business with a certain degree of
specialization entails information and transaction costs and
therefore niches within the business that permit the pursuit
of goals independent of the goals pursued by others in the
firm. For Cyert and March, the fundamental question is how
are the goals of the operation formed given these richly

divergent utility functions and their associated complex

institt

the sum
the mul
of the

thereby
costs a
larger

is in t
the ope
for lar
Vlll te
unambig
the fam
lust Se
Defoe”
operam
"Nbep
and as 4
that thc
Primary

InStitut

 

translat

22
institutional environments.

The difference between this question in the context of
the small firm with a single owner-operator rather than in
the multi-person firm context is that for a small firm much
of the entire operation can and is overseen by one person,
thereby reducing many of the transaction and information
costs associated with more complex systems. Where the
larger firm and the small or large family farm are similar
is in the difficulties in clearly establishing the goals of
the operation. Perhaps to an extent even greater than that
for large and small non-family businesses, a family farm
will tend to have significant problems determining
unambiguous goals to guide the operation. The operator of
the family farm, while perhaps being the principle manager,
must generally deal with the desires and needs of other
members of the family since the performance of the
operation can have a profound effect on every family
member. Different members can have very different goals,
and as a result disagreements are common. Still, assuming
that these conflicts are resolved and have led to certain
primary goals for the farm, the comparatively less complex
institutional environment should permit an easier

translation of these primary goals into operational goals.

 

*2

the op.
I
small 1
occurre
Operat
goals.
enviror
Iespons
Organiz
0r
firm 1
have t:
0f SP61
single
a Comp
each
°°mmon
thQSe
needi:
Each t
OPErat
the in

aIEaS

23
3.2.2 Operational Goals and Functional Areas

The above reasoning leads to the conclusion that one
way to help discern the behavior of a firm is to identify
it's primary goals. But more is needed than knowledge of
the operator's primary goals to model the behavior of the
small firm since the reactions of the firm to day-to-day
occurrences are also a function of the way in which the
operator has organized the firm so as to attain the primary
goals. Different responses are necessary for different
environmental circumstances and the range of possible
responses will be, at least in part, a function of the
organizational structure of the firm.

One way to identify the organizational structure of a
firm is to ask which specialized functions will the firm
have to perform in order to meet the primary goals. A form
of specialization can be thought of as occurring in the
single owner farm operation. The modern farmer, operating in
a complex environment must perform several complex tasks,
each requiring specific skills. Behavioral theory and
common sense would suggest that the way to handle all of
these tasks is to separate them according to the functions
needing to be performed and, to a certain extent, to treat
each task or functional area as a unique and discrete
operation within the farm. Good management requires that
the important linkages between each of these functional

areas be established and more or less continuously

recogn:
behavi;

limitet

perfor:
differe
the ani

attivit

 

Iachine
acting

Ianager
analyst
soneho“
togetbe
priMar}.

nothing

24
recognized no matter what function is being performed. Yet
behavioral theory would also suggest that operators have a
limited and finite capacity to monitor complex linkages, so
that out of necessity as small a number of linkages would be
preferable.

For instance, in many ways the operations needing to be
performed in order to produce livestock rations are totally
different from those involved in the feeding and milking of
the animals, which are in turn very different from the
activities associated with the purchasing and financing of
machinery and buildings. In one activity the farmer is
acting as a crop producer, in the other as livestock
manager, and in the third as a combination investment
analyst and equipment procurer. The operator must be
somehow cognizant of how these disparate activities come
together to form a dairy operation that suits his or her
primary goals, yet in many ways each of the activities have
nothing to do with each other.

The small farm can therefore be viewed as a collection
of functional areas, each guided by operational goals, and
linked by information and or materials so as to achieve the
primary goals.

Such an organization permits a separation of the
various tasks or activities performed according to their
contribution to attaining these goals. This view of firms

is consistent with the behavioral view, which emphasizes

 

that f
multi-
unit h
each

goal I

.4

consci'
organi;
The ma:

perfor:

 

 

have t
attain:
Vill 3]
been c(
for ea:

farmer

25
that firms have a multiple set of goals. In the
multi-person firm context this translates into each sub-
unit having its own goal. For the small firm it means that
each separate area of tasks is guided by an operational
goal (or goals).

It is not necessary to assume the farmer in question
consciously decides to utilize the above described
organizational system in order for that system to arise.
The management of the operation will necessarily focus on
performing certain tasks every day and these tasks will
have to contribute, at least to some degree, to the
attainment of the primary goals of the farmer. The farmer
will also have to have some sense of when these tasks have
been completed. This "sense" implies the presence of goals
for each functional area which will be pursued, whether the
farmer has explicitly chosen such goals or adopted them for
some other reason (i.e. habit or tradition). Given that the
day-to-day operations will be organized around such goals
and that the functions of an operation consistently and
distinctly follow either a daily or seasonal pattern, it
seems reasonable to conceptually structure the behavior of
the farmer around these distinct functional areas and

goals.

 

SOP's. ,
to be a
frequer
operati
operati

activit

 

farmer
and goa
of the
Possibl
"lthin
50 that

A

feedba

inform

26
3.3 Other Applications of Behavioral Theory to Small Firms

Other components of behavioral theory can be applied in
the small farm theory. The first of these is the use of
SOP's. SOP's economize on the amount of thought that needs
to be applied to problems or situations that reoccur
frequently. A farmer could develop and use standard
operating procedures in each of the functional areas of the
operation as well as in the operation-wide management
activities. The standard operating procedures used by a
farmer would necessarily be related to the functional areas
and goals referred to above. The successful identification
of the standard operating procedures should make it
possible to know how the farmer has orientated himself both
within each functional area and within the total operation
so that the primary goals can be achieved.

As for the multi-person business, small farms need
feedback in order to operate successfully. Feedback is the
information about the conditions and situations that face
the farmer that signal the need for action or inaction.
Feedback is also the process by which learning occurs,
where the successes and failures of past actions indicate
to the decision maker the appropriateness of the standard
operating procedures in use. Of the information that the
small firm operator receives in response to different
standard operating procedure governed actions, he will have

to distinguish between the feedback implying the need for

action

 

standa:
thesis

conten‘

treatefi
|
inactit
standa:

lr

functit

 

functic
Operati
entire
suCces
implie
lead t

Opera:

instar
ofa

declsj
Perfo1
this I

degrel

27
action and the feedback implying a need for a change in the
standard operating procedures. No attempt is made in this
thesis to model the process of adaptation and change in the
content of the standard operating procedures. Feedback is
treated only in its role as the indicator for action or
inaction within the decision set bounded by the existing
standard operating procedures.

In the small firm, feedback can occur within each
functional area and it can occur across all of the
functional areas, indicating the performance of the entire
operation relative to the primary goals. Feedback on the
entire system's performance will indicate the failure or
success of the firm relative to the primary goal; failure
implies a problem that needs to be dealt with, and will
lead to changes in behavior that must take place within the
operational, organizational structure of the firm.

Feedback and the resulting responses cannot occur
instantaneously and simultaneously. In the actual operation
of a small or large firm a lag exists between the time of
decisions and the time of experiencing the associated
performance and the time when the decisions in response to
this performance are made. The length of the lag and the
degree of the farmer's recognition of the association
between past decisions and current performance determines
in part how well the operation performs. The decisions made

now are dependant on the decisions made in the past, due to

feedba
percei‘

decisi<

cognit
functi
focuse
firm b
Simple
ShOulc

EXpres

°P€rat
being
the

be St)
which
,

goals

tel-u

area.

28
feedback, so that the form in which past performance is
perceived determines to a certain degree the quality of the

decisions made today.

3.4 Summary of the Small Firm Theory

The behavioral theory of the firm, as used by Cyert and
March for large, multi-person operations, can be applied to
small farms given the theory's emphasis on individual
cognitive abilities and the resulting associated means for
functioning in a complex world. Behavioral firm theory
focuses on the role of goals as fundamental determinants of
firm behavior. If these goals can be determined, the
simpler institutional setting of the single operator farm
should make it easier to specify how these goals will be
expressed in the activities of the operation.

Goals can be divided into two categories, primary and
operational. Primary goals define the firm's reasons for
being in operation and may be used as guides in determining
the organizational structure of the farm. The farm will
be structured around relatively distinct functional areas
which contribute to the attainment of the primary goals.
These functional areas can be characterized by operational
goals and SOP's.

Along with the operational goals and SOP's, feedback
determines the decisions made in a particular functional

area. Feedback also ties the entire system together by

 

indica
goals.

will af

 

29
indicating overall performance, relative to the primary
goals. This feedback can be used to guide decisions which

will affect the entire system.

‘9'

1'85

the

bud

P’V'I
no
'1

fro
0f

8m

65%

\-

So:

CHAPTER FOUR
AV OR SYSTEMS SIMULATION MODE 0 HID- ICHIGAN

QALgY FARM

4.1 Introduction

A Mid-Michigan dairy operation was selected for this
research because of the ready availability of: 1, data for
the many operating sub—systems of dairy farms (including,
in particular, the data provided by the Michigan Telfarm
budgets ([4][5][16][21][22]); and 2, numerous knowledgeable
and experienced observers and practitioners that can be used
as references for specifying actual operating
characteristics of a dairy farm. Expert advice was drawn
from agricultural economists, dairy science and management

practitioners as well as dairy farmers.

No attempt is made here to create a "representative"
farm, as in the use of a programming model based on data
from some stratified sample of all farms [11]. The focus

of this model is instead to reasonably approximate the
environment and behavior of a farm based on the judgement of
experts; not a model of all farms but rather a model
approximating at least one farm. Neither is the model
created in this research intended to serve as a
representation of how a farm should operate, but again, just
as a reasonable representation of an actual farm.

Following the theoretical and heuristic guidelines
established in Chapter Three, the goals of the operation and

some reasonable sources of feedback are established first,

30

'1‘

Ope

ani

COU

355

C61-

Pay

0f

31
below. An appropriate set of sub-systems for the operation
are established as well. Based on these goals and sub-
systems, the manner in which the specified sources of
feedback will affect the guidance of the system is
discussed. All of this information is then summarized in
causal loop diagram models of the hypothesized operation,
representing both the entire operation as well as the
important sub-systems. The chapter concludes with a
description of the problems encountered in the attempt to

create the systems simulation model of the entire operation.

4.2 Primary Goals

The primary goals of the operation are assumed as
follows. The first is that the farmer desires to remain in
operation until retirement. The second is that the farmer
wants to milk approximately 55-60 cows with approximately
the same technology during this period. The rest of the
operation is scaled appropriately for this number of
animals. The farmer will consider milking more or less
cows, subject to maximum and minimum constraints, but the
farmer essentially operates a 60 cow operation.

In association with these two primary goals, a third is
assumed as well: the farmer in question needs to attain a
certain level of revenues above and beyond those needed to
pay for the operation. This level of revenue can be thought

of as family living expenses. The primary goal, the size

and ty
associ
seek a

bv an

 

expens

 

on pro
produc
be inc
of 11
Produc
of an)
maint,

SOUIC«

assum
00mp0
Brain
kinds

are S

the fé

0n 50:

32
and type of the operation selected by the farmer and the
associated costs and revenues, imply that the operator will
seek a level of revenues that are in excess of total costs
by an amount equal to or in excess of the family living

expenses .

4.2.1 Other Characteristics

For simplicity's sake, the emphasis of the operation is
on producing milk. Two other possible substantial revenue
producing activities, cash grain and livestock, are not to
be included in this operation. Any revenues from the sale
of livestock are not the result of intentional livestock
production for sale; revenues may be generated by the sale
of any replacement livestock in excess of that needed to
maintain the desired milking herd size. There are no
sources of income from off the farm.

Another characteristic of this operation that is
assumed for simplifying purposes is that only the roughage
component of the dairy ration is produced on the farm. All
grain concentrates are purchased off the farm. Only two
kinds of roughage are grown; haylage and corn silage. Both

are stored in above ground silos.

This characteristic of producing only the roughages on
the farm, and producing concentrates off-farm, is observable

on some farms in Michigan. But it is more common for dairy

opera:
off ce'
compli_
create
decisi
to be

in the

 

 

 

33
operations to be growing their own grain, if not for sale
off the farm then for on-farm consumption. The
complications and the associated problems that would be
created for a behavioral model of the joint economic
decisions of producing both grain and milk were considered
to be too great relative to the realism that would be lost
in the model by assuming only milk production.

These goals and characteristics are arbitrary, given
the range of types of operations found in central Michigan.
Yet many mid-Michigan operations are in fact similar to the
farm proposed above, both in the operating conditions and
the goals attributed to the hypothesized farmer. These
similarities permit the determination of reasonable

operating procedures based on expert opinion.

4.2.2 Feedback with Respect to the Primary Goals

The primary goal can be summarized as "remain in the
business of producing milk with a 50 to 60 milking cow
operation until the time of retirement, provided that
adequate family living revenues are generated". One piece
of information that can serve as an indicator of
performance relative to this goal is expected cash flow. If
the operation has the desired number of animals and all of
the associated necessary functions generating costs and
revenues are being performed, any shortfall in anticipated

net revenues indicates a need to change.

 

in Cha
involv
enviro
flow r
to use
neetin
desire
then t
short-
this

Succes

 

AlmOStl
Projec

°Perat

34

As indicated by the theory of the small firm developed
in Chapter Three, making the primary goals operational
involves combining different SOP's in a satisficing
environment. The day to day SOP of meeting short-term cash
flow needs can reasonably be considered to lead the farmer
to use cash flow position to determine if the farm is
meeting the primary goals. If the farm has the qualities
desired by the farmer and meets today's demands for cash,
then the primary goal has been met, at least in the
short-run. The task is then to create an SOP that extends
this daily success into the future. One indicator of future
success is projected cash flow over some planning horizon.
Almost all of the cash flow information needed to make this
projection is available from information about daily
operations.

This indicator is not proposed as the best way to
guide the system, but rather one indicator that could be
used given the implications of behavioral theory. Although
discussions with dairy experts did not conclusively point
towards this or any other primary performance indicator, the
general consensus was that such an indicator could
conceivably be used by a real farmer. The point of placing
this hypothetical dairy farm in a simulation model is in
part to determine if such an indicator could permit a farm

to succeed under real world circumstances.

needed
opera:
to the

making

interv
compo:

expend

 

35

If expected net cash flow falls short of the level
needed for family living expenses, a change in the
operation is needed. Defining the set of actions available
to the operator in such an instance is the last step in
making the primary goals operational. Two likely areas of
intervention are in the primary revenue generating
component, milk animals, and in the primary source of
expenditures, the feed ration.

The general consensus among the experts and
practitioners consulted is that farmers do respond to
relative changes in input and output prices. Unfortunately,
no one was willing or able to say conclusively if farmers
respond directly to prices, or through changes in some
other indicator like net cash flow, or how the response
actually occurs. Most were willing to concede that one of
the potential indicators that could be used by farmers to
monitor performance is cash flow.

One relatively common response to difficult financial
circumstances observed among farmers is to slow down the
rate at which input suppliers are repaid7. The practice
amounts to obtaining an unsecured loan and reduces
immediate cash flow pressures. Farmers may also, at times,
be willing to marginally alter the size of the herd
(housing and milking capacity permitting) in order to
improve cash flow position or take advantage of short-term

changes in prices.

N

changeF
feed 1
tender
such c
inves:
feed a
quanti
inerri

ail, -
behavi
t0 Che.
would
improv

r€515:

'EXper

the u
At fir
needir
deerea

Change

—- ————_—_—_———
—_

feedir
can an

98*wa

d

 

36

Many dairy farmers are reluctant to significantly
change the content as well as the relative mix of their
feed in response to changes in input or output prices. This
tendency would seem to be due to the uncertain effects of
such changes on output per cow, as well as the significant
investment of time and effort in learning how to handle and
feed a new ration. An existing SOP relative to the type and
quantity of ration appears to have a large degree of
inertia, meaning that it will change very slowly, if at
all, in response to standard changes in prices. Skinnerian
behavioral theory would say that in order for the ration SOP
to change in response to input price changes, the farmer
would have to consistently experience that such changes
improve net revenue, yet the other factors leading to
resistance against changing the SOP would tend to make the
”experiments" necessary for such experience unlikely.

The willingness to change herd size is in contrast to
the unwillingness to significantly change ration content.
At first glance, this makes sense; the time and effort
needing to be expended in order manage a herd increased or
decreased by five or ten percent is less than that to
change the specific mix of a ration and the associated
feeding practices for the entire herd. Presumably, farmers
can and do marginally modify the quantity and timing of an

established ration mix, and this behavior could be modeled.

f (j

m

(h
1‘
(T)

 

Payab1

affECt

(“3 S

e—J

diStit.

Sjv'sten

 

37

Potentially, all of the interventions discussed so far,
as well as others, such as delaying the purchase of
machinery, could be included in a simulation model of a
reasonable operation. In this proposed model, the farmer is
assumed to have only a gradually changing SOP governing
ration content with the changes limited to increases in the
total quantity of ration, as historically promoted by the
agricultural extension service and adopted by other
producers. Changes in ration type and mix may result from
relative changes in prices, but inclusion of such influences
do not appear necessary to simulate a reasonable operation.

The action set available to the farmer in response to
changes in expected cash flow is limited in this model to
marginal changes in the number of milking animals. The
actual change in herd size to be made by the farmer is
determined by the number of animals the farmer believes
would lead to the highest level of expected net cash flow.
The operator is assumed to handle short term cash flow
pressures by reducing the rate of repayment of accounts
payable, but the long-term direction of the farm is not

affected by this decision.

4.3 Subsystems and Operational Goals
The hypothetical dairy farm in this study has a set of
distinct functions that make up the operation of the total

system. Following the conceptual means for organizing a

pres

11

of‘.

.12

V;

man

and

Ma:

C01-

the

38
single operator firm given in Chapter Three, Table 4.1
presents the functional areas of the operation, hereafter
called subsystems, as well as the primary operational goals
of the subsystems.

Each of the proposed subsystems contributes to the
primary goals of maintaining the operation at a certain
size until retirement while maintaining a certain level of
cash for living expenses. As can be seen in Table 4.1, most
of the subsystems and their associated goals contribute
"operationally" to the primary goals, for example: herd
management maintains the desired number of milking animals
and their replacements in good health so that they can
produce milk, and thereby produce cash; labor management
attains and manages the level of labor needed to operate the

various subsystems of the farm and produce cash, given the
level of labor desired by the owner/operator; and, short
term cash management deals with the payment of accounts as
well as attaining and paying for any short-term loans so as
to permit the continuance of daily operations.

The linkages between these subsystems is either
materials, costs or information. Roughages produced or
concentrates purchased are physically passed to a ration
feeding Operation, which is then converted to milk, while
the costs of producing or purchasing the ration are

monitored and dealt with by the cash management subsystem.

Tab]
Ass:

1.x:
4
l r)
A

|:=

Her<

Hii

r—4
0)
(f

:c (I)
:T‘
o

inf

C05

39

Table 4.1: A Summary of the Farm's Subsystems and the
Associated Goals

 

SUBSYSTEM GOALS
Herd Management Ascertain and maintain the desired herd
size. Raise and or purchase

replacements.

Ration Production Produce the desired quantity of
roughage and concentrates.

Milk Production Determine and feed the appropriate
ration. Maintain adequate milk per cow.

Labor Management Hire the needed quantity of labor.
Maintain the quantity of operator labor
within the desired range.

Capital Purchases Purchase equipment and buildings as

needed.
Debt Management Pay-back long and intermediate term
debt. Alter repayment rate if excess

cash is present.

Short-Term Cash Manage cash to maintain a positive
Management balance.

 

The informational linkages between the subsystems
serves as the source of feedback. The central feedback
information is the relative levels of variable and fixed
costs and revenue which are used to determine the expected
cash flow. Feedback affects the system in two ways. The
first is within a subsystem as various decisions are made.
For instance, the number of replacements to maintain, or
the quantity of capital to obtain, or the amount to be paid

to input suppliers. The feedback on performance of the

subs

dere

SEC

poo

her

ah

9—0:
p)

‘74

1"

4O
subsystem is compared to the relevant operational goal to

determine the necessary response.

The second, over-riding effect of expected cash flow
sets much of the course for the system. If the response to
poor expected performance is limited to marginal changes in
herd size, the operation cannot change either too
drastically or quickly.

All of the materials, costs and information linkages
are discussed in more detail in the following section,
which presents causal loop diagrams of the entire dairy

farm system.

4.4 Causal Loop Diagrams of the Dairy Farm Model
4.4.1 Introduction

The causal loop diagram is a descriptive instrument
that can be used to depict the fundamental components of a
system as well as the causal interactions linking all of
these components. A causal loop diagram consists of three
things; the variables or components of the system
(represented by the circles), the linkages and direction of
causality between the components (the lines and arrows),
and the relationship between the causally linked variables,
defined as either positive or negative (+ or -). The system

changes over time as the variables simultaneously increase,

be?

die

[85

S ‘v'!

p01

he?

Sh"
.'

th

41
decrease or have no affect on the relative size of the
other variables.

The number of variables and interconnections in the
behavioral model of the entire farm make its causal 100p
diagram excessively large, detailed and difficult to either
read or explain. For this reason, the diagram of the entire
system is not presented. The causal loop diagrams of
portions of the dairy operation presented and discussed
below are much easier to interpret and understand while
still demonstrating the important aspects of the entire

system.

4.4.2 The Entire Dairy Operation

The model is organized around the functional subsystems
that operationally help the farmer reach his primary goal.
The primary goal is expressed in two variables (model
variables are identified in the text by all capital
letters), CASH NEEDED FOR FAMILY LIVING and DESIRED
OPERATION SIZE AND TYPE. Four relatively distinct
groupings of subsystems needed to reach the primary goal can
be identified in the model. The first is the policy setting
subset which examines expected cash flow relative to CASH
NEEDED FOR FAMILY LIVING to determine the desired number of
milking animals. The second grouping includes the
subsystems for the production of milk, herd management,

ration production and purchases, and the use of hired and

42
operator labor. The third grouping is the short-term
financial component, including variable costs of production,
farm revenues, and cash flow position which itself includes
accounts payable management and payments on long,
intermediate and short term debt. The final grouping is the
capital, debt and assets subset which provides an
accounting record of debts and liabilities for the dairy

farm.

4.4.3 Cash and Expected Cash Flow

A separation is made in this model between the current
cash flow position and the current expected cash flow. The
fundamental difference between these two variables is the
way in which accounts payable are treated. Figure 4.1
depicts the accounts payable loop.

VARIABLE COSTS, entered exogenously in this detailed
view, has a positive effect on ACCOUNTS PAYABLE, meaning
that as costs go up, so will ACCOUNTS PAYABLE; the converse
is also true. REVENUE has a similar effect on CASH. As
ACCOUNTS PAYABLE increases, so will the rate at which
accounts are paid out (ACCOUNTS PAYABLE OUT). This, in
turn will decrease the ACCOUNTS PAYABLE and CASH due to the
negative or inverse effects on these two variables.

ACCOUNTS PAYABLE and CASH also determine the level of CASH

FLOW PRESSURE experienced by the farmer. As CASH FLOW

43

goommgnoxoa mec3000<\efiétmom :moo Steeliu05m ;\e mujwfi

 

 

 

 

  
   

+.
®3C®>®m
: m o o I
+ I
mmejmmmkm/V I: ejmv
>> 0 E m o x O
mEmExom :mOo \ E a
\. mecsooo<
6mg +
\. / ....\\
/.+ m€moxoa
/.\ ®C_\</AH/._Li_nsim / mWrF—JOUQ/q >
fl E u m. . fliiii, + . m a m o Q ,
, _) \.. ® OCO
// #LCIU \\ _Q . >

 

PRESS!

BORROI

CASH v

its i:

increa

 

decrea
leads

PAYABL'
PAYABL
CASH F
DEBT P

thI'Oug

4.4.4

Ti

 

PIGSEn

CELL 5'
along ‘
REVEXt.
Figure
the 9X]
line

which ‘

44
PRESSURE increase beyond a certain level, SHORT TERM
BORROWING will have to occur. The result is an increase in
CASH which will decrease CASH FLOW PRESSURE.

Another balancing mechanism for CASH FLOW PRESSURE is
its impact on ACCOUNTS PAYABLE OUT. As CASH FLOW PRESSURE
increases, ACCOUNTS PAYABLE OUT will decrease, thereby
decreasing the rate of withdrawals from CASH which again
leads to a decrease in CASH FLOW PRESSURE. But as ACCOUNTS
PAYABLE OUT decreases, the rate of decline in ACCOUNTS
PAYABLE will also decrease, which will tend to maintain
CASH FLOW PRESSURE. SHORT TERM BORROWING will increase
DEBT PAYMENTS, which also increases CASH FLOW PRESSURE

through effects on CASH.

4.4.4 Revenues

The loop representing the generation of revenues is
presented in Figure 4.2. The STOCK OF ANIMALS and RATION
determine MILK PRODUCTION. The STOCK OF ANIMALS (along with
SOP's not depicted here) determines the SALE OF EXCESS OR
CULL STOCK. MILK PRODUCTION and the SALE OF CULL STOCK,
along with the associated prices, determine the level of
REVENUES. REVENUES (which feeds into the cash loop, see
Figure 4.1) feeds back onto the stock of animals through
the expected cash flow loop, depicted here by the dotted
line. The dotted line represents another set of variables

which characterize the expected cash flow process. The

45

 

 

Q00; mmjcm>mm e5 cozgmcmo ”was 839.1

    

  

>>O_|._

@300 IT :80 bofimgxm

     

+. +.
mmscm>mm ,

IT

‘~-—--—-a'

cozujboum

 

3:2 \
y: ....
+ /
//
//. \\
\> //.x.\
\ , \ EOE _ C x
_\ . . .
i C a; O ) ) .. i
r/ p. x / .FC XLQEAM O
/ x /.
/ \.\ / x

farme
calcu

CUTIE

herd
indic
or de
herd
flow
{event
needec
sizes
t0 he:
diffe:
become
the 11

indic.

i

i
i

indepe

the r

invol.

 

46
farmer is assumed to be making EXPECTED CASH FLOW
calculations for increases, decreases and no change in
current herd size..

Each EXPECTED CASH FLOW calculation for the different
herd sizes are ranked, with the highest dollar value
indicating the most advantageous herd size. The increases
or decreases are limited by an SOP that allows changes in
herd size of only five and ten percent. The expected cash
flow calculation is an extrapolation based on current
revenues per cow, variable expenses and DEBT PAYMENTS. The
needed quantity of hired labor changes for differing herd
sizes and the changes in hired labor costs from herd size
to herd size are the primary source of per cow cost
difference among the various calculations. This herd size
becomes the DESIRED MILKERS figure. The question mark at
the link between EXPECTED CASH FLOW and STOCK OF ANIMALS
indicates the dependency of the effect on the outcome of
the ranking.

The farmer's analysis of EXPECTED CASH FLOW is
independent of the ever-present problem of short-term cash
management. While dealing with short-term cash problems
involves changing the rate of payout on ACCOUNTS PAYABLE or
borrowing operating funds, EXPECTED CASH FLOW analysis uses
current information about costs and revenue to make a

strategic decision with an outlook beyond this month.

4.4.5

The S

PRODC

PRQDU

 

 

47
4.4.5 Variable Costs

The Variable Costs loop is presented in Figure 4.3.
The STOCK OF ANIMALS determines the level of RATION
PRODUCTION AND PURCHASES; these two variables lead to MILK
PRODUCTION. The three together determine the TOTAL LABOR
required for the farm. The STOCK OF ANIMALS also determines
the quantity of HIRED LABOR through an SOP not depicted
here. The TOTAL LABOR required minus the HIRED LABOR
determines the quantity of OPERATOR LABOR.

Five variables then contribute to the VARIABLE COSTS
level; STOCK OF ANIMALS (herd management costs), RATION
PRODUCTION AND PURCHASES, MILK PRODUCTION, HIRED LABOR and
all of the associated INPUT PRICES. As VARIABLE COSTS go
up, so will ACCOUNTS PAYABLE. VARIABLE COST increases will
also likely lead to decreases in EXPECTED CASH FLOW,
although this effect may be ameliorated by alterations in
expenses due to differing levels of HIRED LABOR. The loop
is closed by linking EXPECTED CASH FLOW to the STOCK OF
ANIMALS, with the sign of the causality being dependent on
comparison of relative advantages as explained for the

previous diagram.

48

 
  
   
   

—- \

\

hhred (Doeroior

Lobor LGbOV//
-+\\ .//

V

A~
, Rohoo \a
P {’0 Ci U «'11: i i 0 fl \

 

Vorkfifle
(hosts

Accounts

Poyobke
//

 

Figure 4.3: Vorioble Costs Loop

4.4.

Debt
LONG

of t

SL111

r v
‘\

Pre
flo
det
the

0n

49
4.4.6 Capital, Debt and Assets

The final subsystem to be depicted is the Capital,
Debt and Assets Loop, as shown in Figure 4.4. Purchases of
LONG AND INTERMEDIATE TERM CAPITAL occur at the inception
of the operation and when the items wear out. Borrowing
occurs to finance the purchase of the items, leading to the
principle and interest payments which constitute LONG AND
INTERMEDIATE TERM DEBT. SHORT TERM BORROWING occurs when
there are significant CASH FLOW PRESSURES (see Figure 4.1),
which also leads to the accumulation of SHORT TERM DEBT.
Both of these sources of debt combine to form DEBT
PAYMENTS, which are withdrawals from CASH. ASSETS is the
sum of the monetary values of CASH, CAPITAL and STOCK OF
ANIMALS.

The ratio of debts to assets serves in this model as
the principle indicator of credit worthiness. While this
ratio is commonly used by the credit industry to judge
credit worthiness for short and intermediate term loans,
long term loans (for buildings and land) are often judged
on a more comprehensive basis (such as cash flow analysis
to insure sufficient cash for repayments, or the net
present value of the investment in conjunction with cash
flow analysis). In numerous cases, this quantity of
detailed financial information is not readily available so
that lenders often resort to an enumeration of the assets

on the farm to indicate the quantity of collateral

50

A

  

mzzezc<
eo

xooim

 
 

 

Qooo Wfiom< occipoa

 
 

axiom
momm<
oi fiooe

\
IT
\ 38 /
ELQF
@wFLUmiCtCmior:
CCU @COJ /

.5 \\\WT \\V\1//z\\

\\ _otooo \\ ioog

FC ii @F CL c._ WP
®#O_U®FCL®#C_ #LOLW [T

<\

///HWMHMWWWH\\\

.2558

 

Hexv oujmrm

IT
mixitxoo
iooQ
IT
‘mézioueom II
rcuoe
iuogm

prese
world

the I

thres
occur
the

is re
conti
conti

in ch

4.5

 

51

presents. For the sake of simplicity and given the real
world use of collateral information for long term debts,
the DEBT TO ASSET RATIO is used to judge all three debts.

As the DEBT TO ASSET RATIO increases beyond a certain
threshold, less borrowing of all three types of CAPITAL can
occur. The availability of credit decreases continuously as
the ratio increases beyond this threshold until the point
is reached where no borrowing can occur. DEBT PAYMENTS will
continue to be withdrawn from CASH which will lead to a
continuation of CASH FLOW PRESSURES, leading to a reduction

in the ACCOUNTS PAYABLE OUT rate (see Figure 4.1).

4.5 Summary

The expected cash flow calculation is a source of
feedback which directs the course of the operation with the
primary goal being the level of CASH NEEDED FOR FAMILY
LIVING expenses and the DESIRED TYPE AND SIZE OPERATION.
The causal loop diagrams of various components of the
entire system are presented to detail the primary
operations on the farm. Behavior in the cash management
subsystem is significantly affected by the experience of
cash flow pressures. The availability of credit for long
and intermediate term capital as well as short term
operating funds is dependent on the debt to asset ratio.
This ratio serves as an indicator of ability to meet

repayment responsibilities.

b.5

comp:
theor
detai
subsy

the m

 

52

4.5 The Systems Simulation Model of this Hypothetical
Operation

The simulation model of this behavioral model was
constructed using the system dynamics language STELLA, a
package designed for use on APPLE's MacIntosh personal
computer. The simulation model is derived directly from the
theory of the single-owner firm and the behavioral model
detailed above. The operation is organized around the
subsystems that permit the farm to operate day to day, with
the most dynamic sources of feedback being short-term cash
availability, expected cash flow, and the effects of labor
requirements.

The model is not completed. The problems with the model
are too numerous and the model too detailed to permit its
completion within the time constraints of this thesis.
Table 4.2 is a summary of the subsystems in the models as
well as the number and type of variables in each subsystem.
(Definitions for the types of variables in a system
dynamic's model are given in Appendix One.) Both the
total number of variables and the number of state variables
indicate that this is a relatively large model. In
comparison, Meadow's Commodity Cycle model for hogs has
approximately one-sixth the number of state variables and
one-eighth the total number of variables ([20], a

description of Meadow's model is given in Chapter 5).

S—

uh

RHHRLSVHCAu

CL E - T

To

53

Table 4.2: The number of variables, by type, in the dairy
farm simulation model, as compared to Meadow's model. (a/)

 

 

Subsystem State Rate Au 1 iary Total
Herd Management ..... 6 10 39 55
Ration production ... 2 2 l3 17
Milk Production ..... O 0 3 3
Revenues and Costs .. O O 28 7
Land ................ 0 O 7 7
Short-Term Cash

Management ......... 7 ll 11 29
Capital, Debt,

Assets ............. 12 18 27 57
Labor .............. 2 2 18 22
Expected Cash Flow.. 2 2 32 36
Total Per Type .... 31 45 178 254

Total Per Type in
Meadow's Model .... 5 5 23 33

 

The fact that Meadow's model is so much smaller than
the dairy farm model is partly a reflection of Meadow's
skill as a systems' modeler. Yet, this dairy model was
intended to explicitly capture a more complex reaction and
decision process, and the size of the model is a reflection
of this extra degree of complexity.

Major difficulties remain in obtaining a defensible
specification of some crucial variables in the dairy model.
Independent of the likely objections that some researchers
may have to the use of expected cash flow as the primary
source of feedback, and marginal changes in herd size as the

principle method of strategic management in the model,

54
problems remain in the herd management, short-term cash
management and labor management subsystems. In the herd
management subsystem, the current specification of the
culling decisions of young cows and mature cows likely
causes a misrepresentation of the relative mix of young and
mature milkers on a dairy operation. More complex culling
decision processes need to be incorporated.

In short-term cash management, a better specification
is needed for the impacts of changes in cash flow pressures
on the rate at which accounts are paid out. Difficulties
remain in the labor subsystem, primarily in the SOP's
governing the hiring of labor and the total labor
requirements for the dairy operation. These, and other
specifications and technical problems need to be resolved
before a working model of an entire farm can be completed.

As discussed in Chapter One, these problems were
initially not considered major enough problems to prevent
the completion of the model within this research's time
constraints. At the time the problems were first
encountered attention was turned to writing up completed
portions of the research. The writing process as well as
the experience of building the model led to more informed
and better questions, which led to the discovery in the
literature of partial answers to the primary research
questions. The literature explored and the conclusions

drawn from it as well as the experience of constructing the

55

above simulation model are presented in Chapter Five.

re
to
co
us
it:
re
Th
he
an
be

vi

de
d:\'
an

CI“.

in

the

inc

APTE FIV
H 0 ND IC N

BEHAVIORAL SIMULATION MODELS

5.1 Introduction

Models, economic or otherwise, are caricatures of a
real world situation or system, and as such are always going
to be ”unreal" in significant respects. The issue of
concern is whether or not the model is real enough to be of
use in its applications to or exploring the problem of
interest. Deciding upon the degree of reality that a model
requires is then a crucial step in assessing its quality.
This decision requires some classifactory scheme which can
be of help in determining how the models should be judged
and by what criterion. A two-way scheme for viewing
behavioral simulation models is offered below as one way for
viewing behavioral simulation models.

With these considerations in mind, the literature
describing some of the important behavioral and systems
dynamics models is reviewed. The literature provides some
answers to the primary research questions developed in
Chapter One. The literature, and the attempt undertaken in
this research to model a Michigan dairy operation (detailed
in Chapter Four) also indicates ways in which behavioral
theory, system dynamics and neo-classical microeconomic
theory may be jointly used in market-level analysis. These
and other conclusions are discussed at the end of this

chapter.

56

57
5.2 Judging the Validity of Simulation Models

Although other classification schemes are certainly
possible, behavioral simulation models of real and unique
markets and firms can be considered as belonging to one of
two classes: 1, a theoretical model with many of the same
properties relevant to validity and specification of any
other theoretical model; or 2, an attempt to forecast with
some degree of precision the performance of the system in
question. But the distinction between theoretical model
and forecasting model becomes blurred when simulation
techniques are used. This is because any simulation model
of the real world is essentially a theory of the market or
firm, so that a model intended to simulate or forecast
actual performance has some of the same characteristics as
the theoretical models.

Given this somewhat blurred distinction between these
two classes of models, an issue arises as to how to judge
the validity and appropriateness of the model in question,
particularly when the model belongs mostly to the
theoretical class, rather than the more quantitatively
precise representational class. This issue becomes somewhat
slightly more difficult to resolve, in theory, when computer
simulation is used to evaluate the implications of the

model's structure and assumptions.

tiv
soc
net
the
sir
qu:

us:

mo:
1a1
CO]

SO!

av;

0f
5554
the(

L0 h

58

Computer simulation of theoretical models is a rela-
tively new entrant to the field of tools available to the
social science investigator. The fact that the computer
needs and uses quantitatively measured variables to execute
theoretical models often leads to the assumption that the
simulation results need to be judged on some absolute
quantitative basis much like the way we check the math we
use to balance our checkbook. Either the calculated results
are correct or they are not.

But the fact is that computer simulation of theoretical
models is no more than another system of symbols or
language for expressing the set of relationships or
conditions that the theoretician wishes to explore. For
some sets of relationships and conditions the computer
simulation language is perhaps the only suitable means
available for the exploration.

Cyert and March view computer simulation as a
"...fourth language in which the assumptions of a theory
can be expressed and its conclusions derived" [9, page
315]. The other languages available are ordinary prose,
pictorial geometry and mathematics.

Although simulation models can serve in many roles, one
of Cyert and March's primary uses of computer simulation is
essentially as a tool for the building of behavioral
theories of firms. The models constructed are not intended

to be used for forecasting, but rather they are used to

inv

SEC

thc

St?

0
r4)

is

EX

of

EX

pa

mo

IE

no

in

59
investigate the consequences, in theory, of a particular
set of assumptions and hypotheses.

Given this distinction, the issue of obtaining a
properly identified system of variables so as to permit
unbiased parameter estimation, is not necessarily an issue
with respect to judging theoretical simulation models. In
the application of assumptions to create theoretical
structures, the primary issue is the logic and consistency
of the model, as well as its linkages to our sense of what
is a reasonable representation of the real world.

There are no absolute tests of the logic, consistency
and the degree of correspondence between theoretical models
and our own sense of what is reasonable. These are more
properly decided on the basis of informed argument amongst
experts. In other words, an evaluation of the specification
of any theoretical model, no matter what language it is
expressed in, is a matter of debate amongst students of the
subject matter. When we judge a researcher's description of
a thought experiment they performed to try and solve a
particular problem we don't ask if the parameters in the
model are properly estimated. We ask if the conditions and
approximate magnitudes of relationships make sense and are
relevant. In many ways a theoretical simulation model is
nothing more than a complicated thought experiment conducted

in a cybernetic environment.

60

The issue of unbiased parameter estimation does appear
when these theoretical models are adapted or applied to
actual systems or markets in an attempt to recreate the
actual performance of the system in question. One of the
commonly used means for judging the performance of any
simulation model, behavioral or otherwise, is to compare
model output for some historical period to that performance
actually observed. While a model that tracks history well
is considerably more credible than one that does not, there
is no guarantee that the individual parameters in such a
model are necessarily correct. Models can mimic a system's
past performance without actually having important
relationships in the system properly specified and
quantified, meaning that the test of workability is not
interchangeable with the statistical/mathematical conditions
and tests for bias.

But theoretical simulation models are not necessarily
seeking to achieve a representation of an actual system. The
implications derived from the solving or execution of any
theoretical model, whether simulation based or mathematical,
are not forecasts, and as a result theoretical analytical
models should not be judged by the criterion applied to
forecasting models. Simulation models could be more
effectively constructed and their validity judged if the
distinction between theoretical and forecasting models is

kept clearly in mind when the models are built as well as

61

evaluated by outside observers.

5.3 Some Literature on Behavioral Simulation and a General
Theory of Markets

One of the stated purposes of researchers using
simulation models (see below, Cyert and March, Lyneis, and
Meadows) is to develop a set of assumptions into a theory
that is broadly applicable to many situations, as in the
way the neo-classical model is used as a theory for all
markets. As a test of the general model, the basic
structure developed is applied to a real world system.

Leibenstein believes that the Carnegie approach (for
present purposes, the system dynamics models are placed in
this same class) really serve as models of a specific and
unique environment and that they are not general enough to
become a truly coherent micro-micro theory. He cites others
who have come to the same conclusion [18, page 483]. If he
is correct on this point, it is not necessarily a
condemnation of the Carnegie class of models, just a call
for more care to be used when identifying the exact
applications where these models will be the most useful.

The ability of any of the behavioral approaches to
achieve a general theory of markets, including the work of
Leibenstein, appears to be severely limited by the present
state of human behavioral theory. The models proposed will
only be as universal as the universality of the assumed

behavior. As long as we disagree over what constitutes

62

universal or general market behavior the behavioral models
would seem to be best used in the analysis of specific
markets or firms, where the relevant behavior is more
likely to be defensibly specified.

Even if the behavioral applications are not able to
achieve general market results, it remains possible that a
behavioral model of a specific market could be used as a

tool in sector level policy analysis and policy making.

5.4 The Literature of Behavioral Simulation Models
5.4.1 Cyert and March; A Duopoly Model

A general model is constructed by Cyert and March [9]
which attempts to describe the price and output behavior of
firms in an oligopoly. The properties of this model are
also explored in the context of a duopoly.

The basic model deals with three sets of decisions and
their interactions with each other through feedback. The
three decision areas are output, price and sales.

The output decision is based on a smoothed inventory
and production goal. The proposed level of needed inventory
is set by this unit within certain limits defining
excessive or inadequate levels. Proposed production is then
the forecast for sales plus the difference between proposed
and current inventories. Any changes in production must be
within some experientially determined upper and lower

limit. The inventory and the production decision are

63
therefore inter-related although the goals are set
separately.

The price set by the firm is a result of the following:
a relatively complex process involving a comparison between
actual profits achieved in the last period, given the
price, and the profit goal; any changes in the costs of
manufacturing the good; and any changes in the long-run
price behavior of competitors. The pricing mechanism is
analogous to that used by manufacturers in Balderston and
Hoggatt's model (see below).

The sales decision is not treated in as much detail as
the previous two. In response to changes in performance of
various units relative to their goals, pressure is applied
to change the level of sales. Sales pressure can result if
there is a decrease in organizational slack in the firm as
a whole, or a direct result of inadequate profits or market
share. The sales unit is also able to exert pressure on
price setting within the firm, with any serious changes in
the price position of the firm relative to competitors'
being immediately corrected (within the imposed profit goal
/ price limits). Undesired changes in sales level or market
share has a more lagged effect on price.

The marketplace within which all firms interact is also
defined in the model. Market demand is a function of the
weighted average of all firms' prices, sales effectiveness

pressures, sales promotion expenditures and an exogenous

64
factor. The firm's market share is a function of relative
sales effectiveness pressure, relative price level,
relative promotion expenditures and lagged market share.

In the two-firm simulation model application of the
above general model, a multiple, step-wise regression
analysis is applied to the results from several simulations
to attempt to determine the greatest sources of impacts in
the system. Four dependent variables (firm price,
inventory, market share and profit) are made a potential
function of twenty-five others - the final equations
contain approximately eleven variables.

Examples of independent variables seen to consistently
have a strong effect are as follows:

1) The proportional increase permitted in production
(the upper bound).

2) The extent of price response to changes in
competitor's price.

3) The required markup to maintain profitability.

4) Propensity to modify price in reaction to failure on
the profit goal.

5) That firms with adequate slack in their operation,
which permitted an improvement in operations when the firm
experienced input cost changes, tended to be the lower
priced firms in the market.

6) Ability to modify the sales goal relative to current

profits.

65

It is not clear why the multiple regression method of
analysis is used. Aside from the serious problems with the
validity of the results of step-wise regression, it seems
that unless the four equations are estimated as a
simultaneous system, biased and inconsistent coefficient
estimates should result. As in the discussion of Balderston
and Hoggatt (below), the lack of concrete sensitivity
analysis here may be a function of the lack of appropriate
software. In either case, it may have been more reasonable
to perform ANOVA analysis to obtain the type of results
desired by the authors.

It would seem that the analysis used demonstrates in a
non-definitive sense that the system of assumptions and
theory yields performance that could be expected in an
economic system. These results also demonstrate that the
model is "viable" in the sense used by Balderston and

Hoggatt (see below).

5.4.2 Balderston and Hoggatt; A Decentralized Market System
A computer simulation model was created by Balderston
and Hoggatt [l] [2] to permit the analysis of a system of
relationships that they believed are too complex for
standard mathematical analysis. They are attempting to
explore the dynamics of a market system, composed of
different sets of economic actors and linked together by

relatively complex behavior.

66

The market system to be simulated is the West Coast
lumber industry. The industry is seen to be composed of
three classes of vertically linked firms. Goods move from
manufacturers, through wholesalers, to retailers. The
wholesalers act as the intermediaries between the
manufacturers and retailers; information about supply and
demand meets at the wholesale level. Cash flows from
retailers and wholesalers to manufacturers.

The primary goal of the research is:

”... to show how limits on market information,
decentralization of market decisions, and institutional
alignments affect and are affected by economic forces." [2,
page 183].

To achieve this goal, the authors use an eclectic approach
to representing the firms in this system as well as the
operations of the market. The authors feel that their
approach is new. In their own words:

"We had no general theory of market processes to serve
as a precedent for applications; rather, we attempted to use
the occasion of a single, empirically based study for
exploring the potential domain of such a general theory."
[2, page 183].

Although the primary goal of the study is to provide general
analytical results, the authors recognize that the claims of
generality are subject to dispute, based on "...
consideration of the conceptual elements of the model and
of the way in which they have been linked together". [2,
page 183].

In the simulation model, the three sets of economic

actors trade amongst themselves without a centralized

67
trading area. Trade partners are discovered through a
decentralized search for suitable price and output

offerings.

The empirical study of the lumber industry found the
distribution of firms in the industry to be roughly two
manufacturers to one wholesaler to three retailers. The
number of firms in each of these classes is limited in the
model to a maximum of thirty to fifteen to sixty, using the
same proportion as that found in the real market.

Manufacturers have two primary functions; produce goods
and set offer prices. The quantity of goods produced by any
one manufacturer is determined by the quantity of output
needed to return their inventories to some desired level,
relative to expected sales. Expected sales are a four
market period moving average of past sales.

The price offered by each manufacturer is not based on
profit maximization. Maximization would require knowledge of
how own changes in price would affect total demand as well
as the offer prices of other manufacturers. The pricing
rules are also not based on attaining (satisficing) some
aspiration based profit goal, since such a decision also
requires the capability to predict the effects of own price
changes .on relative profitability. The authors decide to
establish a price setting rule which is based on feedback

from the effects of past pricing decisions on profits. In

68
an indirect way, this price setting rule attains a certain
type of satisficing result. The feedback from most recent
experience is given the most weight.

The retailers have three primary functions; establish
the retail price of the good, determine the price which
they are willing to pay for goods, and determine the
quantity of goods they would like to purchases.

The retailers set a retail price that will result in
profit maximization. The retail firms face a downward
sloping demand curve for their final product. The retail
firms are not able to affect demand through sales promotion
so that the demand curve is kept constant throughout the
simulation. The retail firms are monopolistic competitors,
due to local isolation.

Retailers set bid prices in essentially the same manner
as that used by the manufacturers, based on experience of
past bids and the associated profit,

The quantity of goods the retailers would like to
purchase is that needed to return their inventories to some
desired level relative to expected sales. Expectations for
future sales are generated in the same way they are
generated for manufacturers.

Wholesalers do not carry any inventory of goods. They
serve solely as a facilitator of trade between potential
trading partners. The wholesaler searches among

manufacturers and retailers for close matches in offer and

69
bid prices. Once potential partners are discovered the
wholesaler communicates with the two potential partners to
try and confirm the sale. The sale will be confirmed if the
offer and bid prices of the potential traders are
approximately the same.

Since one of the goals of the research is to examine
the effects of the costs of information on the performance
of the system, the model is simulated with exogenously set
levels of message costs for each communication between
wholesaler and potential trader. The unit message cost
varies from zero to some arbitrarily set level.

The range of trading firms that the wholesalers will
search is set exogenously from run to run - the two
possible states being either all retailers and manufacturers
explored in a random fashion or a subset of all potential
firms. The content of the subset is determined by past
experience with the cost of achieving an agreement between
potential partners. The subset will tend to be made up of
retailers and manufacturers who, in the past, the
wholesaler was able to bring together for the lowest
relative cost.

A simulation is divided up into "market periods", with
each period divided up into a "market cycle”. At the start
of each market period the manufacturers and retailers all
set their desired prices and the quantities they would like

to trade. Each market cycle consists of the search for

70
potential trading partners. Cycles continue until all
demand for the period is met. One market period is
equivalent to one month with each simulation lasting up to
sixty months.

The performance of each firm is measured with a
financial balance sheet. Firms that are bankrupt at the end
of the market period are forced out of business and if the
average profits of all firms in a particular class are
above some threshold level, new firms in this class are
permitted to enter. Firms also can change size as a
simulation progresses as they respond to increases or
decreases in the quantity of goods they are trading and the
quantity of working capital available to them.

The authors state that one of the first tests of the
"correctness" of the model is to test its "viability”, a
term which in a limited sense is synonymous with the
commonly used system dynamics concept of model validation.
A model is considered viable, in the authors' use, simply
if the model can be simulated. The assumption seems to be
that given that their focus has been to realistically
represent portions of the system, if the assembled portions
continue to produce and trade output through a simulation
run, then the model has a certain degree of viability (and

validity).

71

The step of systematically examining the simulation
results to determine if the simulated system actually
demonstrates behavior expected from the real system (as
used in system dynamics, see below) is not explicitly
taken. In their defense, though, the authors report on
numerous runs of the model with an eye towards spotting
results that would be inconsistent with their knowledge of
the real system.

The model is used as an experimental environment for
examining the effect of changing certain crucial variables
in the system. Several runs are performed with each run
varying only in the unit message cost of communication and
the type of preference ordering set used by the wholesaler
(a random set or one determined by experience). Under these
different conditions, the authors attempt to measure the
degree of market efficiency, the degree of match or
mismatch between supply and demand and the effects of the
system on the size distribution of firms.

Efficiency is defined here as the relationship of
potential wholesaler revenues at the start of the market
period to those revenues actually received. At the start of
each market period the authors construct an aggregate
manufacturer's supply curve, given their inventories and
offer prices. A similar aggregate demand curve is
constructed for the retailers. These two schedules imply a

certain level of potential revenues (call it Al) for

72
wholesalers if they were to match suppliers and retailers
with a minimum number of communications. At the end of the
same market period, actual wholesaler revenue (A2) is
calculated. The market efficiency measure is e - Al / A2.
The more efficient the market, the closer e is to one, the
less efficient, the closer e is to zero.

Wholesalers do a relatively efficient job. The
calculated average values for different simulations range
from .81 to .92, with the simulations differing on the size
of the unit message cost as well as the type of wholesale
preference set.

The authors also examine the ability of the market to
match supply with demand. Using the same start of market
period supply and demand schedules calculated above, it is
possible to ask if the potential quantity to be supplied by
manufacturers is greater, less than or equal to that
demanded by retailers. Supply is found to be in excess in
all types of simulations with excess being greater under a
random preference set compared to the experiential
preference set.

The authors find that the operations of the system
result in a skewed distribution of the size of firms within
their respective classes, even though all firms within each
class are given the exact same characteristics at the start
of each run. No firm differs according to entrepreneurial

ability or resource availability so that the skewness

73
results purely from the operations of the market. The
higher the unit message cost the greater the degree of
skewness. More skewness also occurs when the wholesalers's
search for trading partners is guided by past experience of
low cost agreements rather than a random survey of all
potential traders.

The authors do not state this, but one of the
implications of these results is that there is a tradeoff
between the advantages of the random preference set, which
results in a less skewed distribution of firm sizes, and
the fact that the random preferences persistently yield a
greater level of excess supply.

In a separate report on this research published in 1962
[2], the authors discuss the current methodological
limitations imposed on the analysis of the causal
mechanisms in simulation models. The authors use their own
statistical indicators of correlation between variables
(such as non-parametric tests) in the model, but they are
not able to perform the kind of sensitivity analysis that
would lead to reasonably conclusive proof of causality.
This limitation appears to be more a function of the
software that the authors had available for model execution
rather than the degree of complexity of the system. Today's
software may permit adequate sensitivity analysis of this

model.

74
In conclusion then, Balderston and Hoggatt have built a
simulation model using non-optimizing goals in a standard
operating procedure dominated behavioral setting. It
appears that their research has shown that such a model
can replicate the functions and economic performance of a

single firm as well as those of an entire market.

5.4.3 Summary of the Implications of these Studies

The two models described above, created by Cyert and
March, and Balderston and Hoggatt represent some of the
potential for behavioral simulation modeling of firms.
Without providing what may be considered irrefutable
evidence, it seems that their models have represented the
operations of firms without a singular profit goal, but
rather with many goals and without maximization or perfect
information. The question remains, can this approach be
used for sector analysis?

The answer would seem to be a strongly conditional yes.
Both models could be used to evaluate different policies,
such as the effects of changing input and output price
levels, or permitting more instability in these prices.
Balderston and Hoggatt's model also could provide a means
for examining many other significant questions. For
example, how would the system perform if there is a
difference in the unit message cost between wholesalers? The

same question might be what if some wholesalers had better

75
access to information about the trading position of
manufacturers and retailers? Or, how would the system
perform if the price discovery process was changed to a
centralized auction market?

The potential for using this approach for sector
analysis must be questioned because of the great quantities
of time and expense that are required to construct these
models. Balderston and Hoggatt estimate that the dollar
cost of their research is approximately $96,000.00, or
$350,000.00 in 1985 dollars (roughly seven man years of
work) [1, pages 280—281]. The cost of their research
includes the two years spent by one of the authors in doing
the empirical study of the industry, which became his PhD
dissertation.

One of the reasons for this expense, which is also the
source of the primary constraint upon using the behavioral
approach, is that the "general" theory is not readily
applicable to specific markets. Each time a model is built
of a specific firm or market system, large costs will be
incurred when the empirical evidence is gathered and the
model specified. Every system specific model is a new
theory and building this theory is a time consuming
process. This expense can be contrasted with the relative
ease that the neo-classical theory of firms and markets can
be applied to a specific situation with aggregate data.

There would appear to be a tradeoff between the richly

76
complex questions that can be asked of the simulation
models described above and the associated expense of the
modeling effort. It would be interesting to compare the
costs of the Balderston and Hoggatt study with those
incurred for a similar neo-classical regional programming
model, such as the 1963 study of the distribution of dairy
and related resources in the Lake States region [11].

Given that the state of technology for simulation model
construction has improved significantly since the time of
the behavioral studies described above, it seems reasonable
to assume that the cost of projects similar to either of
the above models, undertaken today, could be greatly
lowered.

Before exploring in more detail the question of this
methods applicability to sector analysis it would seem
instructive to ask why has computer simulation been used so
little to explore and expand the theory of the firm. A
review of the publications in the major economic journals
for the past seven years reveals no research of the type
presented above. This is in contrast to the statements by
some that the advent of computer simulation would spawn a
great deal of theory building about the inner workings of
the firm (see, [7] and [9, Appendix B]).

Naylor and Vernon [23, pages 463-464] present four
reasons for the lack of applications of simulation to the

development of firm theory. First, such models have a

77
strong empirical base resulting in the same difficulties
outlined above. Second, many more fields than just
economics are necessary to adequately develop these models.
The potential theorist, in this mode, must have a grasp of
decision theory and~theories of group behavior, in addition
to working knowledge of the contributions of such
disciplines as psychology, sociology, political science and
business administration. Studies requiring such a broad
cross discipline approach are tremendously difficult to
undertake and complete.

The third criticism is that not enough attention has
been given to developing effective experimental design in
computer simulation applications. The authors state that
“economists have had little or nothing to say about the
problem of designing simulation experiments with models of
the firm" and for this reason, the results "have proven to
be inconclusive and difficult to interpret". Any
generalizations made on the basis of such models are viewed
with skepticism on the part of economists, in sum resulting
in "questions about the usefulness of computer models of
the firm" [23, page 464]. It is not entirely clear how to
take this criticism. In part, the source of the confusion
about the validity of the technique may be the result of the
causes outlined earlier in this chapter; theoretical models
are different from forecasting models and that the two

types of models can be easily confused in the simulation

78
environment.

The confusion may also be due to the extensive
complexity present in many firm models, leading to
difficulties in sensitivity analysis and model validation.
This type of confusion may be avoidable if researchers
establish clearer purposes for the models and use theory
that is constrained to a degree of complexity that will
permit adequate sensitivity analysis, given the capabilities
of the software.

Naylor and Vernon's final point is that simulation
models have been applied to ”cases where more traditional
methods of analysis may have been more appropriate" [23,
page 464]. Their point is made in reference to sector level
problems. It seems clear that traditional methods of
analysis could not have achieved the type of results given
by Balderston and Hoggatt or those of Cyert and March.
Still, Naylor and Vernon cite the work of Machlup and others
which makes a case for when maximization of goals in a
partial equilibrium environment can achieve satisfactory
results [23, page 464].

There very well seem to be instances where Machlup's
point is true, even in the context of simulation models.
There are other possibilities for using a behavioral
approach in a simulation environment that incorporates
elements of maximizing behavior, without incurring the same

vast costs of the Balderston and Hoggatt's study. Naylor

79

and Vernon's final point will be discussed in more detail
at the end of this chapter, after examples of these
alternative behavioral models of firms have been explored.

The alternative modeling approach examined below is an
application of the system dynamics approach, as developed by
Forrester [l3] and others at MIT. The system dynamics
approach, as summarized by Lyneis, is as follows:

- identification of specific problem behaviors

- construction of a model of the structure and
policies creating the behavior

- developing an understanding of how the structure
and policies create the behavior

- designing policies that improve behavior

- test alternative policy responses under
different plausible scenarios [19, page xiv]

The focus of the approach is on the interaction of the
behavior of the system's actors with the structure of the
system. A computer simulation model is used to perform this

analysis.

5.4.4 Meadows; A Dynamic Commodity Cycle Model

Meadows research goal is to examine commodity systems
and attempt to model the regular cyclical fluctuations
commonly found in these systems. The primary objectives are
to ”... derive a general dynamics model of the structure
underlying long-term commodity production cycles, to
validate that model, and to determine it's implications for

the design of commodity stabilization policy." [20, page

80
2]. In the system dynamics approach, "underlying
structures" are seen as resulting from the real world lags
imposed on materials, information and the adjustment of
behavior.

Although Meadows does not state it as such, the
economic actors in the model are implicitly assumed to
behave in an essentially neo-classical fashion. Meadows
uses standard assumptions concerning the shape and slope of
both the firm level supply and the consumer demand curves
as well as assuming that the actors are rationally acting
to maximize a utility function dominated by economic
returns [20, chapter four].

In an attempt to create a generalizable analytic
framework, a non-commodity specific model is constructed
which emphasizes structural relationships common to many
commodity systems. System behavior is considered to be
rooted in changes in various "state" or "stock" variables
which accumulate or decline over time. The major state
variables in the general commodity system are production
capacity, commodity inventories at the wholesale level and
the price expected by producers.

Production capacity is causally linked to capacity
desired on the part of producers, which is itself causally
linked to their price expectations for both the commodity
and the major inputs to the production process. Inventories

are linked to the amount of capacity being utilized by

81
producers and per capita consumption of the commodity. A
biologically or technologically determined delay exits
between production capacity utilized and the completion of
the production process. Per capita consumption is a
function of commodity price which is determined in the
short-run by wholesalers who set prices in response to the
movements of actual inventories relative to some desired
level. Expected price at the producer level is a weighted
average of the actual commodity price. Expected price then
closes the system's loop by providing feedback to producers
as they generate productive capacity.

Meadows specifies values for the variables in this
structure using published price elasticities and
appropriate biological and technical delays for hogs,
cattle and broilers. Three commodity specific models are
therefore created. The simulations of each of these models
resulted in cyclic production patterns with periods and
magnitudes similar to those actually observed for all three
commodities.

Sensitivity analysis of the hog model is performed.
Some of the areas examined are changes in supply and demand
elasticities and the level of inventory coverage desired by
wholesalers.

Making the supply elasticity of producers more elastic
results in increased instability in productive capacity.

Meadows does not analyze the source of this instability but

82

it is likely due not only to the obvious reason, the
increased response rate of producers, but also to the
effect of such responses on the movement of more pork
through the system, causing more price instability.
Ultimately, this instability feeds back into desired
productive capacity. (An examination of the relative
movement of all variables in the system is possible given
the nature of the software used to construct the model. It
is not clear to me why Meadows did not explore these loops
in more detail, although he may have not seen this as
necessary to achieve the broader goals of the study).

Keeping the original supply elasticity and reducing
the demand elasticity for pork results in even more
unstable productive capacity than that seen for changes in
the supply elasticity.

Meadows also lowers the inventory level coverage
desired by wholesalers by 50 %, which in essence makes the
retail price more sensitive to short-term changes in supply
or demand. Productive capacity also becomes extremely
unstable under these conditions. Meadows notes that any
commodity stabilization policy intended to work at the
producer level will likely be thwarted if the same policy
results in reduced desired inventory coverage [20, page

76].

83
Meadow's own conclusions are that sensitivity analysis
can be used in these models to reveal potential areas of
focus for policy intervention. The implication is that for
certain classes of commodities, the general commodity
production cycle model could be readily adapted for use as

a policy analysis instrument.

5.4.5 Lyneis; Corporate Planning and Policy Design

Lyneis [l9] examines the potential for using a system
dynamics model of a corporation as an aid to corporate
planning and policy design.

Lyneis does not use a behavioral/organizational theory
of firm operations that focuses on the interaction of goals
within the firm, as used by Cyert and March. The
corporation is viewed as one unit with one primary
objective, although the objective 'used in this case is not
profit maximization but rather producing enough output to
satisfy demand. Viewing the corporation as one unit also
implies that the decisions made by the corporation to
achieve this objective are made without any of the
difficulties foreseen by Cyert and March. The use of lags
in the perception of changes in the environment as well as
lags in decision making time in Lyneis' model may serve as
a valid and straightforward approximation of the decision

making process within the corporation.

84

The study is divided into three parts: 1, a general
overview of the applicability of system dynamics to
corporate problems; 2, development of a basic and general
corporate simulation model; and 3, applications of the
model, with variations, to specific problem areas.

Lyneis places the general corporation to be modeled in
a monopolistically competitive market. The corporation's
output is heterogeneous enough to permit it to be
distinguished from the other competitors in the market. The
competitors remain constant in their policies, so that
changes in this corporation's behavior does not result in
an alteration of the other firms' basic operations.

The general model of the corporation is based on, as in
Meadows, the significant stock variables in the system. The
model is essentially an inventory model with many of the
auxiliary variables being determined by the information
implied by the relative levels of the inventory states.

The customer order rate, an exogenous or endogenous
variable, depending on the run, sets in motion a process
that is linked eventually to finished inventory for sale.
The customer order rate signals to the corporation the
quantity of finished inventory needed. This information
compared to the actual finished inventory determines the
needed production rate. The corporation treats the needed
production rate as its current objective. The production of

finished goods does not occur instantaneously but rather it

85
takes a certain amount of time determined by the
manufacturing process as well as the constraints imposed by
the available stock of inputs to the production process.

The stock of inputs is determined in nearly the same
fashion as the finished inventory. The current average of
the production rate (the averaging process implicitly
involves a delay in perception) determines the
corporation's needed parts inventory. Parts are ordered if
there is a shortage, with a time delay occurring between
the time of ordering and parts arrival.

This basic model is used to demonstrate how the
decision rules related to the perception of and response to
needed finished inventory and needed parts inventory are
dominant dynamic links that can create instability or
inefficiencies in this simple system.

In part three of the study the model is applied with
some modifications to a number of corporate problems, such
as:

1) the use of trend analysis of demand
information;

2) interactions with company suppliers;

3) interactions with customers and competitors;

4) interactions with labor;

5) the dynamics of financial control;

6) the dynamics of capacity expansion;

7) the dynamics of professional resource expansion.

A brief discussion of the third problem area is given below

86
to demonstrate application of the model.

In order to permit the examination of the problems
surrounding interactions with customers and competitors, the
customer order rate is made partially endogenous to the
performance of the firm. The customer order rate responds to
changes in two variables: 1, the length of the delay between
the time the customer orders finished inventory and the
time it is delivered (delivery delay); and 2, changes in
the sale price of the finished good.

The delivery delay occurs because the firm does not
anticipate changes in customer order rate. The unexpected
demand causes product completions to lag behind desired
production. The firm responds by increasing the production
rate, which is itself held back by the limitations of
available inputs to production.

Two of the scenarios run by Lyneis should indicate the
model's usefulness. One scenario examines the reactions of
the firm when the customer order rate is linked to delivery
delay. The other to the firm's performance with both a
delivery delay and price response. The firm is forced to
react to the shock of an increase in the customer order
rate, as would occur if a major competitor was suddenly
unable to supply goods to consumers.

In the first set of runs, the lags imposed by the
firm's inventories and response time endogenously causes

the customer order rate to fall below the level before the

87
shock occurred. The production lags result eventually in a
decrease in market share.

The firm could alter the policies governing the
responses to such demand changes and hopefully retain more
market share. The firm could either increase the rate at
which it recognizes a change in demand, thereby increasing
the rate at which it changes desired production or it could
carry more inventory. The first policy change would still
result in the lag due to the shortage in parts; the second
means tying up more capital in inventory than perhaps the
firm really would like. The trade-offs between maintaining
market share and maintaining liquidity are therefore able
to be examined in scenarios of this type.

In the third scenario, with both a delivery delay and
price response, the system becomes unstable, with the
degree of instability a function of the speed with which
the firm alters it's price due to inventory coverage
changes and the elasticity of demand given to consumers.

With demand changes being linked to both delivery delay
and price, the model permits an evaluation of the tradeoffs
between the costs of carrying inventory, the costs of lost
market share and the costs of instability.

One potential shortcoming of Lyneis' approach to
examining the performance of the firm is that the firm's
competitors are permitted to respond to changes in demand

as if they were not constrained by any structural and

88
behavioral lags. It seems more realistic to expect that
unexpected shocks would be felt by all competitors,
creating a dynamic interaction between all firms and the
customers. The observed changes in market share would likely
be considerably different under these market conditions. The
market model created by Balderston and Hoggatt (see above)
placed firms in such a context and their approach may be
applicable here.

The work of Lyneis is similar to that of Meadows' in
that they are both system dynamics approaches; they focus
on the impacts of the stock variables in the system, which
for many businesses means inventories. But, the work of
Lyneis is also significantly different from Meadows. Lyneis
places much more emphasis on the actual operating
procedures within the firm and does not give the firm a
profit maximizing goal.

No discussion is given in Lyneis of the relationship of
the operating structures used in this research and those of
the real world (Meadows explores this relationship to his
model). The lack of such discussion makes it hard to gauge
the validity of the theory behind this model. If, though,
the theory is relatively valid, then it seems reasonable to
conclude that the standard operating procedures used in
this context do permit a good representation of the

performance of a real firm.

89
5.5 Conclusions: Simulation and Sector Analysis
Cyert and March made an explicit statement in their
book concerning how they perceived their behavioral theory
of the firm in relation to neo-classical theory. They

state:

"The (neo-classical) theory of the firm, which is
primarily a theory of markets, purports to explain at a
general level the way resources are allocated by a price
system. To the extent that the model does this successfully,
its gross assumptions will be justified. However, there are
a number of important and interesting questions relating
specifically to firm behavior that the theory cannot answer
and was never developed to answer... Thus, many of the
attacks on the theory of the firm are not so much proper
critiques of existing theory as they are suggestions for the
development of a theory appropriate to a different set of
questions. Ultimately, a new theory of firm decision-making
behavior might be used as a basis for a theory of markets,
but at least in the short-run we should distinguish between
a theory of micro-behavior, on the one hand, and the micro-
assumptions appropriate to a theory of aggregate economic
behavior on the other. In the present volume we will argue
that we have developed the rudiments of a reasonable
theory of firm decision making. Our arguments for the
utility of the theory in developing a theory of markets are,
at least at the moment, more modest." [9, pages 15-16]

The same general conclusion is reached by Cohen and
Cyert in 1975, when they state that for a certain class of
problems in aggregate sectors, the micro-assumptions of
neoclassical theory may be perfectly acceptable ([7, page
51] as cited by Leibenstein [18, page 482]). Simon also
seems to leave open the possibility for the adequacy of the
neo-classical model when he states that:

"Their (economic models) veridicality and usefulness
cannot be judged from the fact that they satisfy, formally,

the Subjective Expected Utility assumptions. In evaluating
them, it is critical to know how close the postulated

90
utilities and future events match those of the real world."
[28, page 16].

Leibenstein senses a procedural difficulty here, noting
that "If a theory will not predict the behavior of a firm,
can it predict the behavior of a collection of firms?" [18,
page 482].

This question deserves a great deal more discussion
than will be pursued here, but it seems reasonable to
conclude that when all firms are examined in an aggregate,
static sense, the shape of the implied supply and demand
curves do approximate, at least to a rough degree, those
expected from neo-classical theory (even the behavioral
models presented above incorporate in a behavioral sense,
relative price information and price response). This price
responsiveness will be present, it appears, even if decision
makers are not maximizing and may in fact be satisficing.
This point does not deal with the issue of how much of a
degree of rough approximation of real firm behavior will be
necessary to permit the neo-classical results to be
accepted with confidence.

In the context of the current discussion of sector

analysis, given the above remarks, the following points can
be made:
1) There is a difference between a general theory of

markets, a general theory of firm behavior and a set of

theoretical approaches to firms and markets that can be of

91
significant help in constructing simulation models of
specific, unique firms and or market systems. This last
distinction makes the usage of theory analogous to the
application of principles learned in a discipline such as
engineering, where no single method covers all of the
designs needing to be constructed.
2) Even though Cyert and March's theory is not intended to
replace the neo-classical theory of markets, they have used
their theory to develop a simulation model of a
hypothesized two firm market. They have represented a firm
with their theory and they have placed it in a market
governed by similar principles. The primary limitation of
their approach is the degree of complexity in their model,
making it extremely difficult to construct and difficult to
analyze. These problems make their method of limited use to
other researchers in sector analysis.
3) Lyneis presents a system dynamics model of corporate
behavior that is less complex than Cyert and March's but may
in fact capture as much of the structure needed to analyze
the behavior and performance of corporations. The greatest
limitation of Lyneis' model is that it places its
corporation in an environment of firms that do not reflect
the same operational realities, making the results of
firm's interactions in markets of questionable relevance.
4) Balderston and Hoggatt's model essentially deals with

this limitation by thoroughly re-creating the Operations and

92
relationships of an entire market. Their model seems to be
potentially the most fruitful for sector analysis, although
their model may be of such great detail and expense that
their approach also holds little promise for practical
research. It is not possible to determine by a priori
reasoning exactly how much of this expense could be avoided
today with the use of better software and hardware, but a
similar project could be expected to cost less.
5) It would be possible to place a collection of Lyneis
type firms in a Balderston and Hoggatt environment. This
approach may be just as complicated as the Balderston and
Hoggatt exercise and therefore subject to the same
limitations. A model of this sort would be of use for
sector analysis.
6) Meadows creates a model of an agricultural commodity
sector that relies primarily on neo-classical assumptions of
firm behavior, but places the actors in an environment
which reflects material, biological and informational
delays to replicate the cycle in productive capacity and
prices experienced in this sector. The sector level policy
applications seem clear if one accepts the validity of the
model; the relative effects of price stabilization policies
can be ascertained, given the different goals and
structural constraints of the different actors in the

system.

93

7) A synthesis of Meadow's approach, which is an
aggregative system dynamics model using some aspects of
neo-classical assumptions, with the Carnegie emphasis on
the micro-behavior of firms, may be able to take advantage
of the relative ease of application of a Meadows model and
the real world benefits of behavioral theory. The synthesis
could take the form of establishing a sector or sub-sector
model using the general system dynamics approach of Meadows
and then imbedding in this model, at crucial points,
sub-models incorporating more explicit behavioral theories.
The crucial points may be those observed in the
Meadows-type model to be of singular importance to the
operation of the sector, given the policy question at hand.

In comparison to all of the modeling efforts reviewed,
the last suggestion given above appears to hold the most
promise of utilizing behavioral theories of the firm in a
simulation environment without spending a vast amount of
time, effort and money. The system dynamics emphasis on
identifying the problem behavior and including only as much
structure as needed to capture the relevant behavior
greatly reduces the scope and complexity of the models.

Such a model could be conceived along lines similar to
those used by Meadows; making use of the neo-classical
assumptions for simplicities sake where relevant and then
focusing in on a specific node in the system, applying more

explicit behavioral theories. The research effort could be

94
divided into two broad stages, the system dynamics stage,
including sensitivity analysis searching for sources of the
undesired performance, and the behavioral modeling stage if
it is found that the model could be improved by such an
application. This model seems more consistent with the

goals of a PHD dissertation and cannot be undertaken in

this current research.

5.6 Answers to Research Questions

The first two research questions from Chapter One can
now be answered with the information learned up to this
point. The first research question asks if firms can be
reasonably characterized in simulation models with
non-maximizing goals pursued with SOP's. Although the
results of the simulation models reviewed above are not
conclusive, they provide strong evidence that behavioral
simulation models can be used to represent firms in this
manner. The second research question asks if there is a
potential for using the behavioral view of the firm to
perform sector analysis. The literature and the modelling
.effort undertaken here indicates that the detailed
behavioral simulation models are costly and time consuming
and may be useful for sector policy analysis only if a
large number of researchers and resources can be committed.
Still, the technique can be seen as holding promise for

less costly research if used in combination with other

95
theories, such as neo-classical market theory.

The third research question is, can a system dynamics
modeling method be used to create behavioral simulation
model of firms and by association contribute to the
analysis of sectors? This question remains unanswered
because my own model was never satisfactorily completed in
STELLA, although a model of the entire firm does not appear
to be impossible to achieve. Although the entire firm model
was not completed, the behavioral models of significant
portions of the dairy farm do indicate that it is a
satisfactory method for representing important areas of
farm operations and could possibly be used to solve farm
level management problems. Chapter Six and Seven presents
one of the dairy farm's subsystems, the herd management

model and is evidence for this conclusion.

QHAETER §IK
H E AN CE I AT OD

6.1 Introduction

The reason for constructing and presenting this herd
management simulation model is significantly different from
the original reason for the design and construction of the
entire farm simulation model. The entire farm model was
intended to help answer the primary question of the thesis;
can behavioral simulation models reasonably represent the
functions of an operation, without maximizing behavior and
perfect information? This question was answered in large
part by the literature (see the conclusions to the previous
chapter). One of the other primary questions being asked in
this thesis (research question three, from Chapter One)
concerns the ability of the system dynamics software STELLA
to be used to design and construct behavioral models for
use in firm level and sector analysis.

Although it was found to be provisionally possible in
Chapter Five to perform firm level and sector analysis with
behavioral models, it could not be determined whether STELLA
can be used to build a behavioral model of an entire farm.
This is because a behavioral model of the hypothetical
dairy operation was never completed in STELLA; in part,
this result implies that the answer to the question would
be no. In fact, this question remains unanswered given that
the skill level and time constraints of this researcher
contributed to the failure to obtain a working model.

96

97

The design, construction and simulation of this herd
management model cannot be used to answer research question
three. This is because the herd management model does not
incorporate enough dynamic economic behavior to permit an
extension of its simulation results to conclusions about
entire farms. The goals of building and simulating the
herd management model are less ambitious:

1) Can STELLA be used to model a specific part of the
dairy operation?

2) Can this model rely on the use of SOP's to guide the
operation?

3) Could such a model, which focuses on specific

problems in the herd management area, be of possible use to
agricultural economists?

6.2 Model Overview

The primary goals of this herd management subsystem are
to maintain a sufficiently healthy population of milking
animals and replacements while pursuing a program of
selection of animals for replacements that will tend to
improve the quality of the milking herd. The goal of herd
improvement could be pursued by either using selective
breeding and culling of replacements on the farm or by
purchasing high quality animals at any stage of maturity.

The herd in this model is divided into four age groups.
The rationale behind the divisions stems from
considerations of the different stages of herd management

and the associated change in management problems. The

98

stages tend to represent ages of animals that are treated
in a manner different from all of the other aged animals in
the herd. The general differences tend to reflect death
rates, feeding rates and differences in culling decisions.

The four main age groups are: 1), calves, from age zero
to four months; 2), heifers, from age five to twenty—seven
months; 3), young cows, which enter this stage upon their
first freshening at approximately 28 months, and; 4), cows,
which enter this stage when the second birth occurs.

A general description of each of these age divisions
and the types of decisions being made is given below. A
complete listing of the equations representing this
subsystem along with a detailed description of each of the

equations follows.

6.3 Calves

Calves enter the herd as the stock of animals of
adequate age give birth. A good rule of thumb is that a
mature cow will give birth once every thirteen monthsg,
although this average can be improved if the farmer's
breeding program is effective. The first four months of
life are the most crucial for the replacement animal since
this is the time they are the most susceptible to diseases
that can lead to death [12]. Death losses in this age group
can be as high as forty-seven percent, but a commonly used

figure is fourteen percent [12].

99

Given the focus of the farm on producing milk and not
livestock for sale, only the animals needed as replacements
for milkers are kept on the farm. All surviving male calves
are sold. All surviving females in excess of the number
needed to replace the loss of mature animals due to death
and the standard cull rate are also sold.

Calves remain in this stage for four months, by which
time they have passed the critical point for susceptibility
to disease. At the end of four months, calves are either

sold or move into the heifer stage.

6.4 Heifers

Heifers are animals from age five to twenty-seven
months, which is the age that on average, the cow will
first give birth and begin milkinglo. Animals are held in
this stage in this model to simplify the simulation of the
decision to cull a young animal due to milking or
reproduction problems (see the detailed description of the
model equations, below, for further discussion of this
separation). The decision to cull these animals can only
occur once they have given birth and have started to milk,
hence the separation of the young animals into non-freshened
and freshened animals. Heifers are subject to death loss
during this stage, with average losses being seven percent a

year for all animals from age five months and olderll.

100

6.5 Young Cows

These are the cows which have freshened for the first
time and have begun their first lactation. The operator is
assumed to examine these animals, looking for reproduction
problems or poor milking characteristics. The assumption is
that, on average, five percent of all young cows are culled
each year due to these "standard" problems. Young cows are
therefore removed from this stage for three reasons; death
losses, culling due to standard problems, and the
maturation of the surviving and remaining animals, reaching
the mature cow stage. The mature cow stage is assumed to
be reached when the animals freshen for the second time, at
approximately the thirty-seventh month. Purchases of young
cows are permitted to occur in some of the simulations of
the model. Purchases occur when there are fewer milkers than

the number of desired milkers.

6.6 Cows

This is the final stage for milking animals on the
dairy farm. An animal that has reached this stage is
producing an acceptable quantity of milk and would
generally not be culled until it has begun to pass its peak
in milk output. The exact age at which this decline in
output begins varies from animal to animal, but on average

the acceptable milking animal is culled from a Michigan

101
herd at the age of sixty monthslz. This rule of thumb is
used for most simulations of this model, although the
effects of lengthening the average life of the milk cow are

explored in some simulations.

6.7 Desired Milkers

The operational goal of this subsystem is represented
by the variable, Desired Milkers. This variable's
specification depends on the type of behavior and causality
actually occurring on the farm. Two specifications are used
to examine the consequences of different types of behavior
and feedback loops.

The first specification involves an exogenous
determination of the value of the variable, implying that
the farmer reacts to information other than the performance
of the herd management sector in setting the operational
goal. The exogenous specification can be changed in the
midst of a simulation, as in a scenario where the farmer is
assumed to desire a rapid expansion of the operation from
one quantity of milkers to another. The second
specification involves an exogenous component, as in the
scenario given above, but also links feedback on the
ability of the operation to meet the desired milker goal
relative to the actual size of the goal. Figure 6.1
presents a causal loop diagram of the operation and

emphasizes the potential importance of the presence or lack

102

 
    

Deotn
Loss
COWGS

,///flu‘\\\\

hIoturoUcm]‘\

Rots For

 
 

Birth
Rote
+

++ _ /\:l/

+:® Colves for
Repmcenmyu
Iwotunahon
Rote For

hhafers

  

 

\ ,
\~.‘ /'7

Young \ ~q
1/ / /“‘“‘\\

Cowsto V; ,*

’ '1 n .. a -.

Herd // // hurehOS:
" roung

/ fl
// \ Lows

\ /’ \\
\
c

hkafers
for ’//,,l;\\
/Rexflocernent ,/ Denna \
X\\r—’/// / Loss ]
//i. Non ;
,,/”// > C r] | .3 3;} ‘/

     
   

+
+ Cows
, //”' ‘“\\,
CuHed / . .\
— Deswed \
M I i i E? I”

Cop /

+ \ ,1 I

+ 7 \\\ // , \
‘” e \
I
I

 
 
 
 

/Deswed‘ //,,i~\\
NMH<€FS /// \*_ —
600i /' PEI'If'QI'I'I’IOI'ICB ‘
P - PPM] I‘ve: i
runwory \\* to Deemed 1/

   

iv] I i i' 13 FE I,/"

Cool \A
\\\\‘~#’///

  

FIgure 6ft Porfiol<33uso|Ltmm>£hogn3nicfi Herd '~3mrI IIIHI

Vth ond WHHKMH on Endogenous Desued MHhLIiJHUi

103
of such feedback.

As shown in the diagram, the size of the discrepancy
between the actual number of milkers and the desired number
is the Desired Milker Gap. First, assume that the loop
connecting this gap to the Desired Milker goal is not
present (the two arrows marked with a "*"), so that the goal
is determined exogenously. In this case, if the number of
Milkers falls short of the goal, purchases of young animals
will occur, tending to bring the system back to an
equilibrium as Milkers approaches and finally meets the
Desired Milker goal. The behavior of the system given this
type of structure is examined in the first set of
simulations, in Chapter Seven.

Now assume that the link between the gap in the number
of Milkers (Performance Relative to Desired Milkers) and
the Desired Milker goal is present, thereby making this
variable a function of exogenous and endogenous influences.
The hypothesis is that the response of many farmers to not
being able to meet their Desired Milker goal will lead to
an increase in the size of the goal, perhaps thinking that
if they have more milkers and therefore more births the
system will be better able to produce replacements and
maintain the desired number of milkers. Such behavior would
be similar to the reaction of some operators to poor
profits where the response is often to increase the size of

the operation rather than improve the quality of

104
management13 [29]. As the Desired Milker goal goes up so
will the purchases of young milkers, thereby increasing the
number of milkers, but also the number of Cows Culled. The
net result of endogenizing such behavior is examined in

another set of simulations, below.

6.8 Cash Flow

The decision to expand the herd, even if only
marginally, may be constrained in the short-run in some
cases by the availability of cash. In the model of the
entire herd, described in Chapter Four, cash flow was seen
principally as a means to monitor performance relative to
the primary strategic goal of the operation. The assumption
was made in the entire herd model that cash would never be
a constraint on the decision to marginally expand the herd.
This assumption was made because of the historical ability
of dairy farmers to obtain cash - primarily due to their
monthly source of income and the willingness of creditors
to lend short-term funds. Although there are times when
these funds would not be available, this problem is not
considered serious enough to warrant its inclusion in this

herd management model.

105

6.9 Detailed Description of the Herd Management Sub-Model

The following is a list of the equations in the
subsystem, organized around the major state variables in
the age group. Text follows most equations giving a verbal
explanation of the reasoning used for equation
construction. Appendix Two contains a system dynamics
diagram of the herd management simulation model and may be
of help in analysis of the structural logic of the model. A
list of all of the equations in STELLA format can also be
found there as well. Appendix One contains a brief review
of the notation used in both the system dynamics equations

and flow diagrams.

(l) Calves
(1.1) L Calves.K - Ca1ves.J + DT * (Calves In.JK -
Calves Out.JK)
(calves)
(1.2) R Calves In.KL - Birth Rate * Cows.K + (Young
Cows To Herd.KL + Heifers For Replacement.KL)
(calves per month)

(1.3) C Birth Rate - .0769
(fraction per cow per month)

The birth rate is essentially a function of the number
of animals mature enough to bear young and the rate at
which conception occurs. A good rule of thumb is that a cow
will give birth once every thirteen months. This basic
length of time will decrease or increase, depending on the
specific birth management capacities of the farmer, but as

a general guideline, thirteen months seems reasonable. For

106
this reason, the birth rate is simply 1/13 times the
number of cows plus the number of heifers and young cows
ready to freshen. Heifers become young cows when they
freshen for the first time and young cows become cows when
they freshen for the second time. Counting the number of
freshening heifers and young cows is therefore the same as
counting the number of calves born (these variables are
defined in detail below). Some seasonality should be
expected in the birth rate, but is not considered
significant enough for the purposes of this model to be
incorporated.
(1.4) R Calves Out.KL - Ca1ves.K * (Death Loss Calves
+ Maturation Rate Calves)
(calves per month)

The number of young calves leaving this stage is a
simple function of the number of calves in the state
variable times the sum of rates at which calves die and
mature. Some of these surviving and maturing animals are
bull calves, and are sold (see below), while the surviving

and maturing females are either sold or are entered into

the heifer stage of the herd (see below).

(1.5) C Death Loss Calves - .14 per four months or
.035
(fraction per month)
The rate of calf death loss will vary from farm to farm

depending on the conditions and the management skills of the

farmer [12], but the generally used rule of thumb is 14 to

107
15 % of all calves will die due to disease by the time they
reach four months of agela. If a 14% loss is assumed, then

each month 3.5% of the animals die.

(1.6) C Maturation Rate Calves - .25
(fraction per month)
The calf maturation rate is simply one every four

months.

(1.7) A Calves Needed.K - Desired Milkers.K * Steady
State Removal Rate
(calves per month)
This is the number of heifer calves needed to maintain
the desired herd size assuming that no purchases will be
made of ready to freshen or freshened animals. It is equal

to the steady state removal rate of mature animals from the

herd times the number of desired milkers.

(1.8) C Steady State Removal Rate - (Death Loss Cows
+ Cull Rate Standard Cows + (Expansion and
Expected Death Loss Factor)
(fraction per month)

(1.9) C Expansion and Death Loss Factor - .017
(fraction per month)
1.8 is calculated by the standard cull rate of milking
animals, plus 5% for potential herd expansion and 15 % for
death loss between calf to milker on an annual basis;

.20/12 - monthly rate.

 

108
(1.10) A Calves For Replacement.K - Calves 0ut.JK *
(l - Death Loss Calves)
(calves per month)
0f the calves moving out of the Calf state, some are

lost due to death. This equation defines the proportion

available for entering the next stage of maturation.

(1.11) A Sale Bull Calves.K - Calves For Replacement.K
* Bull Calf Rate
(male calves per month)

(1.12) C Bull Calf Rate - .50
(fraction males per calf per month)

The sale of bull calves is equal to one half of the

surviving calves. All bull calves are assumed to be sold.

(1.13) A Sale Heifer Calves.K -
If Desired Milker Gap.K > 0
Then 0
Else If Calves Needed.K > Calves For
Replacement.K - Sale Bull Calves.K
Then 0
Else (Calves For Replacement.K - Sale Bull
Calves.K) - Calves Needed.K
(calves per month)

The sale of heifer calves is more complicated than the
sale of bull calves, since most of the animals needed to
maintain the desired milking herd size are generated from
within the farm. The number of heifer calves sold is equal
to the surviving and maturing animals minus the number of
bull calves sold and minus the number of calves needed to

enter the herd in order to maintain a steady state milking

11erd size (neither growing or declining). The fact is that

 

109
there will rarely, if ever, be any excess calves if there
is a desire to expand the herd. For this reason, if the
Desired Milker Cap is greater than zero, then no heifer
calves are sold. Also, if there are fewer heifer calves
available than those needed then none will be sold.

Desired Milkers is either an exogenous or endogenous
variable, depending on the behavior being analyzed.
Assumptions can be made about the farmer's desired size of
operation, and given this goal, the performance of the rest
of the model can be observed. Endogenous effects are
limited to reactions to the degree of attainment of the
desired level of milkers, which is assumed to potentially
lead to the desire to further changes in the herd size.
Consult Chapter Seven for details on how this variable is

treated in various simulations.

(2) Heifers

(2.1) L Heifers.K - Heifers.J + DT * (Heifers In.JK -
Heifers Out.JK)
(heifers)

The stock of heifers, from age 5 to 27 months, at which

point they freshen.

(2.2) R Heifers In.KL - Young Heifers Available.K
(heifers per month)

(2.3) A Young Heifers Available.K - Calves For
Replacement.K - Sale Bull Calves.K — Sale
Heifer Ca1ves.K
(heifers per month)

110

The rate at which heifers accumulate; equal to the
number of maturing and surviving calves minus the number of
calves sold plus the number of young heifers purchased (if

any).

(2.4) R Heifers Out.KL - Heifers.K * (Death Loss
Heifers + Maturation Rate Heifers)
(heifers per month)
The rate at which heifers decline; equal to the number
of heifers times the sum of the rate at which they mature

and the rate of death loss.

(2.5) C Maturation Rate Heifers - .0435
(fraction per month)

The maturation rate for heifers. They remain in
this state for 23 months, or one heifer leaves every 23

months, or .0435.

(2.6) C Death Loss Heifers - .0058
(fraction per month)

Heifers die at roughly 7% per year, or .0058 of

the animals per monthls.

(2.7) A Heifers For Replacement.K - Heifers Out.KL *
(1 - Death Loss)

(heifers per month)

111
The number of heifers moving out is actually the
survivors. The death loss is subtracted from one to
determine the proportion of heifers moving out that are

surviving.

(3) Young Cows
(3.1) L Young Cows.K - Young Cows.J + DT * (Young
Cows In.JK - Young Cows Culled. JK - Young
Cows To Herd.JK
(young cows)

The number of young cows (first freshened animals who
are giving milk) is a function of many variables, one of
which is the purchases of young cows (a state variable
treated separately, below). The milking herd is divided
into two categories to permit the model to more easily
account for the different management decisions made by the
farmer relative to the age of the milking animal.

One of the primary herd management assumptions is that
cows are generally culled once they reach the age of 60
months, so that animals between their first freshening and
60 mOnths are the milk producers. The consequences of
extending the life span beyond 60 months is investigated in
some simulations of the model, but the value of 60 is
assumed to be the baseline or equilibrium value. If there
is a desire to expand the herd, then the farmer is assumed
to prefer to maintain the SOP of culling older animals and

expand by either retaining more young cows or purchasing

young cows. In the case of there being a desire to contract

 

112
the milking herd size, the farmer is assumed to prefer to
cull more older (and presumably poorer milk producing) cows
and maintain a constant rate of young animal retention. Of
course there will be young cows who are terrible producers
and some 60 month cows who are tremendous producers and the
farmer would be expected to cull differently under these
conditions, but such quality details are beyond the scope
of this research. In general, the assumptions made about
the farmers decisions seem to be the most likely to
contribute to an increase of herd average rate of milk per
cow16.

Another reason for separating the two ages of animals
relates to the retention decision of young animals itself.
This decision is one of the most crucial for herd
improvement. It is generally assumed to be nearly
impossible to decide on the future quality ( breeding and
milking characteristics ) of a heifer. It is necessary to
wait until a heifer has carried for her first term and/or
begun to produce milk that a determination can be made of
the animals quality and the likelihood of her retention in
the milking herd17.

Although in this model every milking animal is given
the same breeding and milk producing characteristics,
allowance is made for some young animals to be of inferior
quality (see Cull Rate Standard Young Cow below). The

separation of cows from young cows and the presence of the

 

113
quality variable will permit the introduction of a
stochastic quality element if it is deemed necessary in
future research (a RANDOM designation is available in

STELLA).

(3.2) R Young Cows In.KL - Heifers For Replacement +
Young Cows Arriving.KL
' (young cows per month)

The rate at which young cows are accumulating; equal to
the number of maturing heifers and the purchases of young
cows. Heifers For Replacement is defined above in eq.
(2.8).

Purchasing younger animals permits the farmer to

continue to use the SOP governing the culling of mature

milkers.

(3.3) L Young Cow Orders.K - Young Cows Orders.J +
DT * (Young Cow Orders In.JK - Young Cow
Arrivals.JK)

(young cows)

The state variable defining the quantity of young
animals the farmer is in the process of purchasing. The
assumption is that it takes time to perceive the need to
purchase the animals and it takes time for the animals on

order to arrive.

(3.3a) R Young Cow Orders In.KL - Young Cow Purchases
Needed.K / Time to Perceive Purchase Need
(young cows per month)

114
The rate at which orders accumulates is equal to the

need for purchases divided by the time it takes to perceive

the need.
(3.3b) A Young Cow Purchases Needed.K -
If Desired Milker Gap.K > 0
Then Desired Milker Gap.K
Else 0
(young cows)
(3.3c) C Time to Perceive Purchase Need — 1
(month)
(3.3d) A Desired Milker Gap.K - Desired Milkers -
Milkers.K
(milkers)

The young cows needing to be purchased is defined
as the difference between the desired number of milkers and
the current number of milking animals.

Desired Milkers is specified in two different ways,
given the scenario being performed. The first is where
Desired Milkers is entirely exogenous, at either a constant
level or perhaps increasing substantially at some date to
represent the desire to greatly increase the herd size. The
other specification leaves the exogenous component but also
makes it partly endogenous, responding to ability to meet
the Desired Milker goal (see the model overview above for

more detail).

(3.3e A Desired Milkers.K - (Base Value) + Desired
Milker Gap Average.K
(milkers)

115
The Base Value is generally set at 45 animals, an
arbitrary level. The average size of the gap, either
positive or negative, represents the performance of the
herd relative to the goal and this measure of performance

is used to adjust the goal up or down.

(3.3f) L Desired Milker Gap Average.K - Desired Milker
Gap Average.J + DT* (Desired Milker Avg Gap
In.KL)
(milkers)

This is an averaging process commonly used in system
dynamics that involves a goal-gap formulation [26, p271-272,
& 308-310]. The current value for the Desired Milker Cap is
being averaged into the performance measure, Desired Milker
Gap Average.

(3.3g) R Desired Milker Avg Gap In.KL - Desired Milker

Gap Gap.K / Time to Average Desired
(milkers per month)

(3.3h) A Desired Milker Gap Gap.K - Desired Milker
Gap.K - Desired Milker Gap Avg.K
(milkers)
(3.31) C Time to Average Desired Gap - 6
(months)

The equations 3.3g - 3.31 are a fairly standard
treatment of a goal gap averaging process. The reader who
finds this formulation confusing and would like to know how
it works is advised to consult the textbook cited above.
(3.3j) R Young Cow Arrivals.KL - Young Cows Orders.K /

Time to Purchase Young Cows
(young cows per month)

”Th!

the

116

The rate at which young cows being purchased arrive at

farm is equal to the current number of animals on order

divided by the time it takes for the order to be filled.

The time constant represents the time it takes to arrange

for the purchase and delivery of the animal.

(3.3k) C Time to Purchase Young Cows - 2
(months)
(3.31) A Milkers.K — Young Cows.K + Cows.K
(milkers)
(3.4) R Young Cows Culled.KL - Young Cows.K * (Death

COWS

need

Loss Young Cows + Maturation Rate Young Cows)
+ Cull Rate Standard Young Cows
(young cows per month)
There are three reasons for the removal of young

from the herd; losses due to death, or animals that

to be removed from the herd because they are poor milk

producers or poor reproducers. The last two reasons are

represented by the last variable in the equation.

(3.5)

(3.6)

C Death Loss Young Cows - .0058
(fraction per month)
(Seven % a year)

C Maturation Rate Young Cows - .09834
(fraction per month)

The assumption is that these animals will be

culled only after they have completed their first

lactation, even though they are not average producers. If

it takes ten months to complete a lactation than

approximately 9.8 % of the animals that are going to be

117
culled are removed each month.
(3.7) C Cull Rate Standard Young Cows = .0042
(fraction cows per month)

This cull rate reflects the assumption that a certain
portion of the young cows will need to be culled due to
unavoidable problems with either breeding or producing milk
(or just poor milk production). This portion is likely
partly endogenous to the system (a function of the herd
improvement practices) as well as partly stochastic and
therefore exogenous to the management system. For the
purposes of this research a single, constant portion set
arbitrarily at 5 percent per year is adopted. The Dairy Herd
Improvement Association publishes some Michigan data about
culling in DHIA herds for these reasons but it was not
possible to distill a single number from the data. The 5
percent I selected has no empirical or expert justification
at all and should be more precisely determined at some
time. (It is unlikely that the culling for this reason is
much larger than 5%, for if it was it would seem to be an
important source of loss and would therefore have more
concrete data published about it. If there is error in this
assumed value of 5 percent, it is likely an overestimation
of the losses due to these problems so that the ability of
the system to produce retainable animals is not over
exaggerated. Yet it is possible that these losses are more

significant than what has been assumed.)

 

118

(3.8) A Sale Young Cows.K - Young Cows Culled.KL * (l
- Death Loss Young Cows)
(young cows per month)

Only the proportion of animals removed from the herd
for reasons other than death loss can be sold (for

butchering or as milk producing animals).

(3.9) R Young Cows To Herd.KL - Young Cows.K *
Maturation Rate Young Cows
(young cows per month)
The assumption is that a young cow becomes a
mature cow when it freshens for the second time, which is

essentially the end of the first, which occurs on average

ten months after the last freshening (assuming a nine month

gestation).
(3.10) C Maturation Rate Young Cows - .09834
(fraction per month)
(4) Cows
(4.1) L Cows.K - Cows.J + DT * (Young Cows To Herd.JK
- Cows Out.JK)
(cows)

Animals enter from the young cow state and exit
according to the standard cull rate of 60 month life-cycle,
with provisions for increasing the cull rate to accommodate

a desire for rapid contraction of the milking herd.

119

(4.2) R ' Young Cows To Herd.KL a (see eq. 3.9)

(4.3) R Cows Out.KL - Cows.K * (Death Loss Cows +
Cull Rate Standard Cows) + Cull Rate Variable
Cows.K

(cows per month)
The rate at which cows disappear from the herd. The
variable is a function of the standard rate of culling,
death loss, and a variable cull rate depending on the

desired change in milking herd size change (if any).

(4.4) C Death Loss Cows - .0058
(fraction per month)

This variable takes the same value as that for

young cows, .07 a year or .0058 a month.

A4.5 C Cull Rate Standard Cows - .0455
(fraction per month)

The cull rate based on the age of the animal, where any
animal reaching the age of 60 months is removed from the
herd. The SOP here is that this policy is maintained to
attain the best average milk producing herd. The time spent

in this state is 22 months (from 38 to 60), so that the

variable is equal to 1/22 = .0455.

(4.6) A Cull Rate Variable Cows.K -
If Desired Milker Gap.K < 0
Then - Desired Milker Gap.K
Else 0

(cows per month)

120

The Desired Milker Cap is a measure of the difference
between the number of milking animals desired and the number
actually currently on hand. If there are not enough milkers
on hand (Desired Milker Gap > 0) then purchases of young
cows will occur (see above, eq. (3.3)) and there will be no
change in the mature cow cull rate. (This is not strictly
true for all sensitivity analysis simulations. Some
analysis is done on the effects of retaining mature animals
for longer than 60 months. See Chapter Seven for more
detail.) If there are more milkers on hand than desired,
then the excess milkers will be removed from the stock of
mature cows, the assumption being that on average the

younger animals will have better characteristics.

(4.7) A Sale Cull Cows.K - Cows Out.KL * (1 - Death
Loss Cows)
(cows per month)

The proportion of animals not lost due to death are

sold.

6.10 Summary

Five major components make up the herd management
model. Four of the components revolve around the separation
of the herd according to ages representing different
management problems and needs. These four are calves (0-4
months), heifers (5-27 months), young cows (28-38 months),

and cows (39-60 months). The fifth component is the

121

variable defining the major impetus for change in the

system, the desired number of milkers. Desired Milkers is

specified differently according to the scenario being

conducted.

Four different sets of simulations of this model are

performed and presented in the following chapter.

 

CH P EVEN
MU N R NA M N

7.1 Introduction

The principle reasons for simulating a system dynamics
model is to explore the properties of the system in
question. This exploration primarily takes the form of
examining the system's sensitivity to specific variables and
structural loops within the system. Before this
examination can take place, validation of the model must be
undertaken to determine if it is suitable for sensitivity
analysis. Validation involves the question, does the model
represent a real system? This can be answered by comparing
simulation results to the expectations of persons with
relevant experience, or comparing the results to actual
data, or both. The goal in constructing the herd management
model was not so much to create a "real system" as it was
to answer the questions from the introduction to Chapter
Six. These questions are:

1) Can STELLA be used to model a specific part of the
dairy operation?

2) Can this model rely on the use of SOP's to guide the
operation?

If STELLA is capable of translating assumptions
concerning SOP's and the way they are organized in specific
operations into a simulation model that gives reasonable
results, then it would be possible to answer yes to both of
these questions. It is not necessary for the model to give

122

 

123
completely expected results; unforseen but reasonable
results should be expected to occur given that the
simulation technique permits extremely complex interactions
between relatively straightforward SOP's.

The validity of the herd management model is therefore
a question of whether or not the simulation results seem
reasonable given the design of the model. This is in
contrast to the normally used meaning of validity presented
above. Given that the herd management model is intended to
answer less rigorous questions than a model which
undeniably represents a real system, a less rigorous
criterion of validity should be appropriate. Although a
significant attempt has been made in this modeling effort to
design a system of realistic SOP's, it is left to future
research to more conclusively investigate whether or not we
can develop assumptions concerning SOP's and their
organization that do in fact precisely represent a real
system.

The test of validity used here is essentially that used
by Balderston and Hoggatt, which they called the criterion
of model viability (see Chapter Five, pages 55-59). This
criterion is nothing more than; can the model be simulated
and give results consistent with what one would expect
given the design and specification of the model? The
validity of the herd management model, as judged by the

simulation results, is therefore dependent in part on the

124
congruence between what we would expect, given the
assumptions concerning the relationships among the
important management procedures, and the actual simulation
results.

The test of validity is really a pragmatic test of the
reasonableness of the assumptions that make up the model as
well as the reasonableness of the results, both of which
cannot be judged independently of the intended use of the
model. The bulk of the argument of the model's validity
really centers on informed debate about the quality of the
assumed SOP's and the structure uniting the SOP's into one
system.

Although a more clear specification of performance
criteria would be desirable, for the purposes of judging
the herd management model, it is assumed that congruence
will be demonstrated if the results indicate that the
births and deaths of animals are occurring and that there
are animals being culled and sold. No comparison is made to
data from actual farms, which is a shortcoming of the
analysis, but if the goal is to demonstrate the congruence
between assumptions and results, the comparison of results
to actual data takes on less significance.

The first set of simulation runs below, set I, are runs
of the basic model with marginal changes in some principle
parameters. The simulation results are judged with the

viability-congruence criterion in mind.

125

The third question asked in the introduction to Chapter
Six was:

3) Could a model oriented towards a specific problem
constructed in STELLA be of possible use to agricultural
economists?

A relatively objective test of whether STELLA could be
a valuable tool would be for agricultural economists to
pose actual problems to which STELLA could be applied. If
satisfactory answers were found, then with a certain
degree of confidence the answer would be yes.

The test used here for determining STELLA's problem
solving capacity is more limited and therefore less
conclusive than the above proposed relatively objective
test. Hypothetical herd management practices relative to
the problem of obtaining and maintaining a desired milking
herd size are postulated as a proxy for the questions of
agricultural economists. These questions are asked in Runs
II-IV within the context of sensitivity analysis, meaning
that the runs make changes in certain important variables
and loops to determine their importance to the herd
management system. The results of these runs are judged
relative to the question of STELLA's usefulness to
agricultural economists in the conclusion to this chapter.

The simulation runs detailed below fall into four
categories. The results from each run within a category are
summarized in tables to be found in the text. The

categories of runs are: I, no purchases of replacement

 

126
stock, a condition which allows us to examine the model's
viability and basic properties; II, constant purchases SOP;
III, variable purchases SOP; and IV, linking Desired Milkers
to performance (incorporating feedback directing the entire

subsystem).

7.2 The Herd Dynamics Without Purchases of Replacements

The underlying structure and properties of the model
can be explored by simulating it without any purchases of
replacement stock. Under these conditions, the herd is
forced to react as if it represented the dairy cows in a
completely closed system, like a country with no imports of
live animals. The simulation results reveal the various
strong and weak effects of the death and cull rates on the
herd as a whole. The results also demonstrate the need for
purchasing replacement animals if the individual farmer
either starts with an inadequate level of replacement stock
or is constrained to an inadequate level by housing or

management limitations.

7.2.1 Run IA; Baseline For No Purchases

A simulation model of the type created for the dairy
herd tends to reach an equilibrium or steady state if there
are no unstable cyclic reactions in the model or external
shocks. The steady state reflects a balancing between the

number of animals present in each of the states, Calves,

127
Heifers, Young Cows and Cows, and the losses of animals
from each of these states.

This baseline or steady state is established simply by
simulating the model and noting the final values. Aside from
the specified death losses and cull rates given in Chapter
Six's detailed description above, the starting values of
each of the states need to be established before the run.
For simple expositional purposes, the milking herd is
assumed to be 45 animals; 30 Cows and 15 Young Cows. The
Calves are started at 15 and Heifers at 30. The values and
other beginning conditions of this baseline run with no
purchases is presented in Table 7.1a. The graphical results

from this run are in Figure 7.1a-b.

Table 7.1a: No Purchases, Baseline, Run IA

 

- Cull Rates: Standard
- Death Losses: Standard
- Desired Milker goal: Constant, 45 Milkers

Initial Value of Final Value of
State Variable Variable State Variable
15 Calves 25
30 Heifers 65
15 Young Cows 15
30 Cows 30

Time to Steady State: 334 months

 

a)

h ‘)

fl
(“'1
.2.

bww- awed awvd
Haw—...: wa Has—...;

JOIN-‘-

aw»—

$01M”
.q-J

AWM‘
kw

Low-
k—v—oa

JOIN)‘
5*}

m
'0 O

31

2’5
0

r. .71 JM "was

.500
.500

250

.000

000
500

h—O
Y‘J
x

2 Makers 3 YoungCov.':.Cuhec 4 CowsCulled

A‘Ln‘J

 

 

 

J
l /
q
'1: 2/3/3
‘ 2 4/
4 W?!”
J 4——-'2
J 2/
1
d

fi' ' Ij'Y Y I ' "' I "‘ "lf'v' ' I ' "' 1" ' ‘ 1" 'f‘l
OO 90 000 180.000 270.000 360.

Time
2 Cows 3 YoungCows 4 Heife's

AAJk‘LA‘A-LAIA‘A‘ALAJ

 

 

V/+/
2 rh—
€‘wawwwtfl3
l
*‘"1*“1"*'T'v**1v"1"*1*"1'***':31
0 0 9(1 033 180 00:; ?70.000 er

Time

Tigurvs 7.1:2-1): Baselinv For No Purchasers, Run IA.

129

The time paths of four variables are depicted in Figure
7.1a, Desired Milkers, Milkers (the sum of cows and young
cows) and the cull rates for Young Cows and Cows. The goal,
Desired Milkers, is assumed to be a constant 45 milkers.
The number of Milkers falls drastically at the start of the
simulation to less than 30 animals and eventually reaches
the goal after approximately 334 months.

The reason for this large decrease in Milkers and the
long lag until reaching the goal is the inadequate level of
replacement stock (Heifers and Calves). Although the Young
Cows Culled and Cows Culled fall as the milking herd falls,
the effect of the removals of Milkers overwhelms the
capacity of the system to provide replacements. Eventually,
at approximately month 45 the culling of milkers has
dropped enough to permit the gradually expanding stock of
replacements to rebuild the milking herd. The increase in
Calves and Heifers and the associated increase in Cows and
Young Cows starting in month 45 can be seen in Figure 7.1b.

The entire dairy herd stabilizes eventually at 25

calves, 65 heifers, 15 Young Cows and 30 Cows.

7.2.2 Run IB; Altering Calf Death Losses

One area of intervention in this model of the dairy
herd is at the point of calf death loss. The assumed value
of 14 percent per year is an average for Michigan DHIA

herds although some farms have much lower rates of loss.

 

130
This run uses the same starting point as above except for
decreasing the Death Loss Calves from 14 percent to 7
percent per year. Starting and ending information is in
Table 7.1b, and the graphical results are presented in

Figures 7.2a-b.

Table 7.1b: Altering Calf Death Losses, Run IB

 

- Cull Rates: Standard
- Death Losses: Altered, Death Loss Calves - 7% per year
- Desired Milker goal: Constant, 45 Milkers

Initial Value of Final Value of
State Variable Variable State Variable
15 Calves 27
30 Heifers 65
15 Young Cows 15
30 Cows 30

Time to Steady State: 234 months

 

The variables graphed in this run are identical with
those in IA. Between the two sets of runs, the time paths
of the variables are very similar, except that it takes
much less time for this herd to reach steady state: while
the baseline herd in run IA took 334 months to reach steady
state, reducing the death losses to calves allowed the herd
to reach the goal in 234 months. The values for the state
variables in steady state in this run are identical with
those from the Run IA, except for the increase in two

calves when death losses are reduced.

 

1 DesiredMHKers

50.000
2.500

fiNNd

43.750
2.000

3000-.

37.500
1.500

JOIN-i“

8)

31.250
1.000

sum--

25 000
0.500

$03M“
W\'-‘

1 Calves,

ch DIN-I.
W
(A?
‘1
U1
0
C)

#000"

LNN‘
5...:
F)
{-71
O
C)
O

13)

JOIN)‘
w
~ N
U1
’3
0

AW“
\_.._4
O
0

Figures 7.28-b:

131

 

 

 

 

 

 

 

 

 

 

 

2 Mnmrs 3 CowsCuHed 4 YoungCowsCuHed
'1
q
1
1
J 1 I 1 1 12, 12 12
1 /
1 N 4. 4.
3 /4/3/\_,_3 3
J AZ/
1 2::
ti *vtn *** 1 ' "1'1" ' *‘I ' 1-211 '*‘ *fi'fiv T I "1" l
0.0 90.000 180 000 270.000 380
Tree
2 Cows 3 YoungCows 4 Heifers
1 ... . .
1 - 1
J /
1 V/
1 IM—l 1
1 2M/
1 ‘2
q
1
1'1VT “1 v .4", r'v ~ I . .44 T‘v ' ‘ 1 ‘**" Y‘f‘ ' I * h: 1
0 0 90 000 180.000 270 000 31-0
Toma

Baseline with Calf Death Loss Altered, Run 18.

132

7.2.3 Run 1C; Altering the Cull Rate Standard For Cows

The assumption that on average Cows are culled at the
age of 60 months is a reflection of the pattern of milk
output for the dairy animals. In other words, the
assumption is that on average, milking animals are
sufficiently past their peak at 60 months to warrant their
removal from the herd. What if this point were not reached
until the age of 72 months? It seems likely that increasing
the average age of the Cow should have large effects on the
dairy herd.

Table 7.1c summarizes the starting and ending
conditions for this run. The baseline conditions from Run

IA are used, except for increasing the average age of Cows

to 72 months.

Table 7.1c: Altering Cull Rate Standard Cows, Run IC

 

- Cull Rates: Standard
- Death Losses: Standard
- Desired Milker goal: Constant, 45 Milkers

Initial Value of Final Value of
State Variable Variable State Variable
15 Calves 25
30 Heifers 50
15 Young Cows 12
30 Cows 33

Time to Steady State: 66 months

 

a)

b)

133

 

 

 

 

 

 

 

 

1 Dealrendlkers 2 Milkers 3 CowsCuHed 4 YoungCowsCulled
1
51,} 50.000 _
3} 2.500 11
1
11} 43.750 1 1 ‘2 ‘2
3 '1
4} 2.000 1
1 J
2} 37.500 _.
14"} 1.500 1 L
1 _ A
, 7 ‘ 3 ‘5
§} 31.250 _.
4} 1.000 ;
1 J
§} 25.000 1 - .5215.-. 1 1 1 - m --- firm .-. fill“, --.“ . .-.
4} 0 500 0.0 90.000 180.000 270.00 360C
Tame
1 Ca‘ves 2 Cows 3 YoungCows 4 Heifers
1
£1 50 000
J‘
4 75.000
1
§] 37.500
‘ 56 250 4/\2_,4-_==-‘24-‘==-—_=
1 2/
3] 25.000
4 37.500 ‘/‘-1~ 1
1
4' 18 750
1
g} 00 I V ' 'If' ""V'Y fiwfi'j'fi‘ fi'fir" l" V *‘1 ' 'v'VT ' ' ' 1
4 00 0.0 90.000 180 000 270 000 350
Mme

Figures 7.

Run 1C.

3a—h: Baseline With Cull Rate Standard for Cows Altered,

134
The simulation results graphed in Figure 7.3 a and b show
that both the time it takes to reach steady state and the
time paths of the variables are significantly different
than either IA or IB. The principle effect of the increased
age is that there are more births of calves in the herd.
The relatively rapid increase in Calves and Heifers
(Figure 7.3b) feed more animals into the milking herd. By
month 66, there are so many heifers that there is pressure
on the milking herd to increase beyond the desired level.
The sharp increase in Cows Culled (Figure 7.3a) in month 66
reflects the effect of the culling of these excess Cows.
Within a few months the herd reaches it's steady state with
the levels of animals in the various states different from

those in either IA or IE (see Table 7.1c).

7.2.4 Discussion of Runs IB and IC

Either of these two changes in the baseline, reducing
calf death losses or increasing the average age of Cows
could be incorporated into the herd management model. In
fact, the death losses could be made endogenous to the
system, responding to a level of management expertise
following some sort of learning curve. Or, average cow age
at the time of culling could be a stochastic, normally
distributed variable, moving randomly between 60 and 72
months, or any desired age. The zero point of the normal

distribution could be altered over time to reflect the

135
changing genetic composition of the herd.

Although these modifications are possible, for
expositional purposes in all of the following runs it is
assumed that the values of cull rates and death losses are
those used in IA. Still, the basic implication of all of
these runs is the same; the farmer needs to either have
sufficient levels of replacement stock or must be able to
purchase animals if the Desired Milker goal is to be reached

in fewer than five to thirty years.

7.2.5 Run ID; Adequate Stocks of Replacements

Starting the herd off with the steady state number of
replacement animals causes the results you would expect;
Milkers equals Desired Milkers from the start, with
essentially no deviation from the initial condition. The

results are summarized in Table 7.1d. Graphs of the results

Table 7.ld: Adequate Stocks of Replacements, Run ID

 

- Cull Rates: Standard
- Death Losses: Standard
- Desired Milker goal: Constant, 45 Milkers

Initial Value of Final Value of
State Variable Variable State Variable
25 Calves 25
65 Heifers 65
15 Young Cows 15
30 Cows 30

Time to Steady State: 0 months

 

a)

b)

1 Desrredeers

%} 50.000
‘} 2500
1 ..

2} 43 750
i} 2.000
1

a} 37.500
4} 1.500
1

§} 31.250
d 1.000
1

g} 25 000
4} 0.500
1 Caves

1

§} 50.000
4 75.000
1

g} 37 500
4 56.250
1

a} 25.000
4 37.500
1

g} 12 500
4 18.750
1

2] 0.0
3

4 0.0

136

 

 

 

 

 

 

 

 

 

 

 

2 Mrlkers 3 CowsCuHed 4 YoungCowsCuhe—d
'1
j 12 12
1 4 A
1"— . v
1*- 3 3—
r
1
1
J
11v Vi‘I" v'v I'vrvrv 1 v """ 1"" V 1*v*'**'1 "‘ ' 1 ' ' v 1
0.0 90000 180.000 270.00 360.
Turns
2 Cows 3 YoungCows 4 Heifers
.,
4
1 A A
1 Y V
q
J
1
1
2 2
1
, 1 1
I
j 3 3——-—-
I
1
1
7*v'T""'1' Tr'T*'I7"'IT"'I"‘ "7711
00 90.000 180.000 270.000 360
'11an

Figures 7.4a-b: Baseline With Adequate Stocks of Replacements, Run ID.

137
are in Figures 7.4 a and b. The simulation is aborted after

100 months due to lack of interest in straight lines.

7.2.6 Model Viability and Validity

The herd management model performs well relative to the
criteria of reasonableness of results and their congruence
with the design of the model. Births, deaths, culling and
sales all are occurring, and although the steady state
levels of animals in the herd varied according to changes
in fundamental parameters, the results are consistent with
the assumptions of the run as well as the SOP's in the
model. Lacking more formal criteria, the conclusion to be

drawn is that the model is viable.

7.3 SOP on Constant Purchases of Replacements

Not all farmers will be able to raise the quantity of
replacement stock demonstrated in Run ID that is needed to
maintain the herd at the Desired Milker level. The
constraints on animal housing, ration supplies, labor and
management skills could limit the total number of
replacements the farmer can maintain. In this set of runs,
for expositional purposes the farm is assumed to be able to
raise only as many Heifers as there are Milkers. Calves can
only reach to two-thirds of the number of Milkers. Before
examining a constant purchases SOP it will be helpful to

demonstrate the effects of these constraints.

138
7.3.1 Run IIA, Constrained Stock, No Purchases
The constraints on the stock of replacement animals
forces the farmer to purchase replacements if the Desired
Milker goal is to be met. Figures 7.5 a and b present
graphs of the results of a simulation with these

conditions. The main features of the run are summarized in

Table 7.2a.

Table 7.2a: No Purchases and Constrained Stock, Run IIA

 

- Cull Rates: Standard
- Death Losses: Standard
- Desired Milker goal: Constant, 45 Milkers

Initial Value of Final Value of
State Variable Variable State Variable
25 Calves 17
45 Heifer 45
15 Young Cows ll
30 Cows 20

Time To Steady State: 80 Months

 

The removals of both Cows and Young Cows initially
pulls the number of Milkers down as before in the runs in
I. The difference is that Young Cows and Cows are never
able to recover to Desired Milkers as any Heifers in excess
of the 45 animal constraint must be removed from the herd.
The dairy herd's steady state condition is only 31 Milkers,

with 11 Young Cows and 20 Cows. There are fewer calves than

a)

b)

1 DesredMuikers

%} 50 000
4; 2.500
1

2} 43.750
3’} 2.000
1

a} 37.500
4} 1.500
1

a} 31.250
4} 1.000
1

§} 25.000
4} 0.500
1 Cakes

1

§] 50.000
4 75.000
1

3} 37.500
4’ 56.250
1

§} 25.000
4 37.500
1

§} 12.500
4 10.750
1

2} 0.0
3

4 0.0

139

 

 

 

 

 

 

 

 

 

 

 

2 Merrs 3 CowsCuHed 4 YoungCowsCuHed
7
4
1
n1
1
‘ 1 I I
1
I
4
‘ ‘ 4 1
1 3 2 1
.. 2:; 2"
1
1

*7 1'1 777777 1W1 1'1 """" 1 ' 1'17 ***** 1 1 ' ' ''''
00 90.000 180.000 270.000 360.
Time

2 Cows 3 YoungCows 4 Heifers

‘ ‘r ‘.

3 3 3
[F'**" 1 """ 1'Y'T" 1"" ' 1 ‘‘‘‘‘ 1'7fifi'i ***** '*'* 1
0.0 90.000 180.000 270.000 360.

Time

Figures 7.5a-b: No Purchases and Constrained Stock, Run IIA.

140
the 25 animal constraint because of the lack of animals

giving birth.

7.3.2 Run IIB. Constant Purchases SOP

The farmer could have an SOP associated with the
Desired Milker goal of 45 animals which dictates the
purchase of two Young Cows every month. Such an SOP could
be the result of the farmer's own experience or it could be
learned from an experienced farmer. If a son takes over a
dairy operation from his father, and if the father always
purchased two animals a month then the son may very well do
the same.

Under the hypothetical conditions of this model, this

SOP will permit the farmer to attain the Desired Milker

Table 7.2b: Constant Purchases SOP, Constrained Stock, Run
IIB.

 

- Cull Rates: Standard
- Death Losses: Standard
- Desired Milker goal: Constant, 45 Milkers

Initial Value of Final Value of
State Variable Variable State Variable
25.0 Calves 21.7
45.0 Heifers 45.7
15.0 Young Cows 20.4
30.0 Cows 25.3
45.0 Milkers 45.7

Time to Steady State: 0 months

 

I [‘esiredMilkem

%} 50.000

4} 2.500

g} 43 750

4} 000

1 r

g} 37.500
6) 3} 1 500

1

§} 31.250

4} 1 000

1

g} 25 000

4} 0500

1 Cakes

1

a] 50.000

4 75.000

1

g} 37 500

4 56.250

1

g} 25 000
b) 4 37 500

1

13>] 12.500

4 16750

1

g} 00

4 00

Figures 7.6a-b: Constrained Stock, No Purchases, Run IIB.

141

 

 

 

 

 

 

 

 

 

 

 

 

2 M1lkers 3 CowsCulled 4 YoungCowsCuued

q

1
4
1 4 4 a
1- 12 12 12
- 3 3 3,
1

1
1
d
1
1
1

"'V'fiT" ' "1 7 7 "1 ' V" 1 7" " V" "_l V" "1 ""fi 1
0.0 90.000 180.000 270.000 360.

Tune
2 Cows 3 YoungCo-ws 4 Heifers

q
1
:1
1
d
1
.1
1
‘ 4 ‘r ‘.
1 o 9
..1 2 c. ..
1
1‘ 'r 1t 1'1
1
1
1
4
1

''''' 1“ . 1 1*r" ‘ 1 1" r'W'rvww "' ‘ 1 7 "7 1"'*"‘1
00 90.000 180.000 270.00 360.

Time

142

goal. Milkers, in fact will exceed Desired Milkers because
the SOP introduces slightly more "animal" than needed.
Excess Cows are being culled as a result (Figure 7.6a and
Table 7.2b). There is also a slight excess of Heifers as
the maturing animals cannot be removed fast enough (Figure
7.6b and Table 7.2b). These types of errors are called
steady state errors and are the differences between goals

and actual levels when the system reaches equilibrium.

7.3.3 Run IIC, Constant Purchases SOP, Expand the Herd

A reasonable shock to this dairy herd system would be
to expand the operation significantly. The forces leading
to the decision to significantly expand could be made
endogenous to this model, although it would mean modeling
the farmer's "personal" goals and values. The decision to
expand the operation can come from sources other than a
desire to profit maximize. For example, if a farmer wishes
to take advantage of new technology to minimize his
physical strain, capital will need to be borrowed for the
purchase. The cash flow requirements of the borrowed
capital will likely require a larger milking herd. If a
farmer wishes to bring a son into the business, perhaps as
a partner, then more milking animals will be needed to
generate sufficient incomes for the two owners. Neither of
these two examples involve profit maximization, and if the

loans can be arranged the expansion would likely occur.

143

Holding all of these considerations aside, the simple
solution is adopted in this set of runs whereby the dairy
model is shocked with an expansion of the herd to 90 milking
animals in month 60. The exact timing of this expansion
would be hard to predict correctly for a real farm even if
assuming profit maximization. This problem is avoided by
assuming that something is going on in the farm in month 60
that leads to this desire to expand.

In addition to the expansionary shock, which is
initiated by the purchase of 45 Young Cows, other changes
in the model need to be made to accommodate the expansion.
The constraints on the maximum number of Calves and Heifers
is raised proportionately. The SOP on the purchase of Young
Cows is changed from two animals per month to four. Figures
7.7 a, b, c, and d are graphs of the simulation results.
The run is summarized in Table 7.2c.

Before month 60 the state variables follow the same
pattern as the previous run. In month 60 a sharp increase in
Desired Milkers occurs (Figure 7.7a). The expansionary
purchases of Young Cows raises Milkers up to this level at
the same time. The increase in Young Cows also leads to the
dramatic increase in Young Cows being Culled, which
temporarily pulls down Milkers below the goal. In less than
15 months, by month 75, Milkers is slightly in excess of

Desired Milkers.

144

Table 7.2c: Constant Purchases, Expansion in Month 60
(Exogenous Shock), Run IIC

 

- Cull Rates: Standard
- Death Losses: Standard
- Desired Milker goal: 45 animals until month 59, 90 cows

thereafter

- Constraint on Heifers: 45 animals until month 59, 9O
thereafter

- Constraint on Calves: 30 animals until month 59, 6O
thereafter

- SOP on Purchases of Young Cows: 2 animals per month

59, 4 animals per month thereafter

Initial Value of Final Value of
State Variable Variable State Variable
30.0 Calves 43.2
45.0 Heifers 91.4
15.0 Young Cows 40.7
30.0 Cows 50.7
45.0 Milkers 91.4

Time to Meet Desired Milker goal after expansion: 15 months

 

The herd is able to recover because of the increase in
the replacement stock which occurs in response to the
births from the increase number of Milkers. Calves In
(Figure 7.7c) rises just as dramatically as the number of
Milkers, leading to greater numbers of Young Heifers
Available for replacement. The rapid rise in Heifers and
Heifers For Replacement in turn increase the milking herd
size (Figure 7.7d). Once the number of Milkers stabilizes
above Desired Milkers it is the SOP of four purchases per

month that keeps the system stable.

 

 

 

 

 

 

 

 

 

 

 

 

1 Desireszfkers 2 MUM-us 3 CowsCoueo‘ 4 Your‘gCowsCulled
11,} 100.000
§} 8.000 1,
.1 12 ,2
5} 81.250 ‘1
3 n .1 7 4'
1 . ‘
2} 82.500 3
a) 3} 3.250 1
1 4.—1
1,} 43.750 . 123,:
3
4} 1.875
1,} 25.000
'1"'* 1‘Y""'1‘7'**hT'T'1r' 1'r'7‘T'1 1 7'1'1 ' "1* 1 r‘v 1
a} 0500 010 90.000 180.000 270.000 360
me
I Calves 2 Cows 3 10009001125 4 Heifers
1
'3’} 100.000 -
1
" . 4 4.
1 .1
g} 75.000 1
4 1
’ 1
g} 50.000 1 7—~—2
4 1r—— ‘37 ‘3
, 1
§} 25.000 «E72
4 1f
1 1
a 0-0 111111 IfiVWH 1" '77*1** "T*1**" ' 1 ' ‘ :1 7*"1 FY'T‘: 1
I. 0.0 90.000 180.000 270.000 380.
Tums

Figures 7.7a-b: Constant Purchases, Expansion in Month 60 (Exogenous
Shock), Run IIC.

C)

d)

I Ca1ves

1 100.000
2

3 15.000
4

1, 75.000
3} 11.250
4

1 50.000
2‘

31 7.500
4

1 25.000
2

3} 3.750
4

1 0.0
2

3}
4

I Heifers

1 100.000
2

3 15.000
4

1 75.000
2

3} 11.250
4

1 50.000
2

3] 7.500
4

1 25.000
2

4

1 OO
2

3} 00
4

Figures 7.7c-d: Constant Purchases, Expansion in Month 60 (Exogenous

Shock), Run IIC.

146

 

 

 

 

 

 

 

 

 

 

2 Calvesln 3 CalvesOut 4 YHe’rfersAvaHable

J
1
1 2} 23
1
4
1
1 2 1 1
d “I I
1
‘L.
1 I
1
1
1

' "' I "' ‘ I "' hi """ T"" " "VV "1 ' j 1'I" "fij
0.0 90.000 180.000 270.000 360.

Time
2 Hezfersln 3 HeifersOut 4 HeifersFRep’ace

'1
1 1, 1
1
1
1
JF—‘1
.‘ 2‘5 23
1 I
q 1 I
1
1[:IE
1

'f3" T '37-,31 ' V ' I"" ' 1 """" '3 '7‘31 7" ' Ifir'7333—1
00 90.000 180.000 270.000 360.

Tlme

 

147

7.3.4 Run IID, Constant Purchases, Expand Young Milkers and
Heifers

The sharpness in the dip in Milkers immediately
following expansion in the previous run could cause the
farmer to over react. If the farmer does not want to lose
this many milking animals, one corrective policy would be
to purchase all of the needed replacement Heifers at the
time of purchase of the Young Cows. Figures 7.8 a through d
present this scenario and demonstrate that the purchases of
Heifers permits Milkers to immediately meet the Desired
Milkers goal. The response of the variables in all four

graphs is correspondingly rapid.

Table 7.2d: Constant Purchases, Expand Young Cows and
Heifers in Month 60; Run IID

 

- Cull Rates: Standard

- Death Losses: Standard

- Desired Milker goal: 45 animals until month 59,
90 thereafter

- Constraints on Calves and Heifers: Same as IIC

Initial Value of Final Value of
State Variable Variable State Variable
30.0 Calves 43.2
45.0 Heifers 91.4
15.0 Young Cows 40.7
30.0 Cows 50.7
45.0 Milkers 91.4

Time to Meet Desired Milker goal after expansion: 0 months

 

a)

b)

148

 

 

 

 

 

 

 

 

 

 

 

 

1 Desimdekers 2 Mulkers 3 YoungCowsCuHed 4 CowsCuHed
1
1,} 100.000
i} 6.000
12 12 12
5} 81.250
3:} 4.625 3 3 3
4. 4. 4
1
3} 62.500
3} 3.250
1 3—4
g} 43 750 124:,
4} 1.875
;} 25.000
3 1 *v ' I ' "' r ' ""1"" f'Trrvr rfi'V "fir ' Y'Y' V" *1
4} 0500 0.0 90.000 180.000 270.00 360
me
I Cahes 2 Cows 3 YoungCows 4 Heifers
1
3 100.000 1
4 1 4. 1. 4
5 l
3 75.000 1
4 1
1
11 1
3 50.000 1 2 2
1
$1 1
a} 25-000 7%.?
4 1r-
11 l
2 4
3 0° 'V"r 1'1V' "1 ‘ V v'1 """ 1" ' 'TU'1 ' Y T'Tfi' " V'T ' I
4] 0.0 90.000 180.000 270.000 1 360.000
Time

Figures 7.88-b:
Young Cows, Run

Constant Purchases, Expand the Herd with Heifers and
III).

 

C)

d)

1 Caives
1 100.000
2
3} 15.000
4
1 75.000
2
3} 11.250
4
1 50.000
2
3] 7.500
4
1 25.000
2
3] 3.750
4
1 0.0
2
3] 0.0
4
1 HOWB'S
1 100.000
2
3} 15.000
4
1 75.000
71
3} 11.250
4
1 50.000
2
3} 7.500
4
1 25.000
2
3} 3.750
4
1 0.0
2
3}
4

Figures 7.8C-d:
Young Cows, Run

 

 

 

 

 

 

 

 

 

 

 

 

2 Caivesin 3 CaivesOu! 4 YHeiiersAvaiiable
1 2L 23 23
J
l
1
4
1
1
1 23-1 1 l l
1
JL.
:~__l
i
""" I '* "l"'* ' I'r ' "I * "'v1 ""1 I "frr I vr'v'fi
0 0 90.000 180 000 270.00 360,
Time
2 Heifersin 3 HeifersOui 4 HeifersFRepiace
7
1 1 1 1
1 i
1
1
1
‘ J
q
jr——————1
; 23 23 23
1 I 4. 4. 4
:2:th
3
1
***** I *‘1 r I" V "1" *fi'1 1 ***' *"' "I"" v "v "7
0.0 90.000 180.000 270.000 360
Time

Constant Purchases, Expand the Herd with Heifers and

Ill).

150

7.3.5 Run IIE, Constant Purchases, Expansion only in
Heifers

A better strategy for expansion might be to expand only
in Heifers and then let the replacement herd generate the
Milkers. The lag of Milkers behind Desired Milkers in this
case is 17 months (Figure 7.9a and Table 7.2e), rather than
15 months when Young Cows were purchased for the expansion.
The immediate increase in Heifers pulls the other states up
during this period without the drastic dip in milker

numbers experienced in Run IIC.

Table 7.2e: Constant Purchases, Expand Herd With Only
Heifers Only in Month 60, Run IIE

 

- Cull Rates: Standard

- Death Losses: Standard

- Desired Milker goal: 45 animals until month 59,
90 thereafter

- Constraints on Calves and Heifers: Same as IIC

Initial Value of Final Value of
State Variable Variable State Variable
30.0 Calves 43.2
45.0 Heifers 91.4
15.0 Young Cows 40.7
30.0 Cows 50.7
45.0 Milkers 91.4

Time to Meet Desired Milker goal after expansion: 17 months

 

This type of expansion may be cheaper than either the
purchase of Young Cows or the purchase of Young Cows and
Heifers. In comparison to adding Young Cows, only two

months are added until the goal can be reached but there

a)

b)

1 DesiiedMiikers

}} 100.000
3

4} 6 000
3,} 81.250
3

4} 4.625
2,} 62.500
3‘

4; 3.250
;} 43.750
3

4} 1.875
;} 25.000
3

4} 0.500
1 Caives

1

2

3} 100.000
4

1

2

3} 75.000
4

11

2' 50 000
3J .

4

1

§} 25.000
4

1

2

3} 0.0
4

 

 

151

 

 

 

 

 

 

 

 

 

 

 

 

2 Mil-«ers 3 CowsCulled 4 YoungCowsCulied

1 ,2 ,2

4

J

T 4. A.

7 3 3

7

J 4_4

_: 123::

l

‘r'r'Vrir' **7*T ' "*‘1 V" f'V"" 'rl"‘T** I '''' 'Wrrfi'l
0.0 90.000 180.000 270 000 360.

Time
2 Cows 3 YoungCows 4 Heifers

1 {F4—
1

i 9 2

i 11 it
J§::::T2

lr’-‘

J

1

*fi**' I "*‘r 1'—' ' 1'1 *‘Y "I" ' v'F'rr "1 V "Y 1 ' *‘r‘n
0.0 90 000 180.000 270.000 360.

Time

Figures 7.9a-b: Constant Purchases, Expand Herd with Heifers Only, in
Month 60, Run IIE.

152
is no loss due to the culling of milking animals. When
purchasing only Heifers is compared to expanding with both
Young Cows and Heifers, the cost of purchasing the Young
Cows is avoided, although there is the 17 month lag of
Milkers behind Desired Milkers. The purchases of Heifers
appears to be a viable expansion strategy under certain

conditions.

7.4 SOP on Variable Purchases of Replacements

A farmer may decide to calculate each month the number
of animals he needs to purchase based on the difference
between the current number of Milkers and the Desired
Milkers goal. The milking animals purchased in this model
are assumed to be young cows. The calculation of Young
Cows Purchased in the model involves a delay of two months
to account for some time to perceive and act on a shortage
of Milkers. This delay should give different results from
those with the constant purchases SOP because making the

calculation each month may result in less over-purchasing.

7.4.1 Run IIIA, Baseline‘for Variable Purchases

The run begins with the same replacement stock levels
as used in the constant purchases SOP. It is assumed that
it takes the farmer two months to perceive and be able to
act on the shortage of animals. Figure 7.10a shows that

Milkers is not able to meet Desired Milkers when the system

153

Table 7.3a: Baseline for Variable Purchases, Run IIIA

 

- Cull Rates: Standard

- Death Losses: Standard
- Desired Milkers: Constant, 45 animals
- Time to Purchase Young Cows: 2 months
Initial Values Final Values
of Variables Variable of Variables
30 Calves 20.9
45 Heifers 45.6
15 Young Cows 15.1
30 Cows 29.0
45 Milkers 44.1
N.A. Desired Milker Cap .90
0 Young Cows Purchased .90
N.A. Heifers for Replacement 2.2
N.A. Young Cows Culled 1.7
N.A. Cows Culled l 5

 

Table 7.3b: Altered Time Constant from Baseline, Run IIIB

 

- Cull Rates: Standard
- Death Losses: Standard
- Desired Milkers: Constant, 45 animals

- Time to Purchase Young Cows: 4 months
Initial Value of Final Value of
Variable Variable Variables
30 Calves 20.9
45 Heifers 45.6
15 Young Cows 15.1
30 Cows 29.0
45 Milkers 44.1
N.A Desired Milker Cap .90
N.A Young Cows Purchased .90
N.A Heifers for Replacement 2.2
N.A Young Cows Culled 1.7
N.A Cows Culled 1.5

 

154

has reached steady state. The steady state error is given
by the Desired Milker Gap and is equal to (approximately)
one cow. The steady state error results from the interplay
of the purchasing delay, death losses, cull rates and birth

rates .

7.4.2 Run IIIB, Increasing the Purchasing Delay to Four
Months

The steady state difference between Milkers and Desired
Milkers is the same when the purchasing delay is four and
two months. All of the final states for the two runs are
also identical (Tables 7.3 a and b). The only difference
between the two runs is the initial dip in the number of
Young Cows before purchases of Young Cows returns the the
herd to the steady state. Young Cows in this run declines
slightly more when the purchase delay was only two months
(comparing Figure 7.11b to Figure 7.10b).

The difference between Young Cow Purchases in this run
(Figure 7.11) and the previous (Figure 7.10c) is the cause
of the increased decline. It takes slightly longer for
purchases to reach an adequate level when the delay is four
months. As a result, the dip in Milkers in this run is also
slightly greater than under the previous conditions,
although the differnce is almost imperceptible in the
figures (comparing 7.11a to 7.10a). The main point is that

altering the value of the time delay has no impact on the

a)

b)

C)

I DemredMotkers

1

3} 75.000
4} 2.500
32} 62.500
3

4} 2.000
‘32} 50.000
3

4} 1.500
'2} 37.500
5

‘} 1.000
1,} 25.000
2} 0.500
I Caives

1

22} 50.000
4 75.000
1

2} 37.500
3 56.250
1

g} 25.000
4 37.500
1

g} 12.500
4 18.750
1

g] 0.0
4 0.0

I YCWPurchases

;} 7.000
i} 2.500
} 5.250
2 2.000
.1, 3.500
§ 1500
}} 1.750
‘3‘} 1.000
2.} 0.0
g} 0.500

155

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2 Mulkcrs 3 0011115003100 4 YouryCowsrfuf'cc
1 $ 4
‘5 2% 1'.
1
j 12 17. ‘2
J
7
1
fir-wv'vv'v‘jv-r fVVVT V'—v'v ["7 T "_T TWer 1 'V' Y YWV V "Y i ‘
0.0 90.000 180.000 2 10.000 363 000
2 Cows 3 YoungCows 4 Heifers
q
1
1
1
1
1
‘ 4. 4 4
.‘ 2 2 2
1 E 1 1
j I 1 I
.1
1
1
J
1
""'*"Y"‘ jV'VT"'_'"'I"TVT' Tf"'l
0.0 90.000 180.000 270.000 360.000
2 HenfersFReplace 3 YoungCowsCulled 4 CowsCulled
3 3 3
4. 4~ 4
i
1
4
,r~———2 2 2
1
1
d
l v 1 1 E
1
j'fi 1 ‘fi"'1 ‘ ' ' ‘ "' I ' ' ' ' ' "1 V r"' " ' "
0.0 90.000 180.000 270.000 360.000
Tnme

Figures 7.10a-c: Baseline for Variable Purchases, Run IIIA.

a)

b)

I Desireerikers

1

§} 75.000
4} 2.500
1

§ 62.500
4 2.000
1

3,} 50.000
3} 1.500
1

§} 37.500
4} 1.000
1

§} 25.000
4} 0.500
I Calves

1

§] 50.000
4 75.000
1

£1 07500
4 56.250
1

§} 25.000
4 37.500
1

§} 12.500
4 10.750
1

g} 00
4 0.0

1 YCWPurohases

g} 7.000
2} 2.500
}} 5.250
fi} 2 000
g} 3500
§} 1.500
‘} 1.750
E} 1.000
% 0.0
4 0,500

156

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2 Milkers 3 CowsCulied 4 YoungCowsCuIled
.k 4. ‘.
.L 5*
j i 1, 1.
q
1
' ' 1 I ' ' ' Y ' ‘ ' ' ' " ‘ ' ' ' F ‘ 'fi ' fil
00 90.000 180.000 270.000 "60.!
2 Cows 3 YoungCows 4 Heifers
1
1
4
' 4 4
, 2 2
1
§ 1. I
lv_' 3 VL
'1
4
1
1
1
"V" r "Y ' Y ' "v I ''''' 1 '''''' T"" ' I ' Y '*I" 4~4~1
0.0 90000 180.000 270.000 360C
2 HeiiersFReplace 3 YoungCowsCulled 4 CowsCuHed
1
4
1
1
.' -‘—~ 3 3,
«M 4 4
l—~————2—————2~———
1
1 1 1
1
' 'TV'IV1 ' 'VF""fi " 'V' I v ' 7V ' 'vv ' ""Y vafifi
0.0 90.000 180.000 270.000 360.0

Figures Ila-c: Altered Time Constant from Baseline, Run 1118.

157
steady state condition.

The variable purchases SOP (Run III) does not perform
as well as the constant purchases SOP (Run II) if
performance is judged by ability to meet the Desired Milker
goal. Although the Desired Milker Gap of one animal when
purchases are variable is only two percent of the goal, the
absolute value of the gap will amount to more animals as
the herd size increases. If two percent error is acceptable
then the variable purchases SOP may be preferable to the
constant SOP since the variable purchases does not result

in over-purchases at any time.

7.4.3 Run IIIC, Variable Purchases, Expansion in Month 60

Expansion occurs in this run by purchasing Young Cows
in month 60. Desired Milkers and the constraints on
replacement stock are also increased. As shown in Figure
7.12s, Milkers falls off from Desired Milkers for
approximately 15 months, but eventually returns to a steady
state error of approximately 2 animals (Desired Milker Gap
in Table 7.3c). The initial dip in Milkers results from the
drop in Young Cows as they are culled in greater numbers.
The drop in Young Cows can be seen in Figure 7.12b and the
increase in the cull rate in Figure 7.12c. Animals are also
moving out of Young Cows resulting in a growth in the number
of Cows (Figure 7.12b). The variable purchases SOP also
tries to compensate for the deficit in Milkers by

purchasing more animals (Figure 7.12c), but the herd will

158

Table 7.3c: Baseline with Expansion in Month 60, Run IIIC

 

- Cull Rates: Standard
- Death Losses: Standard
- Desired Milkers: 45 animals until month 59, 90 animals

thereafter
- Time to Purchase Young Cows: 2 months
Initial Value of Final Value of
Variable Variable Variables
3O Calves 41.7
45 Heifers 91.2
15 Young Cows 30.3
30 Cows 57.9
45 Milkers 88.2
N.A. Desired Milker Gap 1.80
N.A. Young Cows Purchased 1.80
N.A. Heifers for Replacement 4.5
N.A. Young Cows Culled 3.3
N.A. Cows Culled 3.0

 

 

not reach the steady state until there is a sufficient
number of Heifers.

No analysis of expansion by the purchase of Heifers is
performed with the Young Cow variable purchases SOP since
the results will essentially be identical to those obtained

in Run IIE (except for some steady state error).

a)

b)

C)

I DesnedMilkers

1
13,} 100.000
‘} 6.000
1
§ 01250
g 4625
1 62.500
3
g 3250

1.875

25 000
0 500

}
}
} 43.750
}
1
}

75 000

25.000
00

}
I
1
1

15 000
6.000

1 3.750
1.875

% 00
4 0.500

1

59

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2 Milkers 3 CowsCulied 4 YoungCowsCuiIed

1

1 ‘7' ‘1

.1

4

1

4

.‘ ‘L 4

1 3 3

jag—a

1

1 "T fir """ I VVVVV jfiYfi' Y ''''''' Y '''''' 1'" ' ' V V hi i

0.0 90.000 180.000 270.000 360.
2 Cows 3 YomgCows 4 Heiiers

1 4 4

1

4

3 :— 2

1 . 1 1

4

1 LL .1.

;:1

JV—‘——'

1

' fi' 1 ' "‘*I ' ' ' 4 v fivvr'v v -4 ,fii v v v v vvvfi

0.0 90.000 180.000 270.000 360.
2 He.iersFRepiace 3 YoungCowsCuiied 4 CowsCulled

1

1

1

d

1

4

1

1

J L 35

J 4 4

1 2— 2

.4

harm—.3421

”—723 ' ‘

1 .

1 I

'fi V'Ti' V" r "7'1 V ' j‘fir‘7' ' ' V """" T "‘ " VVVVVV
00 90.000 180.000 270 000 360.1
Time

Figures 7.12a—c: Variable Purchases with Expansion in Month 60,

Run 111C.

160
7.5 Linking the Desired Milker Goal to Feedback

Although introducing meaningful feedback to the Desired
Milker goal from the financial performance of the farm
should result in some interesting dynamic results, this
would involve linking the herd to finance and cost
components. Leaving this type of analysis for future work,
some dynamic feedback effects can be observed by linking
the ability of the herd to meet the Desired Milker goal to
the goal itself.

Figure 7.13 is a causal loop diagram of one conception
of linking Desired Milkers to feedback. A new definition of
Desired Milkers is used in this formulation. Desired
Milkers Base represents the basic number of Milkers the
farmer wants to manage, while Desired Milkers Adjusted
represents an adjustment of the Base as a result of
feedback. The difference between Desired Milkers Adjusted
and Milkers determines the Desired Milker Gap. This is the
gap used to determine how many Young Cows should be
purchased (as in Run III). A perceptual and procedural delay
occurs between the gap and purchases. The gap can also be
used to adjust Desired Milkers Adjusted. As the gap
increases an increase will occur in the average degree of
error which leads to an increase in the Desired Milkers
Adjusted goal. This, in turn leads to purchases of Young
Cows. The averaging of the Desired Milker Gap implies a lag

or delay.

161

 
     
    
     
   

 
     
   
   

Desued
MerrS
Bose

Average '

Desked
Merr

Cop

Desued
Milkers
Adjusted

 
    
   

  

Desued

MerrS
Cop

 
   

. ‘\

   

Young
.1. COW W
Purcnoseg/

Figure 7.13; Cousol Loop Diogrom Linking
Feedbock to Desired Milker Cool

162

Using the degree of error to change the goal would make
sense if the hypothesized farmer were responding to too few
Milkers by trying to increase the size of the operation and
not by altering management techniques. This formulation is
analogous to farmers responding to inadequate profits by
increasing herd size and not improving management
practices.

Two principle loops result from this structure; the
loop leading to purchases and a decrease in the gap and a
loop linking the gap to the goal. The distinction between
Desired Milkers Base and Desired Milkers Adjusted is
important. If Desired Milkers were not held back by a base,
as a gap begins to appear and the purchases of Young Cows
lags behind the increase in the Desired Milker goal, an
upward spiral would result. The system would be unstable
with the goal being chased but never caught by the actual
number of Milkers. The assumption is that the farmer will
remember that he has a basic goal and that this remains
invariant even if there is not enough Milkers.

Given this base, the time path of the simulation of
this model depends on the degree of delay in the purchases
of Young Cows relative to the time it takes to average the

experience of the Desired Milker Gap.

“in

163

7.5.1 Run IVA, B, and C, Analysis of Averaging Times for
Feedback

These three runs, summarized in one table (Table 7.4a),
explore the impact of different averaging times for the
Desired Milker Cap on the herd. One run was made for
averaging times of 6, 4, and 2 months. The simulations show
that the system is not unstable with any of these averaging
times, although the greatest degree of stability results
when 6 months is used. Figures 7.14a through c compare the

time paths of these three runs.

Table 7.4a: Analysis of Averaging Times for Feedback: 6, 4
and 2 months, Runs IVA, IVB and IVC

 

- Cull Rates: Standard

- Death Losses: Standard

- Desired Milkers: Initial 45 animals, changes according to
feedback

- Time to Purchase Young Animals: 2 months

- Time to Average Desired Milker Gap Ratio: 6, 4 and 2

months

Initial Values

of Variables Variable Final Values of Variables
6 4 2
months months months
30 Calves 21.3 21.3 21.3
45 Heifers 46.5 46.5 46.5
15 Young Cows 15.4 15.4 15.4
30 Cows 29.6 29.6 29.6
45 Milkers 45.0 45.0 45.0
N.A. Desired Milker Cap .92 .92 .92
0 Young Cow Purchases .92 .92 .92
0 DMG Ratio .02 .02 .02
45 Desired Milkers 45.9 45.9 45.9

Time to Reach Steady State: See text

164

In Figure 7.14a the averaging time is 6 months. A
slight dip in Milkers occurs at the start of the simulation
and the Desired Milker goal is forced upward slightly (see
the final value for Desired Milkers in Table 7.4a). In
comparison, decreasing the averaging time to 4 months
(Figure 7.14b) causes a little more oscillation, although
the steady state is identical to the previous run. In
Figure 7.14c, with an averaging time of 2 months even more
oscillation occurs and the system does not stabilize until
after 100 months. Still, once this system reaches steady
state it is identical to the previous two runs. The
instability that occurs when two months is used is also
evident in the quantity of Young Cows Purchased and Young
Cows Culled (Figure 7.14). The fact that the averaging
time for the ratio of the gap is the same as the delay on
the purchases of Young Cows is the likely source of this
instability.

At steady state, the milking herd with any of these
averaging times remains at 45 animals, and the Desired
Milker Adjusted Goal stays at approximately 50 animals.
The implication of the results from the runs in III were
that the goal would have to be raised above 45 if the herd
were going to reach 45. In contrast, these runs demonstrate
that the degree of steady state error can be used to adjust

the goal to reach the Desired Milker Base level.

 

DesiredMilkers 2 Milkers 3 CowsCulied 4 YoungCowsCulied

100 000
6.000

81.250
4.625

3.250

 

43.750 12—__12__gz_____
1.875 11:39 1‘- :9

 

 

25.000 1
""" I V V V 1 V "V V VVV I V V V V V V I "V
0500 0.0 90000 100.000 270.000

V

i
}
}
}
} 62 500
i
}
}
}
}

V 1
360.

DesiredMiikers 2 Milkers 3 CowsCulied 4 YoungCowsCuiIed

 

 

'rfiv" I v'v V

0500 0.0 90000 '

Y "Tfiw 'VVV‘.’

I V V V If' V V V V 1
180.000 270.000 360.

d
C"
C
u-
5
Cl
i
}.
‘P
V

2 Miikors 3 00145001100 4 YwanowsCuiiec

75 000
2 500

AAAALA

52.500
2.000

U1
0
O
O
C)
‘ A..-

 

 

m V ‘V v v v Y ‘7 1 v V ‘V 1 v V v

I I V V
0500 00 90 000 160.000
Time

. , v v
270 000 3601

Figures 7.14a-c: The Effect of Differing Averaging Times for
Feedback; 6, 4, and 2 months, Runs IVA, IVB, IVC.

166

7.5.2 Run IVD, Linked Goal, Constant SOP on Purchases,
Expand with Heifers

The runs in II were made with constant purchases of
Young Cows. Run IIE shocked this system by expanding the
herd with 45 Heifers and increasing Desired Milkers to 90
animals. As depicted in Figure 7.9a, Milkers lags behind
Desired Milkers for 17 months. What if this type of shock
was applied to the system when feedback from the performance
relative to Desired Milkers is incorporated?

Figures 7.15 a and b and Table 7.4b present the results
of this simulation. The increase in Heifers registers
immediately (Figure 7.15b). Although Young Cows and Cows

Table 7.4b: Endogenous Desired Milkers, Constant SOP on
Purchases, Expand with Heifers

 

- Cull Rates: Standard

- Death Losses: Standard

- Desired Milkers: Initially 45 animals, changes according
to feedback. Expands to 90 animals in month 60.

- Time to Purchase Young Animals: N.A.

- Time to Average Desired Milker Gap Ratio: 6 months

- SOP on Purchases: Constant, 2 animals per month

Initial Values Final Values
of Variables Variable of Variables

30 Calves 43.4

45 Heifers 91.5

15 Young Cows 40.8

30 Cows 50.6

45 Milkers 91.4

N.A. Desired Milker Gap -l.4

0 Young Cow Purchases 0.0

0 DMG Ratio 0.0

45 Desired Milkers 90.0

Months to Reach 90 Milkers after expansion: 16 months

 

a)

b)

167

 

 

 

 

 

 

 

 

 

1 Desireerlkers 2 Mnlkars 3 CowsCulled 4 YoungCowsCulled
1
7} 150000 _
2} 6.000 ; ‘
1
j 3
g} 118 750 1 H
i} 4.625 : ‘v—4—4-
.1 3 3-
1 1 12
4} 3.250
1 I
§} 56.250 4?:
1.875 *
i
3,} 25.000 1
3 VVVVVVV I VVV VVIVVVVVVVfi VVVVVVI V VVV V VVV I‘VVVVVVT V V VV“!
4} 0500 0.0 90000 180.00 270.000 360.
Time
1 Calves 2 Cows 3 YoungCows 4 Heifers
1
§ 110.000 1
4 1 4
2 1
3} 82.500 '1
“ 1
g 55 000 1 2
a ‘ 1 s 2
\_1.
' 1 l't-~‘._3;—'_—__—_
1
3;] 27.500 - 2
4 1 ‘
3, €
3 0'0 VVVVV T VVVVVV I V V V'I VVVfiVTVV V VVFVVV VVTVVVV VVIVVVVVfV—I
4 0.0 90 000 180.000 270.000 360.:
Tame

Figures 7.153-b: Endogenous Desired Milkers, Constant SOP on Purchases,
Expand with Heifers, Run IVD.

 

168
both increase, the Desired Milker Gap leads to a rapid
increase in Desired Milkers Adjusted (Figure 7.15a). But
the gradual accumulation of Heifers reduces the gap
relative to Desired Milkers Base (not depicted), so that
Desired Milkers Adjusted falls and Milkers rises. Sixteen
months after the expansion, Milkers reaches 90 animals, and
at this time Desired Milkers Adjusted has risen to 116
animals. Milkers are culled as both this variable and
Desired Milkers Adjusted gets pulled down to the base of 90

animals.

7.5.3 Run IVE, Linked Goal, Variable Purchases SOP, Expand
with Young Cows

Run IIIC explores the effects of expanding the herd
with the purchase of Young Cows when the farmer is making
monthly purchases of Young Cows to fill the Desired Milker
Gap. There was essentially no lag between Milkers and
Desired Milkers in this run, except for steady state error
(Figure 7.12a). Essentially the same result occurs in Run
IVE, when 45 Young Cows are purchased in month 60 and the
Desired Milker Gap feedback is linked to Desired Milkers.
The only striking result is that the number of Young Cows
Culled rises greatly during the period of adjustment of the
numbers of Milkers and Young Cows (Figures 7.16s and b).
Young Cow Purchases also remain high for a period

(approximately 24 months) after the expansion to fill any

169

Table 7.4c: Endogenous Desired Milkers, Variable SOP on
Purchases of Young Cows to 90; Run IVE

 

- Cull Rates: Standard

- Death Losses: Standard

- Desired Milkers: Initially 45 animals, changes according
to feedback. Expands to 90 animals in month 60.

5'1

- Time to Purchase Young Animals: 2 months
- Time to Average Desired Milker Gap Ratio: 6 months
- SOP on Purchases: Variable, linked to Desired Milker Gap

Initial Value Final Values
of Variables Variable of Variables

30 Calves 42.6

45 Heifers 93.1

15 Young Cows 30.8

30 Cows 59.1

45 Milkers 90.0

N.A. Desired Milker Gap 1.8

0 Young Cow Purchases 1.8

0 DMG Ratio 0.2

45 Desired Milkers 91.9

Months to Reach 90 Milkers after expansion: Essentially no
time to attain the steady state error of 1.8 animals.

 

developing Desired Milker Gap while Young Cows and Cows
balance out. Young Cow Purchases eventually settles down to

a steady state level (Figure 7.16c)

a)

b)

C)

I DesideHkors

% 150.000
4 6000
1

§} 118.750
4} 4625
I

§} 37.500
4} 3.250
1

g} 56 250
4} 1.875
1

3} 25000
4} 0 500
1 Calves

5

3 110000
4

i»

3 02.500
4

1

3 55 000
4

i»

3 27.500
4

1

3 00
4

I YCWPurchases
I

3 15.000
‘ 6000

7.500
3.250

%} 3.750
4 1075

‘ 00
31

170

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2 Maikers 3 CowsCulled 4 YoungCowsCulled
.1
d
: F‘ 1:12:
‘ 3— '5
1
1
1
1::
J”
1
1
V V V V V V I V V V V V V I V VVV VVV V'IVV VVVVI V V V 1
0.0 90.000 180.000 270.000 360.
2 Cows 3 YoungCows 4 Heifers
1 4‘4.—
1
1
1
1 2 2
I11
1
1F.— 0 C
1 3 3
q
1:.
1
1
1
V VVVVIVVVV V I VVV VV VVVVV I V VVV V V V IVVVVVVV VVVV V
0.0 90.000 180.000 270.000 360
2 HeifersFRep'ace 8 YoungCowsCuued 4 CowsCulled
q
1
4
1
1
1 L 3
‘ 4..~____4__
1 2 2
1
1 3
1"_’j 1......
1 '
VVVVVV I VVV V T V VVV IVV VVVVI V V V V V T VVVVVVIVVVV VV1
0.0 00.000 180.00 270.000 360.
Time

Figures 7.16a-c: Endogenous Desired Milkers, Variable SOP on
Purchases of Young Cows to 90, Run IVE.

171
7.6 Summary of Simulation Results

When the model is simulated without the purchases of
replacement stock and with "baseline" death losses and cull
rates (Run 1), it takes approximately thirty years for
Milkers to reach Desired Milkers. The lag is caused by the
inadequate levels of on the farm replacement stock.
Eventually enough stock is generated internally to meet the
goal. If the average age of Cows is raised to 72 months the
Desired Milker goal can be met in 5 years. These results
indicate the need for purchases of replacements if the
farmer cannot maintain or house sufficient numbers of
Heifers and Calves.

The purchase of two Young Cows a month enables the
farmer to meet the Desired Milker goal although some over
purchasing of animals occurs (Run II). Attempting to expand
to 90 milkers by purchasing Young Cows leads to a 15 month
lag until the new Desired Milker goal is reached. Expansion
by the purchase of Young Cows and Heifers results in no
lag. Expansion by the purchase of only Heifers leads to a
17 month lag, but there are no excess milking animals and
therefore no culling of unnecessary Milkers.

Rather than purchasing the same number of Young Cows
for replacements each month, some farmers may base their
purchases on the gap between the number of Milkers and the
Desired Milkers goal. This SOP, used in Run III, leads to

an inability to meet the Desired Milker goal. The gap never

 

172
disappears in steady state and represents two percent of
Desired Milkers. The steady state gap is attributable to
the time it takes to perceive and arrange for the purchase
of the Young Cows. Increasing this delay in perception does
not alter the steady state consequences. Under the shock of
expansion, Milkers lags behind Desired Milkers by 15 months
which is the time it takes to accumulate an adequate
quantity of replacement stock.

The gap between Desired Milkers and Milkers is used to
alter Desired Milkers in Run IV, representing an attempt by
the farmer to overcome the two percent steady state error
when purchases are variable. Although there is an increase
in instability if the time to average the gap information
is too small, the herd behaves very similarly to the runs
without the endogenous Desired Milker goal; a steady state
gap of two percent still remains. Endogenizing this goal by
linking it to the gap has very little new impact on the

performance of the system.

7.7 Conclusions
The introduction to this chapter asks three questions:
1) Can STELLA be used to model a specific part of the
dairy operation?

2) Can this model rely on the use of SOP's to guide the
operation?

3) Could a model constructed in STELLA and oriented
towards a specific problem be of use to agricultural
economists?

173

Using the criteria of model viability-validity discussed in
the introduction as the measurement standards, the runs in
I indicate that the model is viable and valid. In turn, to
whatever extent these criteria are considered appropriate,
the answers to the first two questions can be considered
positive.

The answer to question three is as subjective as the
answers to the previous two, and depends on the person
viewing the evidence. The only conclusion that really can
be drawn from the results of Runs II-IV is that different
herd management strategies under the conditions of stress
from significant expansion of the milking herd will lead to
specific and distinct results.

The implication, although not rigorously conclusive, is
that STELLA is useful as a problem solving tool. As for
the simulated response of the operator to the performance
of the operation relative to the Desired Milker Goal,
models of an entire range of management goal oriented
behavior should be possible to construct. Such models,
although useful for problems involving only one management
goal/response, become particularly useful when a system of
interacting management practices need to be explored in

order to understand the management problem.

HA ER IGHT
SUMHA Y AND CONCLUSIONS“ BEHAVIORA MODEL AND
H O X ON

8.1 The Primary Research Questions

The primary questions asked in this research are
answered in Chapter Five. It was found, although not
conclusively, that simulation models can be used to
represent the standard operating procedures of firms in such
a way so as to recreate much of their business behavior.

The usefulness of these models for performing sector level
policy analysis was found to be much less promising. While
sector level models using behavioral theory have been
constructed, the time and expense needed for their
construction may be prohibitive. The problem lies in the
lack of a universally applicable theory of economic
behavior, so that a new theory needs to be created for every
system considered. Some promise was found in the system
dynamics models of Meadows, that combined the aggregative
strengths of neo-classical theory with some
behavioral/system dynamics constraints to create an
apparently useful model of the hog industry.

The behavioral, systems simulation model of a Mid-
Michigan dairy farm was never completed in this research.
The time and effort that went into this model was
significant, and contributed to the conclusion that although

detailed behavioral models of a single-owner farm operation

174

175
are possible, building such models for sector level market
analysis may require more resources than are practical for
most situations.

A systems simulation model of the herd management
portion of a dairy operation, constructed with the systems
dynamics language STELLA, permitted the examination of a
number of different management issues that may face a dairy
operator. The results indicate that STELLA is a useful tool
for the examination of on-farm management problems where
farmer behavior needs to be represented by standard
operating procedures not completely consistent with the

assumptions of neo-classical theory.

8.2 A Return to the Issue of Maximization

Chapters One and Two distinguish between neo-classical
and behavioral economic theory by identifying the apparent
differences in their assumptions about maximizing behavior.
But a closer examination of the behavioralist position
indicates that this distinction centered on maximization may
be inappropriate and perhaps not even the relevant and
important issue.

The two assumptions common to theories of market
economies cited at the start of Chapter One are that: 1)
people make choices that are intended to have favorable
impacts on their self-interests; and 2) that people are

rational. The third assumption offered to distinguish neo-

176
classical theory is that people choose courses of action
that lead to a maximum level of utility (or profits). This
particular distinction may be artificial because all three
of these assumptions can be summarized into one reasonable
statement that both neo-classicists and behavioralists could
probably agree upon: "people are, for the most part, doing
their best to get the best possible situation for
themselves".

Assuming that this statement is an accurate
representation of the intuitive root of the maximization
assumption, the distinction between the neo-classicists and
the behavioralists really centers on the way these two
schools define: l) the means and procedures that a person is
assumed to utilize in order to get their best or desired
outcome; and 2) the types of outcomes that a person desires,
the outcomes that they would consider "best".

The assumption that people are seeking to get the best
they can explains nothing without also defining how people
decide what the best is and how they go about getting it.
For example, a person has before them two choices, 1 and 2.
You cannot predict their choice by saying that they will
make the selection that will give them the best possible
outcome, or that will maximize their utility. More
information is needed on how they perceive 1 and 2, and how
they might value these different choices. And you have

essentially explained nothing if, after they have selected

177
2, you say that they selected it in order to maximize their
utility, that they were seeking to get the best they could.

Such a contention amounts to no more than saying that
people do what they do because they do it. Yes, of course,
we know that they do "it", that they are "seeking the best",
but what we need to know is how they go about seeking the
best and what defines as best the particular outcomes being
sought. Not only is the contention that people do it
because they do it incorrect, it is also not meaningful
because it tells us nothing at all about what actually
happened and why.

Neo-classical market/price theory does not sat that
"people do it because they do it”, although some arguments
defending the maximization assumption adopt this type of
logic18. Neo-classical theory would be useless if it did
so. On the contrary, the theory defines peoples' goals and
their abilities to pursue them simply enough to allow
maximizing or best seeking behavior in a competitive economy
to result in the familiar condition of equilibrium. As a
result, under these assumed conditions the theory has a
great deal to say about likely aggregate market reactions to
changes in costs, supply, incomes and numerous other
factors. The theory and its implications are generally
consistent with the observed phenomenon of consumers wanting
to purchase less in aggregate as prices rise and producers

wanting to produce more (and vice versa).

178

Neo—classical theory's predictive capabilities may be
due, in part, to consumer and producer SOP's that include
enough neo-classical price response characteristics to make
aggregate, average supply and demand curves behave neo-
classically. But for explaining and predicting an
individual's decision process, or how they will act as
individuals in their own unique production and consumption
circumstances, the theory loses a great deal of its
relevance. For even though under these circumstances people
are likely doing their best to get the best they can for
themselves, what really matters are the factors determining
the best that the person can do and the person’s goals.

Cyert and March clearly share this same view in the
passage from their book cited above in Chapter Five. The
crucial points they make are that:

"... many of the attacks on the (neo-classical) theory

of the firm are not so much proper critiques of

existing theory as they are suggestions for the

development of a theory appropriate to a different set

of questions."
and that:

"... we should distinguish between a theory of micro-

behavior, on the one hand, and the micro-assumptions

appropriate to a theory of aggregate economic behavior

on the other." [9, pages 15-16]
As discussed in Chapters Two and Three, the Carnegie school
has taken an entirely different approach to studying these

factors. Recognizing the cognitive limitations of

individuals, and their complex decision environments, the
/

179
school has been more interested in developing a theory of
individual decision-making processes that can bear up to
closer empirical scrutiny than neo-classical theory. This
interest has lead them to inquire into what actually happens
at the individual's level.

Although a behavioral theory of markets is desired by
the Carnegie school, there appears to be no willingness to
achieve this by sacrificing a behavioral theory that can
explain and predict individual behavior. As a result,
behavioral theory has a great deal to offer the study of
decision making processes in individual and group
situations, but it still has yet to develop a comprehensive

and consistent theory of markets and prices.

ENDNOIES

1. This simplified caricature of neo-classical theory
ignores some of the other important, and more rigorous,
assumptions that permit its use in the analysis of markets.
For instance, an assumption is generally made about the
degree of homogeneity of the production function. Yet these
three assumptions, plus using profits as the proxy for
utility, permits a reasonable comparison to a competing view
of human economic behavior which follows in the body of the
text.

2. See Chapters Two and Three for citations and discussion
of the work of Herbert Simon and others from the behavioral
school.

3. See Chapter Five, pages 82-83, for a brief discussion of
this issue and citations.

4. Milton Friedman's argument, that economic actors only
need to behave as if they conformed to these assumptions
[cite?, throughout], is often offered in response to

Leibenstein's question.

5. The organization of this thesis does not conform to the
standard "modernist" structure (ie, Assumptions and Theory,
which lead to, Deduction of the Model, Testing of the Model,
and a Summary of Results and Policy Implications). The
reason for not using the modernist organizational structure
(see McCloskey, D.M., Economical Writing, V. 24, No. 2
(1985), and by the same author, The Rhetorig of Economics,
University of Wisconsin Press, Madison (1985)) is that it
would be fundamentally misleading to present these results
as such. The intent is to add to the value of this research
by bringing some light to the process by which the
conclusions were achieved.

6. There is no one style of neo-classical analysis, and
over time a number of significantly different styles and
approaches have emerged. For example, the industrial
organization (1.0.) approach, with its utilization of
structure, conduct and performance as its primary analytical
framework has developed into a major alternative means of
viewing markets and industries.

I.O. analysis is not homogeneous either. A major body
of its practitioners, including those like Bain who
initiated this type of analysis, view the elements of
structure to be limited to conditions of market
concentration (analogous to the conditions of competition,
monopoly, monopolistically competitive and oligopoly in neo-

180

181

classical theory) and view conduct to be essentially
limited to the traditionally neo-classicist profit and or
utility maximization. As such, this approach to 1.0.
analysis is clearly neo-classical in its roots.

1.0. has also developed into other areas, including
those that are distinctly behavioral, utilizing the
flexibility provided by the structure, conduct and
performance paradigm to consider factors other than
concentration as affecting structure and performance, and
considering conduct other than profit maximization.

This discussion is provided to highlight some of the
short-comings of the characterization of neo—classicism
presented here. Still, this view is not entirely unfair -
the emphasis on profit maximization does dominate,
particularly when the neo-classicist attempts to
econometrically model a market system.

7. The rules of thumb for standard operating procedures
discussed in this section are derived from conversations
with H. Bucholtz, Dairy Scientist, Michigan State
University, and 8.8. Nott, Agricultural Economist,
Michigan State University, as well as dairy farm managers.

8. From conversations with a farm credit officer.
9. From conversations with Bucholtz.
10. From conversations with Bucholtz.

11. From unpublished data available from the Michigan Dairy
Herd Improvement Association, East Lansing, Michigan.

12. From Conversations with Bucholtz.
13. From conversations with Bucholtz.
14. From conversations with Bucholtz.
15. From conversations with Bucholtz.
l6. Bucholtz.

17. From conversations with Bucholtz and J.R. Black,
Agricultural Economist, Michigan State University.

18. In general, the arguments tend to hold that no matter
what people do, no matter what choices they make, they can
be seen as maximizers if we take enough of their special
circumstances into account. For example, if someone is not
maximizing profits, then they are maximizing profits and
leisure. And if not these two, then profits, leisure and
psychic costs, and on and on. These arguments are basically

182

saying that if you can identify what it is that people are
doing it is possible to find some way to characterize it as
maximizing. As Leibenstein points out, all meaning (and
usefulness) of the word maximizing is lost if it defines all
behaviors. Maximizing means nothing if ”non-maximizing" is
not permitted to have the opposite meaning. It would seem
beneficial to all if these tautological arguments in support
of maximizing could be dropped. Economists could then
proceed to explore the important and more interesting work
of deciding for each of the theoretical problems faced the
appropriate set of micro-level behavioral assumptions. (See
Leibenstein for further discussion of these issues [18,
p494-496].)

APPENDICES

X N ' SY EH YNAHICS E U TION AND F 0W DIAGRAMS

Systems Dynamics simulation models have three basic
types of variables that are used to represent the functions
of systems. The first and generally most important are the
stock variables (also referred to as state or level
variables). The stock variables are the most important
variables in a systems model because they represent the
portions of the system that are actual conditions or states
in the world. All of the other variables serve to change
these states, or support the variables defining these
changes in states.

Stock variables are the parts of the system that
accumulate or decummulate over time. In physical models,
the stock variables signify the physical materials that have
inertia, that take time in order for their condition to
change. A good example from the dairy model is the stock of
animals representing the milking herd. Other stock
variables in the dairy model might be the stock of feed
rations held on the farm, or the stock of farm machinery or
the stock of working capital held by the operator in the
bank. Stock variables are not limited to representing only
physical objects. For example, a person's perceptions of a
situation do not change instantaneously, but gradually over
time. A stock variable can be used to represent this state

of mind.

183

184

The variables defining directly how the states change
are the Late variables. The rate variable defines how much
the stock variables are growing and or decreasing in size
over time. In the dairy model examples of important rate
variables are "Calves In" and "Calves Out". These variables
represent how many calves are entering the stock of calves
over time (due to births and or purchases) and how many are
leaving (due to maturation or death loss).

The final class of variables is the auxiliaries, which
are used to represent information or relationships in the
model that do not take any time to change. An auxiliary
variable is generally used in conjunction with other
auxiliaries and the stock variables to determine what is
taking place in the rate variables. An example of
auxiliary variables in the dairy model are all of the death
loss variables, or the variable keeping constant track of
the number of milking animals in the herd (the sum of the
stock variables Cows and Young Cows). The time variables
are all auxiliary variables that define the length of time
it takes for various processes to take place (ie the time it
takes to purchase a heifer once the decision has been made
to make the purchase).

Figure Al is a diagram of these three types of
variables in a configuration commonly seen in a systems

model.

185

Auxiliary2

 

 

 

Stock
Rateln RateOut
Auxiliary 1
Figure Al: The three classes of variables in a systems

model

Ammo: Hzm2m0<z<z 55: mm; no Emu/:9 mUHrész mZmHmwm < "03H. xHQzumnEdg

 

 

 

 

 

 

 

186

187

DEF NITION 0 THE VARIABL S N THE MODEL
State Variables:
CALVES: Animals from age 0 to 4 months.
COWS: Milking animals that have freshened a second time.

DMGR-AVG: Desired milker gap ratio, average; a state
variable that serves as the source of feedback for the
system, and is used to adjust the DESIRED MILKERS goal.

HEIFERS: Animals of age 5 to 27 months.

Y-CW-PURCHASES: Young cow purchases; a state variable that
represents the purchases by the operator of young milking
animals to make up for any shortage of milking animals, as
defined by the desired milker gap (DES-MILKER-GAP). This
variable may or may not be used in the model, depending on
the scenario.

Y-HEIF-PURCHASES: Young heifer purchases; a state variable
that represents the purchases by the operator of heifers for
replacement to make up for any shortage of these animals, as
defined by HEIFERS-NEEDED. This variable may or may not be
used in the model, depending on the scenario.

YOUNG-COWS: Milking animals who have only freshened once,
and are in the first lactation.

Rate and Auxiliary Variables:

B-CALF-RATE: Bull calf rate, where half of all of the
calves born are assumed to be males.

BIRTH-RATE: The rate at which calves are born. It is
assumed that a cow, young cow or heifer will have one calf
every 13 months.

CALVES-FOR-REPLACEMENT: The calves that are available to
serves as replacements to the milking herd.

CALVES-IN: A rate variable defining the rate at which calves
are being born.

CALVES-NEEDED: The number of calves that are needed by the
operation simply to maintain the herd at a steady state
size, given the assumed death losses from cows, young cows
and heifers.

188

CALVES-OUT: A rate variable defining the rate at which
calves leave the calf stage and become heifers.

COWS-CULLED: A rate variable defining the rate at which
cows are culled form the herd.

CULL-RATE-STANDARD-COWS: The rate at which cows are removed
from the herd due to their reaching the assumed age beyond
which milk production declines.

CULL-RATE-STANDARD-Y-COW: The rate at which young cows are
removed from the herd due to assumed standard death losses.

CULL-RATE-VARIABLE-COWS: The number of cows to be culled in
excess of the standard rate, the number being given any
excess milking animals.

DEATH-LOSS-CALVES: The assumed standard rate at which calves
die due to disease, illness or other health problems.

DEATH-LOSS-COWS: The assumed standard rate at which cows
die due to disease, illness or other health problems.

DEATH-LOSS-HEIFERS: The assumed standard rate at which
heifers die due to disease, illness or other health
problems.

DEATH-LOSS-Y-COW: The assumed standard rate at which young
cows die due to disease, illness or other health problems.

DESIRED-MILKERS: The number of milking animals that the
operator desires for the farm.

DESIRED-MILKER-GAP: The difference between the current
number of milkers and that number desired by the operator.

DES-Y-HEIF-GAP: The difference between the number of
heifers available for replacement and that number needed by
the operation to maintain the herd at a steady state size.

DMG-RATIO: A ratio between DES-MILKER-GAP and DESIRED-
MILKERS, used to adjust the state variable, DMCR-AVG (see
above).

HEIFERS-F-REPLACEMENT: The number of heifers available to
serve as replacement animals for the herd.

HEIFERS-IN: A rate variable defining the rate at which
calves are entering the heifer stage.

HEIFERS—NEEDED: The number of heifers needed to maintain

189

the herd at a steady state, given the standard assumed death
losses and cull rates of milkers.

HEIFERS-OUT: The rate at which heifers leave the heifers
stage to become either young cows, or are lost due to death.

MAT-RATE-CALVES: The rate at which calves are assumed to
mature and reach the heifer stage.

MAT-RATE-HEIFERS: The rate at which heifers are assumed to
mature and reach the young cow stage.

MAT-RATE-Y-COWS: The rate at which young cows are assumed to
mature and reach the cow stage.

MILKERS: The number of milking animals (cows + young cows).

RETE-RATE-STEADY-STATE: The steady state retention rate
used to calculate CALVES-NEEDED and HEIFERS-NEEDED.

SALE-B-CALVES: The number of bull calves sold.
SALE-CULL-COWS: The number of cull cows sold.

SALE-H-CALVES: The number of heifer calves sold (that
number in excess of those needed for replacements).

SALE-Y-COWS: The sale of young cows (those in excess of the
number needed for replacements).

SOP-Y-COW-PUR: An SOP defining the monthly purchases of
young cows. The variable may or may not be used in the
model, depending on the scenario.

TIME-AVG-DMGR: The assumed number of months over which the
moving average for the desired milker gap ratio is
calculated.

TIME-PUR-Y-CW: The time it takes the farmer to purchase
young cows once the decision has been made to make the
purchase. The value of this variable changes depending on
the scenario.

TIME-PUR-Y-HEIF: The time it takes the farmer to purchase
heifers once the decision to make the purchase has been
made.

Y-CW-PUR-DES: Young cow purchases, desired; the number of
young cow purchases the farmer would like to make.

Y-CW-PUR-GAP: The difference between the number of young

190

cow purchases the farmer desires and the number of purchase
the farmer is currently in the process of completing.

Y-CW-PUR-IN: A rate variable defining the rate at which
the purchases of young cows occurs.

Y—HEIFERS-AVAILABLE: The number of calves that are actually
used as replacement animals and entered into the heifer
stage.

Y-HEIF-PUR-DES: The number of young heifers the farmer
would like to purchase.

Y-HEIF-PUR-GAP: The difference between the number of
heifers the farmer would like to purchase and the number
that the farmer is currently in the process of purchasing.

Y-HEIF-PUR-IN: A rate variable defining the number of
heifer purchases the farmer will commence each month.

YOUNG-COWS-CULLED: A rate variable defining the rate at
which young cows leave the herd, due to death loss.

YOUNG-COWS-IN: A rate variable defining the rate at which
heifers are entering the young cow stage.

YOUNG-COWS-TO-HERD: A rate variable defining the rate at
which young cows are leaving the young cow stage and
entering the cow stage.

BI BLIOGRAPHY

BIBLIOGRAPHY

[l] Balderston, F.E., and A.C. Hoggatt, "Simulation Models:
Analytic Variety and the Problem of Model Reduction," in
A.C. Hoggatt and F.E. Balderston (Eds.), SYMPOSIUM ON
SIMULATION MODELS. Cincinnati: South-Western Publishing
Company, 1963.

[2] Balderston, F.E., and A.C. Hoggatt, SIMULATION OF
MARKET PROCESSES, Berkeley: Institute of Business and
Economic Research, 1962.

[3] Brown, H.L., S.B. Nott, "Business Analysis Summary For
Specialized Michigan Dairy Farms, 1979 Telfarm Data".
Michigan State University Agricultural Economics Report #
379: East Lansing, Michigan; 1980.

[4] Brown, H.L., S.B. Nott, "Business Analysis Summary For
Specialized Michigan Dairy Farms, 1981 Telfarm Data".
Michigan State University Agricultural Economics Report #
413: East Lansing, Michigan; 1982.

[5] Brown, H.L., S.B. Nott, ”Business Analysis Summary For
Specialized Michigan Dairy Farms, 1982 Telfarm Data".
Michigan State University Agricultural Economics Report #
439: East Lansing, Michigan; 1983.

[6] Clarkson, G.P.E., and H.A. Simon, "Simulation of
Individual and Group Behavior," AMERICAN ECONOMIC REVIEW,
50, 1960.

[7] Cohen, K.J. and R.M. Cyert. THEORY OF THE FIRM:
RESOURCE ALLOCATION IN A MARKET ECONOMY. Second Edition.
Englewood Cliffs, N.J.: Prentice-Hall [1965] 1975.

[8] Cyert, R.M., and C.L. Hendrick, "Theory of the Firm:
Past, Present and Future; An Interpretation," JOURNAL OF
ECONOMIC LITERATURE, 10(2), pages 398-412, June 1972.

[9] Cyert, R.M., and J.G. March, A BEHAVIORAL THEORY OF THE
FIRM. Prentice Hall: Englewood Cliffs, N.J., 1963.

[10] Good, D., "Potential Impact of Environmental Pollution
Abatement Alternatives on the Michigan Dairy Farming
Industry". PHD Thesis, Dept Ag Economics, Mich. State
Univ., East Lansing, Michigan. 1972.

[ll] "Equilibrium Analysis of Income Improving Adjustments
on Farms In the Lake States Dairy Region, 1965," University

of Minnesota Ag Experiment Station, Technical Bulletin 246,
October 1963.

191

192

[12] Ferris, T.A., and J.W. Thomas, "Management Factors
Influencing Calf Mortality and Blood Immunoglobin Levels in
Michigan Dairy Herds". Research Report #271, Michigan State
University Agricultural Experiment Station: E. Lansing,
Michigan; 1975.

[13] Forrester, J.W., INDUSTRIAL DYNAMICS. Cambridge, Mass.:
Wright-Allen Press, 1961.

[14] Heiner, R.A., "Imperfect Decisions In Organizations
Toward a Theory of Internal Structure," revised for the
JOURNAL OF ECONOMIC BEHAVIOR AND ORGANIZATION, 1986.

[15] Heiner, R.A., "The Necessity of Delaying Adjustment of
Imperfect Decisions and Expectations," Economics Department:
Brigham Young University; Provo, Utah, February 1986.

[16] Kelsey, M.P., A. Johnson, "Business Analysis Summary
For Specialized Michigan Dairy Farms, 1978 Telfarm Data".
Michigan State University Agricultural Economics Report #
359: East Lansing, Michigan; 1979.

[17] Knoblauch, W., "Level and Variability of Net Income for
Selected Dairy Business Management Strategies". PHD Thesis,
Dept Ag Economics, Mich. State Univ., East Lansing,
Michigan. 1976.

[18] Leibenstein, H., "A Branch of Economics is Missing:
Micro-Micro Theory," JOURNAL OF ECONOMIC LITERATURE, Vol
XVII, pages 477-502, June 1979.

[19] Lyneis, J.M., CORPORATE PLANNING AND POLICY DESIGN: A
SYSTEM DYNAMICS APPROACH. Cambridge, Mass. MIT press, 1980.

[20] Meadows, D.L., DYNAMICS OF COMMODITY PRODUCTION
CYCLES. Cambridge, Mass.: MIT press, 1970.

[21] Nott, S.B., V.L. Brodek, "Business Analysis Summary
For Specialized Michigan Dairy Farms, 1984 Telfarm Data".
Michigan State University Agricultural Economics Report #
472: East Lansing, Michigan; 1985.

[22] Nott, S.B., S. Reynolds, "Business Analysis Summary
For Specialized Michigan Dairy Farms, 1983 Telfarm Data".

Michigan State University Agricultural Economics Report #

456: East Lansing, Michigan; 1984.

[23] Naylor, T.H. and J.M. Vernon, MICROECONOMICS AND
DECISION MODELS OF THE FIRM. Harcourt, Brace and World: New
York, 1969.

‘K

193

[24] Pugh, Alexander, L. III. DYNAMO USER'S MANUAL. MIT
Press: Cambridge, Ma., 1983.

[25] Richmond, 8., A USERS GUIDE TO STELLA. (Draft). High-
Performance Systems, Inc.: Hanover, N.H., 1985.

[26] Roberts, Nancy, David F. Anderson, Ralph M. Deal,
Michael S. Garet and William A. Shaffer. INTRODUCTION TO
COMPUTER SIMULATION: THE SYSTEM DYNAMICS APPROACH. Addison-
Wesley: Reading, Ma., 1983.

[27] Simon, H.A., "A Behavioral Model Of Rational Choice,"
QUARTERLY JOURNAL OF ECONOMICS, 69: 99-118, 1952.

[28] Simon, H.A., REASON IN HUMAN AFFAIRS. Stanford
University Press: Stanford, Ca., 1983.

[29] Stoll, T.L., "An Analysis of the Effect of Expansion on
Cash Flow, Management Income, and Various Management
Factors", Masters Thesis, Dept. Dairy Science, Mich State
Univ., East Lansing, Michigan. 1974.

[30] Tversky, A., and D. Kahnemann, "Judgement Under
Uncertainty: Heuristics and Biases," SCIENCE 185: 1124-31,
1974.

"1.11111111111111