DEFAULT AND LIQUIDITY IN THE MANAGEMENT
OF THE BNDA LENDING PROGRAM IN IVORY COAST
BY
BERNADETTE DIA
Dipl •• Ecole Nationale Superieure Agronomique. 1978
M.S•• University of Connecticut. 1981
THESIS
Submitted in partial fulfillment of the requirements
for the degree of Doctor of Philosophy in Agricultural Economics
in the Graduate College of the
University of Illinois at Urbana-Champaign. 1986
Urbana. Illinois

Hi
DEFAULT AND LIQUIDITY IN THE MANAGEMENT
OF THE BNDA LENDING PROGRAM IN IVORY COAST
Bernadette Dia. Ph.D.
Department of Agricultural Economics
University of Illinois at Urbana-Champaign. 1986
Default influences liquidity as well as being influenced by liquidi-
ty.
While such effects have been discussed in previous studies on li-
quidity management by farm borrowers in developing countries.
they were
not explicitly determined in the empirical analyses.
Instead these
studies assume that a net lending cost of zero combined with an all-cash
loan disbursement will bring the borrower to value the credit reserves of
the lending program and hence repay his debt as much as he can.
This
study expands on previous work by empirically testing this hypothesis for
the case of the "Pret de Faisance Valoir Normalise" (PFVN). a BNDA credit
program for individual medium-sized Ivorian farmers.
In our work we modify a Liquidity Specified Linear Programming
(LSLP) model to include default variables and expand it over a period of
two years with a 4 season-specification per year.
We validate this
default-augmented form of the LSLP model with data gathered in a survey
of farmers from BNDA's center district Bouake'.
The model is then used to
determine the conditions for a zero net lending cost by varying (1) the
cost of default.
(2) the interest rate for BNDA loans.
(3) the credit
limit for BNDA loans.
and
(4)
the cost of delinquency.
It is also
modified to include liquidity management vectors for BNDA credit in order
to determine the combined effect of a zero net lending cost and an all-
cash disbursement of BNDA loans.
The net lending cost for BNDA can be reduced to zero by increasing
(1) the default cost to 2.40 CFAF. (2) the interest rate to 29%. and (3)

iv
the loan limit by 30%.
Gains to the borrower also are reflected in a
larger value of the objective function and in cash available.
I f in
addi tion to these three condi tions BNDA a d o px s a policy of an all-cash
loan disbursement.
it earns some profit from
its operation.
But while
the borrower still gains from
a larger value of the objective function.
cash available decreases.
This decrease. however. is offset by higher
credit reserves.

v
AO<NOWLEDGEMENTS
Several people contributed to the achievement of this study.
and I
wish to express sincere appreciation to all of them.
First and foremost.
I would like to thank Professor Chester B. Baker. my advisor and director
of the study.
for his intellectual stimulation.
guidance.
encouragement
and patience.
I would like also to thank the other members of my commit-
tee. Professors George G. Judge. Peter J. Ba r ry , and Dr. Aziz Bouzaher
for their invaluable comments and suggestions.
I am especially grateful to Dr. Emmanuel R. Kamgnia for his constant
support and patience in guiding me in the area of Numerical Analysis and
sharing his editorial skills.
The data used in this study were collected
by Dr. Kinimoz R. Yabile to whom I am deeply indebted for granting me the
permission to use them.
I
am also indebted to colleagues at
the
C.I.R.E.S. of Abidjan for their invaluable advice in the choice of the
topic for the study.
I would like also to thank the Ecole Nationale Superieure Agronom-
ique (E.N.S.A.) d'Abidjan for granting a leave of absence to pursue the
P~D. studies in the United States.
My appreciation is also extended to
the USDA for its financial support during the last year of the study.
In addition.
special thanks are extended to Professors Jane and
Raymond Leuthold. and Jean and John Due and families for their warm
hospitality and constant moral support.
I am also grateful to many
friends and especially fellow Ivorian students for their encouragement
and friendly cooperation throughout the study.
I am especially thankful
to my mother Suzanne Achy and my father Mamadou A. Dia for their love.
understanding and care.

vi
Finally I wish to dedicate the thesis to my son Armand ~
M. Amous-
sou who has been patiently waiting for a mother love and care for 8
years.

vii
TABLE OF CONTENTS
CHAPTER
ONE
INTRODU CTION • . . . . . . . . . . . . . . . . .
1
1.1
Effect of Default on Lending Institutions. • • • •
2
1.2
Government Supported Credit Program in Ivory
Coast:
the "Banque Nationale pour le Developpe-
ment Agricole" (BNDA)
• • • •
• • • •
3
1.3
Default Management in BNDA
• • • • •
• • ••
9
1.4
Research Objectives. • • • • • • • • • • • . • ••
10
1.5
Organization of the Thesis
• • • • • • • • • • • •
11
TWO
LITERATURE REVIEW
14
2.1
Behavioral Studies
15
2.1.1
Ames' Study [2]
15
2.1.2
Pradhan and Sharma's Study [49]
17
2.1.3
Octavio's Study [44]
• • • •
20
2.1.4
Best's Study [14]
• • • • •
21
2.1.5
Montiel's Study [41]
. • • •
23
2.2
Policy-Oriented Studies • •
26
2.2.1
Liquidity Management and Small Farm
Borrowers' Attitude
26
2.2.2
The Simulation Study of Rosegrant and
Herdt [54]
30
THREE
METHOD OF ANALYSIS • •
35
3.1
Description of the Liquidity Specified Linear
Programming Model • • • • • • • • • • • • • •
35
3.1.1
Theoretical Framework
• • • • • • • •
35
3.1.2
The Liquidity Specified Linear Programming
Model
.
• • • . •
. . . . .

.
.
.
• .
40
3.2
Incorporating Delinquency and Default in the LSLP
Model • • • • .
.
.
.
.
.
.
.
.
.
.
.
• •
.
.

47
3.2.1
Specification of Delinquency and Default
Activities • • . .
. • . . • . .
. • • •
48
3.2.2
Lenders' Response to Delinquency and
Def aul, t
.
.



.
.
.
.

.
.
51
3.2.3
Liquidity Effect of Delinquency and
Default
.
.
.
.
.
.
.
.
. .
.
. . .
52
3.3
The Data






.

.


.


.
53
3.4
Characteristics of Bouake Region
54
FOUR
MODEL
62
4.1
An Overview
. . . .
. . . . 62
4.2
Objective Function
65
4.3
Constraints. • • • •
65

viii
TABLE OF CONTENTS
(continued)
CHAPTER
Page
4.3.1
Production. Marketing and Consumption
Constraints
• .
. • . • . . . .
.
. .
.
• •
70
4.3.2
Financial Constraints
• • • • • • • •
72
4.4
Activities
.
.
. .
.
.
. .
. . . .
. .
.
. . .
76
4.4.1
Production. Consumption and Marketing
Activities • • • • • •
76
4.4.2
Financial Activities. • • • • • • • • •
81
FIVE
MODEL RESULTS AND VALIDATION.
90
5.1 J'lodel Resul ts • • • • • • • • • • • • • •
90
5.1.1
Basic Optimal Plan.
• • • • • •
90
5.1.2
Marginal Values
• • • •
• • • •
95
5.1.3
Effects of Non-Optimal Variables • • • • • • 104
5.2
Validation of the Model • • • • • • • •
• 107
5.2.1
Verification of the Mathematical Formula-
tion of the Model
• • • • •
• 107
5.2.2
Sensitivity Analysis • • • • • • • • • • • • 108
5.2.3
Evaluating the Conformity of the Model
Output With the Observed Values
• • • • • • 114
SIX
ANALYSIS OF POLICY REFORMS •
• • 118
6.1
Ove rv.i.ew


• •


• • • •

• 118
6.2
Description of the Policy Variables and the
Model Variants
· · · · ·
· · · · · · · 119
6.2.1
Default-Penalty Cost ·
· · · · 119
6.2.2
Interest Rate
· · · ·
· 121
6.2.3
Amount of the Loan ·
· · · ·
· 123
6.2.4
Delinquency-Penalty Cost · · · · ·
· 123
6.2.5
Use of BNDA Credit Reserve as a Source of
Liquidity
· · · ·
· · · · · ·
· 124
6.3
Model Variation Results ·
· · · · · ·
· 125
6.3.1
Effects of Varying the Default-Penalty
Cost . . · · · · ·
· · · · · · · ·
· 125
6.3.2
Effects of Increasing Interest Rate for
BNDA Loans · · · · · · · · · · ·
129
6.3.3
Effects of Increasing the Credit Limit for
BNDA Loans · · · · · · · · · · · · · · · · · 141
6.3.4
Effects of Varying the Credit Limit for BNDA
Loans When the Interest Rate is at the
Break-Even Point · · · · · · · · · · ·
· 145
6.3.5
Effects of Increasing the Delinquency-
Penalty Cost · · · · · · · · · · · · · · · · 148
6.3.6
Effects of All-Cash Disbursement for BNDA
Loans
. · · · · · · · · · · · ·
· 153

Ut
TABLE OF CONTENTS (continued)
CHAPTER
SEVEN
SUMMARY. CONCLUSION. SUGGESTIONS
• • 161
7.1
Summary and Conclusion. ••
• • • •
• 161
7.2
Suggestions for Further Research.
• • • 165
REFERENCES •
• 168
APPENDIX •
173
VITA ••
214

x
LIST OF TABLES
Table
1.1
Delinquency Rates in Selected Institutions
4
1.2
BNDA Loan Volume and Percentages by Credit
Activities • • •
• • • • •
7
1.3
Default Rates, Lending Costs and Profitability in
BNDA's Credit Programs (% of Loan Outstanding)
8
2.1
Results from Applying the LSLP Model to BNDA's
Credit Program for Individual Large Farmers
• • • • •
29
3.1
BNDA Large Farmer Credit Program:
Debt
Management
46
3.2
Number of PFVN Loans as % of (l) BNDA Total Number of
Loans Made and (2) Total Excluding the Pr~t de
Soudure
. . . . . . . . . . . . . . . . . .
55
3.3
Indicators of the Agricultural Sector:
The Bouake'
Region . . . . . . . . . . . . . . . . . . . .
57
3.4
Seasonal Specification in Bouake Region
58
3.5
Farm Characteristics in the Major Regions of Ivory
Coast in 1975
• • • • •
59
4.1
Description of Production, Marketing and Consumption
Constraints
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
66
4.2
Description of Financial Constraints •
68
4.3
Description of Production, Marketing and Consumption
Activities • • • • • • • • • • • • •
77
4.4
Description of Financial Activities
78
4.5
Reservation Prices of Cash and Credit
83
5.1
Model Output and Observed Values
91
5.2
Borrowing and Repayment Behavior of the Farmer •
94
5.3
Marginal Value of Extra Resources or Requirements
96
5.4
Conditions for Repaying BNDA Loans, Activity Level
and Variables Leaving the Optimal Solution •
• • 106

xi
LIST OF TABLES (continued)
Table
5.5
Right-Hand Side Ranges for Selected Constraints in the
Basic Model
• • • • • • • • • •
• • • • • • • • 109
6.1
Performance Measures for Evaluating the Model
Variations
.
. • .
• . • . .
. . .
. . . . . . 126
6.2
Effects of Varying the Default-Penalty Cost on the
Farm-Borrower .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
• 127
6.3
Effects of Varying the Default-Penalty Cost on BNDA • • 128
6.4
Effects of Varying the Interest Rate for BNDA Loans
on the Farm-Borrower
• • • • • • • • • • • • • • • • • 133
6.5
Effects of Varying the Interest Rate for BNDA Loans
on the BNDA .
.
.
.
.
.
.
• • .
• .
• .
• .
.
.
.
• .
. 134
6.6
Effects of Varying Default-Penalty Cost When the
Interest Rate is Set Equal to 40% on the Farm-
Borrower
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. 138
6.7
Effects of Varying Default-Penalty Cost When the
Interest Rate is Set Equal to 40% on BNDA •
• • • 139
6.8
Effects of Varying BNDA Credit Limit. Other
Variables at Their Initial Values. on the Farm-
Borrower
.
. . . • . .
. .
• . • . •
. .
. . 142
6.9
Effects of Varying BNDA Credit Limit. Other
Variables at Their Initial Values. on BNDA
• • • 143
6.10
Effects of Varying the BNDA Credit Limit When the
Interest Rate Equals 40%. on the Farm-Borrower
• • • • 146
6.11
Effects of Varying the BNDA Credit Limit When the
Interest Rate Equals 40%. on BNDA • • • • • •
• • • 147
6.12
Effects of Varying BNDA Credit Limit. When Interest
Rate Equals 40% and Default-Penalty Cost is 2.20 CFAF.
on the Farm-Borrower
• • • • • • • • • • • • • • •
149
6.13
Effects of Varying BNDA Credit Limit When Interest
Rate Equals 40% and Default-Penalty Cost is 2.20 CFAF.
on BNDA • • • • .
• • • • • • • • • • • • • • • • • • • 150

xii
LIST OF TABLES (continued)
Table
6.14
Effects of Changing the Delinquency-Penalty Cost on
the Farm-Borrower •
• • • • • • • • • • • . • • • • 151
6.15
Effects of Changing the Delinquency-Penalty Cost on
BNDA





• •
• •








152
6.16
Effects of Interest Rate. Default-Cost and Credit Limit
When They Equal 29%. 2.40 CFAF and +30% of Initial Value
Respectively. on the Farm-Borrower
• • • • • • •
155
6.17
Effects of Interest Rate. Default-Cost and Credit Limit
When They Equal 29%. 2.40 CFAF and +30% of Initial Value
Respectively. on BNDA • • • • • • • • • • • • • • • • • 156
6.18
Effects of BNDA Credit Limit Which is Valued in Reserve
on the Farm-Borrower
. . . . . . . . . . . . . . . . . 157
6.19
Effects of BNDA Credit Limit Which is Valued in Reserve
on BNDA . . . . . . . . . . . . . . . . . . . . . . . . 158

xiii
LIST OF FIGURES
Figure
3.1
Liquidity Value of an Asset
37
3.2
Liquidity Value Curves • • • •
39
3.3
Liquidity Management Vectors for Credit in an LP
Model Incorporating Risk Behavior
• • • •
43
6.1
Equilibrium in the Use of Loan Funds •
131

1
CHAPTER ONE
INTRODUCTION
The economies
of
many developing countries
are still dependent on
agriculture.
In most of these countries where the industrial sector is
embryonic.
the agricultural sector is the major supplier of foreign
exchange earnings.
Two major characteristics of this agriculture are (1)
the predominance of small family size farms.
and (2)
the lack of capital
investment needed to increase production and expand marketable surplus.
In order to promote this vital component of the economy. the govern-
ments of these countries have developed several programs
to assist
the
agricultural sector.
While the form and type of these programs may vary
from country to country. one program. the government supported credit
program (GSCP). is found almost everywhere.
This program is designed to
provide production loans at subsidized interest rates to farmers with
limited income and assets.
The emphasis on agricultural credit programs
stems from the perception that (l) such programs are easy to establish
and administer. (2) capital investment is vital in trying to diversify
and boost production.
and (3) low interest rate loans can be used to
offset
the low commodity
prices and high production cost.
Loans under
GSCPs therefore are made at subsidized interest rates and disbursed in a
way to limit the use of the proceeds for production purposes.
These loans are expected to generate enough income to pay back the
loan and s t i l l leave the farmer better off.
Evidence. however. shows
that most GSCPs have failed to achieve some of the objectives they were
created for.
Common deficiencies are (1) maintaining equity and effi-
ciency in farm production. and (2) assuring the financial viability of

2
the lending institutions. hence limiting the outreach of the program
owing to high cost of lending.
Indeed.
the excess demand for loans
caused by low interest rates has forced lenders to resort to numerous
rationing devices which in most cases raise the loan transaction cost for
small potential borrowers.
This results in the concentration of asset
ownership instead of the intended equity.
However. the most serious
problem affecting the majority of GSCPs is the high delinquency and
default rates.
Table 1.1 reproduced from a recent study by Montie1 [41]
illustrates the seriousness of the problem in some selected countries.
It shows that delinquency and default rates can be as high as 70 percent.
1.1
Effect of Default on Lending Institutions
The viability of the lending institution can be undermined by de-
fault as seen in the following equation [5]
d
L
= f
c
+ a + R ( l + f + a)
(l.1 )
where Lc is the lending cost. f the cost of 10anab1e funds to the lender.
a the lender's cost of administering the loan and d the default rate.
All are expressed in terms of principal loaned.
For an interpretation
purpose. (1.1) can be rewritten as
where b
=
c
a + f is the basic cost. and k = d/ (l-d) the risk factor.
In
the absence of defau1 t , lending cost is reduced to the basic cost.
But
when d 2. 50%. k 2. 1 and. the basic cost is at least doubled increasing
lending cost.
Even when d <50%. the lending cost is still increased due
to default.
In light of this.
the observed high default rates will

3
certainly push the lending cost to unacceptably high levels.
The lending
institution must therefore require high interest rates in order to break-
even.
For example. if f and a are 10% and 4% respectively. with a 3%
default rate lending cost is
.03
L
=
1-.03 (1
c
.10 + .4 +
+ .10 + .04) = .17
In this case interest rate must equal 17% for net lending cost to be
equal to zero.
However. if default rate is 50%. f and a are are un-
changed. lending cost is 128% of principal loaned and the rate of in-
terest would be increased to 128% in order to break-even.
But with the
interventions of local governments to maintain interest rates at a low
level. the lending institutions must either cut back their services or
seek more subsidies from the government in order to at least break-even.
In the current study.
we intend to analyze agricultural loan default in
Ivory Coast.
1.2
Government Supported Credit Program in Ivory Coast: the
"Banque Nationale pour le Developpement Agricole" (BNDA)
Like many other developing countries.
Ivory Coast is trying to
stimulate growth in the agricultural sector through the use of GSCP.
Institutional loans to the agricultural sector are provided by the '~an-
que Nationale pour le Developpement Agricole" (BNDA).
BNDA's main ob-
j ective. is to promote agriculture in general and farming in particular.
through the extension of loans at concessionary interest rates.
It
operates a variety of credit activities that can be grouped into five
categories:
(1) the "Pret de Soudure" (PS). (2) the cooperative credit
program (GVC). (3) the individual credit program. (4) the SODE 1 credit

4
Table 1.1
Delinquency Rates in Selected Institutions
Country
Institution
Arrears to
Annual
Year
Source
Portfolio
Arrears
Rates
Afghanistan
ADBA
37
77
1970-72
World Bank (1975)
Bangladesh
ADB
43
76
1969-73
World Bank (1975)
Bolivia
Agric. Bank
51
1964-71
AID (1973)
BAB
47
1977
Ladman and
Tinnermeier(1977)
Colombia
Caja Agraria
19
1971
World Bank (1975)
INCORA
4
16
1970-71
World Bank (1975)
Costa Rica
BNCR
55
1979
Table 5.21
Chile
-INDAP
16
60
1970
World Bank (1975)
El Salvador
ABC
37
1971
World Bank (1975)
Ethiopia
CADU
50
1971
World Bank (1975)
Ghana
ADB
33
1969-71
AID (1973)
Honduras
BANADESA
53
49
1981
Delgado and
Bueso (1982)
Haiti
BCA
60
1978
Daines (1979)
India
Coops
38
1970-71
Ames (1973)
PLDP
34
7
1971
World Bank (1975)
Iran
ACE I
44
1971-72
Donald (1975)
Kenya
GMR
25
0-22
1971
AID (1973)
AFC
51
36
1971
World Bank (1975)
Morocco
SO CAP
50
1968-72
World Bank (1975)
Malaysia
MADA
19
1975
Samik (1977)
Niger
CNCA
11
29
1971-72
World Bank (1975)
Nigeria
WNACC
52
49
1972
AID (1973)
Peru
ADB
58
1967
Finn (1973)
So. Korea
NACF
7
9
1968-72
Donald (1975)
Sri Lanka
NCS
50
40-50
1969-73
Donald (1975)
Sudan
Coops
26
1968-72
World Bank (1975)
Tanzania
NDCA
28
50
1970-73
World Bank (1975)
Thailand
BAAC
35
5-50
1967-71
Donald (1975)
Tunisia
Credit Unions
50
1970
World Bank (1975)
BNT
66
50
1971
World Bank (1975)
Turkey
TRAB
29
43
1968-70
World Bank (1975)
Source [41]

5
program.
and (5) the credit program for large private farm companies.
A
complete description of each of these programs is provided in [19. 43.
61. 62] and will not be repeated here.
We should. however. note that the
loan categories reflect the income level of the farmers involved.
The PS
provides loans for consumption purposes to small farmers in groups of 6
to 30.
Each group constitutes a borrowing unit and assumes the joint
liability of loans received by individual members.
The cooperative
credit is a group lending program.
In it. loans are given to groups of
medium farmer.s for commercia1iz ation purposes.
The individual credit
program comprises two subprograms:
(1) the "Pret de Faisance Va10ir
Normalise" (PEVN) under which loans for operating expenses are given to
medium farmers.
producers of coffee and cocoa; and (2)
the "Pr~t de
Faisance Va10ir Ordinaire" (PFVO) which provides loans for investment
purposes to large individual farmers.
and covers farming activities such
as banana and pineapple production.
and poultry raising.
The SODEs are
autonomous development agencies placed under the supervision of the
technical ministries (Agriculture and Animal Production ministries).
Their obj ectives are to design and to implement agricu1 tura1 proj ec t s ,
They obtain loans either as financial intermediaries between BNDA and the
farmers they supervise.
or as development agencies that undertake maj or
public agricultural projects.
The last credit program deals with struc-
tured. business-like private farms.
While loans in the PS are extended in cash and without any restric-
tions on how the funds should be used.
loans in other programs are
disbursed in kind and/or cash. for specific purposes.
In order to ser-
vice the maj ority of the farmers. the country is divided into six BNDA
districts.
Each district comprises an agency located in the largest city

6
of the district.
and several bureaus established in the other large
cities.
The volume of the loans made by BNDA has grown from 6.516.0 millions
CFA franc in the fiscal year 1971-1972 to 49.163.0 millions in 1981-1982.
as illustrated in Table 1.2.
This growth. however. has not been equally
shared among its programs.
For example. the percentage of funds allo-
cated to the PS dropped by half.
from 17% to 8%.
Loans to farm com-
panies. either public or private. have remained dominant in the portfolio
of BNDA loans. -The percentage of funds allocated to the large individual
farm borrowers.
the PFVO. has decreased substantially over the last four
years.
More attention.
however.
has been given to both the PFVN and the
GVCs.
Table 1.3. which illustrates the default problem in BNDA credit
programs. shows that repayment in both the PS and the PFVN has been quite
satisfactory.
In these programs default rates varied but around 3% and
7%. respectively.
These rates. which are higher in the GVC. do reach
their peak in the remaining programs. where they can be as high as 80%2.
as shown in the last column of Table 1.3.
The last two rows of Table 1.3
show the lending cost and profitability of corresponding credit programs.
Lending costs are computed using equation (1.1);
because of lack of
information on the administrative cost and the cost of fund in each
credit program. these are assumed all equal to the estimated costs for
BNDA as a whole. 6.53% and 8.25%.
respectively.
Profit in each program.
which is given as the difference between interest rate charged and the
lending cost. is negative throughout. revealing that BNDA is operating at
a loss in all its loan activities.
Hence. in order to at least break
even. BNDA has to reduce default rates and/or raise interest rates to the
level of its lending costs.

Table 1.2
BNDA Loan Volume a and Percentages by Credit Activities
Pret de Soudure
Loan to Coop-
Individual Loans
Otherb
erative GVC
PFVNc
PFVOd
Fiscal
Total
Year
Loan
Volume
as a %
Volume
as a %
Volume
% of
Volume
% of
Volume
as a %
of total
of total
total
total
of total
j
1971-72
6.516.0
1.106.3
17.0
180.0
2.8
29.7
0.5
436.9
6.7
4.763.1
73.1
1972-73
8.445.0
1.545.0
18.3
212.0
2.5
43.5
0.5
558.3
6.6
6.086.2
72.1
1973-74
11.746.0
1.391.0
11.8
101.0
0.9
59.3
0.5
856.6
7.3
9.338.1
79.5
1974-75
18.212.0
1.236.2
6.7
283.0
1.6
143.4
0.8
1.390.5
7.6
15.159.3
83.2
1975-76
30.220.0
1.350.0
4.5
391.0
1.3
278.1
0.9
2.369.5
7.8
25.831.4
85.5
1976-77
23.725.0
1.366.0
5.8
557.7
2.4
541.2
2.3
4.201.6
17. 7
17.127.2
72.2
1977-78
21.744.0
1.419.0
6.5
832.2
3.8
1.068.0
4.9
6.430.9
27.1
12.023.9
55.2
1978-79
29.217.0
2.093.6
7.2
1.760.7
6.0
2.095.5
7.2
5.9
0.002
23.261.3
79.6
1979-80
34.214.0
3.016.0
8.8
3.123.0
9.1
2.312.3
6.8
45.7
0.001
25.717.0
75.2
1980-81
47.943.0
3.523.3
7.3
5.596.4
11.7
2.581.7
5.4
60.2
0.001
36.181.4
75.5
1981-82
49.163.0
4.243.1
8.6
8.014.0
16.3
3.136.2
6.4
63.0
0.001
33.706.7
68.6
Source:
BNDA Annual Reports
aThe unit is million CFAF; 1 CFAF = US $0.0031 in 1982 and US $0.0025 in 1985
bIncludes SODE and Private Farm Companies
......
c Pr e t de Faisance Valoir Normalise
d pr e t de Faisance Valoir Ordinaire

8
Table
1.3
Default Rates. Lending Costs and Profitability in BNDA's
Credit Programs (% of Loan Outstanding)
Pret de
Cooper-
Individual Credit
Farm
Soudure
atives
Compan-
Fiscal
(PS)
(GVC)
PFVN
PFVO
ies*
Year
1974-75
2.00
15.70
8.00
24.30
2.20
1975-76
2.00
13.90
9.00
28.20
4.00
1976-77
3.50
16.50
2.00
32.70
3.80
1977-78
2.50
14.00
3.00
37.00
1978-79
3.20
10.00
7.00
28.00
55.00
1979-80
1.90
17.00
3.00
18.00
62.00
1980-81
5.90
12.00
6.00
19.00
81.00
1981-82
5.10
31.00
21.00
35.00
13 .00
Average Default
Rate
3.30
16.00
7.40
26.50
32.30
Interest Rates (l)
8.50
10.50
10.25
11.00
12.37
Lending Costs
(2)
18.80
37.25
24.06
56.30
69.69
Profit = (1)-(2)
-10.30
-26.75
-13.81
-45.30
-56.63
Source:
BNDA Annual Reports
*Inc1udes SODEs and large private farm companies

9
1.3
Default Management in BNDA
Very often.
default rates are the primary criterion used to measure
the performance of agricultural credit programs in developing countries.
Yet the main objective of these programs has always been to increase
production as well as agricultural income.
If the major criterion that
determines whether the farmer will default or not is his ability to
repay. then default might be due to the failure of the credit programs to
achieve their objective.
As noted by Baker [7]. a direct production
effect of a successful agricultural credit program for small farmers is a
reduction of interest rates given reliable access to timely loans.
This
would increase the range of production alternatives that yield positive
returns to the small farmer.
As well. an agricu1 tura1 credit program
which provides lower cost credit to substitute for cash and other cash
substitutes in reserves management would help boost production although
indirectly.
and possibly enhance repayment of loans.
Therefore.
a pro-
gram which meets all the financial requirements of the small farmer is to
be sought.
The farmer's inability to repay. however. is not the only cause of
delinquency and default.
The farmer's unwillingness to repay.
due either
to a poor quality of the loan services or to the misunderstanding of the
terms of the loan contract is well recognized as a cause of loan default
in developing countries [48. 50. 51].
Improved loan services coupled
with sanctions and rewards.
therefore.
should be considered as means to
control default.
To date. BNDA has tried to control loan delinquency and default
through penal ty measures.
These include additional charges.
refusal of

10
subsequent loans.
and legal actions.
For instance.
an additional charge
of 0.70% per month is imposed on past due loans.
This charge which is
effective beginning the 26th day after the unmet maturity date.
applies
to all BNDA borrowers.
In the specific case of the Pret de Soudure. a
past due loan results in the denial of subsequent loans to the entire
group. and the program is discontinued in bureaus with a delinquency rate
exceeding 5%.
Since September 1977. legal measures have become BNDA's
maj or policy to prevent loan default among its clients.
Legal actions.
in general. are instituted against clients who have the means to repay
their loans but fail to do so.
In these cases. the matter is referred to
a lawyer for collection.
The client is often given a 60 days grace
period.
but may benefit from a longer period due to delays in the imple-
mentation of the decision.
When notes are filed for legal actions the
client is charged 20% as attorney's fee. in addition to the basic loan
interest rate and the monthly 0.70% charge on past due loans.
The penalties seem to have reduced default among small farmers.
However.
default has increased among large farmers.
Is this due to the
fact that the penalties are more severe for small farmers.
or is it due
to the fact that large farmers can find ways around the penalties? Thus.
what set(s) of sanctions or rewards would effectively change the default
behavior of larger farmers?
1.4
Research Objectives
The purpose of the current study is to identify lending policy
reforms which will reduce delinquency and default rates in BNDA's credit
program for individual medium-sized farmers. the PFVN. 3 while maintaining
growth in the agricultural sector.

11
Terms that are important in a loan contract are:
(l) the interest
rate.
(2) the amount of the loan. and (3) the collateral or the equity
lenders require.
These terms could be changed by lenders.
whenever
necessary. in order to influence the behavior of their clients.
For
example. the BNDA has tried to control default by increasing the interest
rate on past due loans and by denying subsequent loans to defaulters.
Here. we intend to study how varying
(1)
the delinquency cost
(2)
the default cost
(3)
the interest rate. and
(4)
the loan limit.
would affect loan delinquency and default among the medium-sized Ivorian
farm-borrowers.
We hope that these changes would lead to a set of
policies that would help to reduce delinquency and default. and thus make
the borrower value BNDA credit reserve when its loans are disbursed all
in cas~
It is expected that the resulting credit program will provide
the desired characteristics for a government supported credit program in
Ivory Coast.
In order to achieve these objectives.
we will use a farm-
level decision-making model that incorporates delinquency and default
behavior.
1.5
Organization of the Thesis
Review of work related to the problem of loan default in developing
countries in general is presented in Chapter Two.
The method of the
analysis is described in Chapter Three.
Chapter Four describes the
mathematical programming model used in the study.
The results of the
model and its validation procedure are presented in Chapter Five.
Chap-

12
ter Six describes the policy variables selected for the analysis of the
policy reforms and presents the results of the effects of the changes in
these variables on the farm-borrower and the BNDA.
The summary. the
conclusion of the thesis. and the suggestions for future research are
presented in Chapter Seve~

13
NOTES
1. SODEs (Societe de Deve10ppement) are organized by product and the
principal ones are SATMACI (coffee and cocoa).
SODEPALM (palm oil and
coconut).
SODEFEL (fruit and vegetables). SODESUCRE (sugar). MOTORAGRI
(agricultural motorization).
and SODEPRA (livestock).
2. This is not the case in the last fiscal year. 1981-1982.
In that
year. default rate increases in all the programs except for farm
companies.
The overall increase in default rate is due to the cli-
matic conditions which were not favorab1e for agricultural production
nationwide.
Default rate for the farm companies decreased because of
the government's development program to phase out a number of SODEs.
The debts of these SODEs have been repaid. hence reducing the default
rate of the credit program.
3. In the current study. we emphasize on the PFVN program despite its low
rate of default shown in Table 1.3 for two reasons:
(1) it is an
important credit program for the BNDA and (2) we dispose of data only
for farmers who obtain loans from the BNDA under this program.
How-
ever. we expect the model to be developed in this study to be suitable
for the analysis of default in the other credit programs.

-
14
CHAPTER TWO
LITERATURE REVIEW'
The problem of loan default in developing countries has been the
subject of several studies.
In a country where default is a major finan-
cia1 problem. policy makers are interested in studies that not only
explain the problem but also can be used to design policies to alleviate
it.
Stewart [56] notes that since the investigator is trying to give
/
advices which will help policy makers to tackle the problem. it is im-
portant that he know s (1) how individuals reac t to various policy mea-
sures. (2) why they react as they do. and (3) how they should react in
the future.
In general. an adequate representation of the system from
which the problem has developed would help to answer these questions.
Because the default problem is inherent in the problem of farm-firm and
household growth. its analysis requires a model that (i) incorporates the
necessary production. consumption. marketing. and financing relation-
ships.
(ii) is adaptable to variation in input.
and (iii) generates
output measurable in terms relevant to questions on growth and cash
flows.
as well as default.
Like the phenomenon of growth. default is a
time dependent phenomenon. implying that analyses which are static in
nature are inadequate.
In what follows we are going to review some of
the existing empirical studies.
We restrict ourselves to these studies
because their corresponding models have a "built-in" procedure for either
testing hypotheses or generating policy decisions.
In this review we
will be mainly interested in finding out whether or not these studies
fulfi11 the requirements implied in the introductory analysis that pre-

15
cedes.
The review will be divided into two types of studies:
(1) be-
haviora1 studies and (2) policy oriented studies.
2.1
Behaviora1 Studies
Behavioral studies attempt to explain the phenomenon of loan default
by establishing possible relationships between default rates and selected
socio-economic characteristics of the farmers and the lending institu-
tions.
2.1.1
Ames' Study [2]
In this study Ames examines the relationship between the repayment
of crop production loan and farm borrowers as well as lending institu-
tions' characteristics.
He uses cross sectional data of 136 farm bor-
rowers from 35 agricultural credit cooperative societies in Mysore State.
India.
Ames first compares averages of farmer resources.
cropping pat-
terns.
borrowing practices and problems by farm size for defaulters and
non-defaul terse
His findings are:
(a) defaulters have fewer assets in
land.
livestock.
equipment and grain stocks than do non-defaulters.
(b)
defaulters have larger average currently financed capital investment than
non-defaulters.
(c) defaulters have on the average a lower net output per
acre for all crops. less farm income and own fewer irrigated acres than
non-defaulters. and (d) the combination of financial obligations to repay
other loans and crop production loans is greater than defaulters' limited
earnings could support.
Ames uses least squares mu1 tip1e regression analysis to identify
farm borrowers' socio-economic characteristics which explain either
farmers' total amount of outstanding loans or the amount of crop produc-

16
tion loan overdues.
In his regression model. Ames uses nine explanatory
variables. and subdivides the sample into two groups according to the
size of the farms:
small and large.
He then runs three regression
analyses. one for the small farms. one for the large farms and one for
all the farmers put together.
Of the nine variables used in the model.
those representing the capital investments. and the percentage of ex-
penses covered by the loans are the only significant factors in ex-
plain~ng loan overdues among small farmers.
In this group loan overdues
increase with both the capital investments and the percentage of expenses
covered by loans.
For the large farmers.
the coefficients of the vari-
abIes representing the capital investments. the farm assets and the
annual consumption expenditure are significant.
The loan overdues are
positively related to both capital investments and farm assets. but
negatively related to annual consumption expenditure.
When the observa-
tions on all the farmers are taken into account in the regression analy-
sis. the variables representing the capital investments. the farm assets.
the consumption expenditure.
and
the percentage of expenses covered by
the loans are significant with the same signs as above.
Four observations can be drawn from Ames' study:
(i) the model used
is not rich enough to give an adequate representation of a system as
complex as the farming system in developing countries: (ii) only the
coefficient of the variable representing the capital investments appears
to have the correct sign. suggesting that most of the hypotheses made
should be reconsidered:
(iii) the model was designed to explain the
phenomenon of loan default and as such does not help in answering ques-
tions such as how farmers react to policy measures. why they react as

17
they do.
and how they should react in the future.
and (iv) Ames' study is
based on a single period analysis.
Because default is dynamic in nature.
an appropriate study should evaluate expected future impacts of policy
taken today.
Such a study will have to be dynamic or time dependent.
2.1.2
Pradhan and Sharma's Study [49]
The main obj ective of this study is to identify factors that dis-
criminate borrowers in crop loan repayment of a commercial bank 1 in
India.
The authors use a discriminant analysis to identify variables
which classify the farmers into non-defaulters and defaulters.
and de-
faulters into non-wilful defaulters and wilful defaulters.
Wilful de-
faulters are those defaulters who have enough money to repay at least 50%
of their loans but choose not to repay.
Non-wilful defaulters are those
defaulters whose repayment capacity falls below 50%.
Of the 80 farmers
who participated in the loan program. 55 defaulted. among these 32 wil-
fully.
Because the analytical tool employed in the study requires ap-
proximately the same number of individuals in each of the two groups
being analyzed. two random samples of 25 borrowers each were selected to
represent the four groups being considered:
the non-defaulters. the
defaulters. the non-wilful defaul t e r s , and the wilful defaul terse
The
discriminant function comprised 15 variables.
The results of the study indicate that the variables representing
the siz~ of operated farm. the percentage of loan under crop production.
the percentage of cash expenditure in total expenditure. and the loan
efficiency are significant in discriminating between non-defaulters and
defaulters.
The estimated function shows that other things being equal.

18
an increase in either loan utilization for crop production or loan effi-
ciency2 raises the discriminant score.
hence placing the borrowers into
the group of non-defaulters.
But an increase in either operated land or
percentage cash expenditure reduces the discriminant score:
in this
case. the borrowers fall in the group of def au1 terse
The mean discri-
minant scores for the non-defaulters and defaulters are 4.6954 and -
0.5802. respectively.
The mean discriminant score for the two groups is
2.0576.
This implies that if the discriminant score of a respondent that
is determined on the basis of the significant variables of the current
analysis and information on the respondent is more than 2.0576. this
respondent can be predicted to be a non-defaulter. otherwise he is likely
to be a defaulter.
Of the 15 socio-economic and behaviora1 factors used in the analy-
sis.
only 3. those representing the educational level.
the percentage of
total expenditure in total income.
and the percentage of earning adults
in the total number of adults are the significant discriminators between
the non-wilful defaulters and the wilful defaulters.
The discriminant
score. in this case. decreases with the level of education. but increases
with either the percentage of total expenditure in total income or the
percentage of earning adults in the total number of adults.
The mean
discriminant score is 22.2185 and 11.5842 for the non-wilful defaulters
and the wilful defaulters. respectively.
The discriminating score for
the two groups is 16.90135.
This score indicates that a respondent whose
computed score is greater than 16.90135 should be considered as a non-
wilful defaulter. otherwise he is a wilful defaulter.
Based on the results of their study.
the authors make the following
recommendations:
(1) every lending institution should see that efficient

19
loaning is ensured in each crop loan transaction: (2) the agricu1 tura1
field officers of the lending institutions should try their best to see
that diversion of loan to purposes othe~ than the one for which the loans
have been sanctioned is minimum:
(3) the lending institutions should be
more inclined toward the small farmers because they have a tendency to be
good pay masters;
(4) the lending institutions should use the prediction
criteria on the basis of the significant factors. in order to a priori
assess their lending risk.
and (5) the lending institutions should follow
a discriminant credit policy when dealing with the non-wilful and wilful
defaul ters.
Pradhan and Sharma's approach differs from Ames' approach in that
the former uses a discriminant analysis whereas the latter is based on a
least squares regression analysis.
The discriminant analysis allows (1)
to test for mean group differences and to describe the overlaps among the
groups. and (2) to construct a classification scheme based on the es-
timated discriminant function in order to assign previously unclassified
observations to appropriate groups.
But as in the case of a regression
analysis.
the model only explains the phenomenon of loan default.
Using
the constructed mean score. the model can also be used to predict whether
an individual will default or not. and if he defaults. whether he does so
wilfully or not.
Despite these differences.
the same criticisms 1eve1ed
against Ames' study can be 1eve1ed against Pradhan and Sharma's study as
well.
·For example the 5 recommendations given above do not provide
enough information in designing policies that will reduce default.
and
evaluating the impact of these policies.

20
2.1.3
Octavio's Study [44]
This study analyzes loan repayment and technical assistance among
Masagana-gg 3 farmers in Bulacan,
the Philippines.
In a multiple regres-
sion analy sis on cross sectional data of 6 rural banks, 21 technicians
and 120 farm borrowers, Octavio shows that:
(a) loan repayment is posi-
tively related to the age of farmers, the farm size, the quantity of
palay sold, the size of the loan, and the provision of irrigated water;
(b) repayment is negatively related to the farming experience of farmers,
the size of the household,
the level of education,
and the distance
between the farm and the rural bank; (c) the household size, the quantity
of palay sold and the availability of irrigation water are the most
significant factors determining loan repayment, and (d) the age of the
farmer is a relatively significant indicator of loan repayment.
Octavio's findings contradict those of Pradhan and Sharma as far as
the effect of farm size on repayment is concerned.
Octavio's analysis
indicates that repayment increases with farm size while Pradhan and
Sharma show that the farmer will be a defaul ter if he operates a large
farm.
Logically,
one would expect a large-sized farmer to dispose of a
higher repayment capability hence repay more often his debts.
But a
farmer may also default because he finds ways around the penalties.
This
is likely to be the case of large farmers.
Hence, taking into account
factors other than the ability to repay, one may accept Pradhan and
Sharma's result.
But, although the two studies indicate that the size of
the farm is a variable which significantly accounts for the farmer's
default behavior, one cannot use it to derive general policy measures for
reducing default owing to the difference in the conclusions reached.

21
Such policies would only be specific to groups of individuals studied for
specific regions in countries.
Overall.
Octavio's model seems to have performed better than Ames'
model.
Of the 10 explanatory variables considered. 5 are significant and
have the expected signs.
But as in the case of Ames' study. Octavio's
study is based on a single-period model and does not provide adequate
answers to the policy decision questions mentioned earlier.
However.
Octavio's results are revealing as far as the relationship between repay-
ment and important farm borrowers' socio-economic characteristics is
concerned.
These results could be very useful in answering questions
pertaining to what ought to be done about the default problem.
2.1.4
Best's Study [14]
In his study. Best looks at the repayment performance among farmers
in three provinces of the Philippines. during the wet and dry seasons.
This study is based on information gathered on two groups of farmers:
one comprising 181 farmers for the 1974 wet season and the other 121
farmers for the 1975 dry season.
A multiple regression analysis is used
to measure the effects of 9 socio-economic variables. and 3 season and
province dummies on the repayment rates.
The dependent variable used in
the model is a discounted repayment rate.
measured so as to reflect the
fact that many farmers make payment on loans past the due date.
When a
f a r me r makes a payment on a past due account. the amount paid is dis-
counted by the opportunity cost of his keeping the money past the due
date.
Three measurements of income--net farm income per hectare.
yield
per hectare. and total income--are identified and used in separate re-
gression runs in order to determine which measure would explain the

22
greatest amount of variation in the repayment rate.
The major findings
are:
(a)
repayment rates are positively related to net farm income per
hectare. yield.
technician's visits. and family size;
(b) repayment rates
are negatively related to the variable representing the quality of the
management of the bank. loan amount per hectare.
farm size. capital
assets. educational level and the dummy variables; (c) net farm income
per hectare. yield. bank management. loan amount per hectare.
farm size.
and technician's visits are the most significant factors determining
repayment rates;
(d) the effect of capital assets on repayment rates is
negligible.
and (e) educational level and family size are insignificant
determinant of repayment rates.
Despite the fact that Best's model takes into account the possibili-
ty that farmers might pay the loan after the due date his results are not
superior to Octavio's.
For example.
the signs of the coefficients of the
variables representing capital assets. and family size do not conform
with logical expectations.
This author notes that capital asset. defined
as machines or animals the farmer owns and uses on his farm. measures the
wealth position of the farmer and hence would be expected to positively
affect repayment.
But his results reveal the opposite effect.
Also. his
results indicate that farm size. which he defines as another indicator of
the farmer's wealth position.
is negatively related to repayment.
which
is opposite to Octavio's results for the same country.
These conclusions
furthe~ support our argument that the results of these studies would lead
to policy decisions
which only
are applicable in the cases being
analyzed.
The same critical remarks made earlier therefore are appli-
cable to this study as well.

23
2.1.5
Montiel's Study [41]
Montiel.
more recently [1983].
studied the problem of loan default
in developing countries in general. and in Costa Rica in particula~
In
his study.
he stresses on the fact that default is not a managerial
problem. instead it is a structural problem.
He argues that agricultural
credi t programs in developing countries have a "buil t-in" default rate
due to the high risk of the activities financed. the high leverage al-
lowed. and the existence of perverse incentives for both farmers and
banks.
Using-a contingent claims analysis.
an extension of option
pricing models. Montiel specifically underlines the financial incentives
for borrowers as well as the implicit subsidies in agricultural credit
programs created by default on the loans.
After reviewing several case studies of the default problem. Montiel
points out the methodological problems common to studies which are based
on statistical analysis:
for example. how to measure default.
to select
an appropriate set of variables to be used in explaining default. to
account for the interaction of production and consumption.
and finally
how to account for the fungibility of credit.
Montiel believes that
these problems might be used to explain the conflicting conclusions
reached by several researchers in trying to explain the problem of loan
default in developing countries.
But despite these remarks. he uses
multiple regression analysis.
a statistical method of analysis.
to iden-
tify factors contributing to the default problem in Banco National de
Costa Rica (BNCR).
The data used in the study are cross sectional data
of 895 loan cases from the Rural Credit Department and the Branches
section of the Commercial Department at the BNC~
The regression model

24
includes 7 explanatory variables of which 3 are dummy variables repre-
senting the unit granting the loan. the sector being financed. and the
type of payment arranged.
Montiel performs two regression analyses using
two types of measurement of default:
(1) the number of days a farmer is
delinquent in the payment of interest or principal and (2) the percentage
of principal in arrears.
The major findings of this study are:
(a) loan default is positive-
ly related to the amount of loan outstanding. and the number of years
since the loan was granted;
(b) loan default is negatively related to the
interest rate. and the term of the loan;
(c) loan default increases under
bank-branch loan granting.
when the loan is classified as new.
and when
the repayment arrangement is such that all principal is due at the end of
the term.
and
(d) the most significant determinants of loan default are
the amount of loan outstanding.
the interest rate.
the term of the loan.
the number of years since the loan was granted. and the unit granting the
loan.
Montiel's analysis differs from the previous ones in that it
emphasizes on the financial environment of the farmers.
Only variables
that relate either to the lending institution or to the borrowing be-
havior of the farmers are used as explanatory variables.
But several elements of criticism are in point here.
Using Mon-
tiel's results.
one would recommend that the bank (1) makes longer term
loans. (2) requires higher interest rates.
(3) reduces the amount of loan
out s t and Lng,
(4) avoids the aging of its loans.
and (5) extends the loan
itself.
The first recommendation could be unrealistic when the bank is
the major supplier of agricultural loans and the majority of its clients
are small farmers.
We would expect the second recommendation to lead to
a decrease in the amount borrowed.
in which case the bank's performance

25
will improve only if it can make loan to new clients.
The third recom-
mendation seems to contradict the first one.
Indeed.
a bank cannot make
long term loans without having to deal with loan outstanding.
Loan
outstanding might be low if the limits on the loans are kept very low.
But such a policy might in turn lead to the farmer's inability to gen-
erate enough income to repay his loan. as seems to be implied by Octa-
vio's resu1 t s,
Following the fifth recommendation could lead to a too
high concentration of the services the bank can provide.
and hence fur-
ther limit the outreach of the program.
The perception is that the
characteristics differentiated by the dummy variable for the unit
granting the loan imply little policy decisions.
Overall. like the
previous studies.
this study only explains the phenomenon of default and
does not help answering the questions relevant to policy decisions men-
tioned earlier.
All the studies seen so far tend to explain the phenomenon of loan
default rather than suggesting steps to be taken to alleviate the prob-
lem.
This was to be expected since they all use a statistical approach
in modelling the phenomeno~ Unfortunately. statistical model fitting
often results in models which are not rich enough to give an adequate
representation of the system being studied.
The models above involve
each at most 15 variables which are expected to be the major determinants
of loan defau1 t , In all the five cases. however. less than 50% of these
variables have the expected sign and could significantly explain the
phenomenon.
For example. the five studies indicate that the farmer's
weal th position is a significant determinant of loan default.
but they
cannot agree on the direction of the relationship maybe because of their
failure to account for additional information which is relevant to the

26
problem.
In order to answer important questions such as how do indi-
viduals react to various policy measures. why they react as they do.
and
how they should react in the future.
the system must be modelled with as
much information as possible.
Since the answers to these questions are
critical in designing policies used to combat the problem of high loan
default.
other approaches must be considered.
For example. po1icy-
oriented studies. which we consider next.
2.2
Policy-Oriented Studies
Those who are in charge
of
the credit programs are not just
interested in knowing why the programs are not running properly. they are
also interested in steps to be taken to alleviate loan delinquency and
default.
These steps could be prescribed by policy-oriented studies.
There exists.
although in limited number.
policy-oriented studies that
have attempted to look at the default problem.
We will review some of
these studies.
2.2.1
Liquidity Management and Small Farm Borrowers' Attitude
Here we consider studies done in India [15]. Cameroon [35]. Ivory
Coast
[62] and Philippines [21. 46].
Although the major concern here is
liquidity response to risk. the authors did not ignore the default prob-
lem.
They all have one common obj ective:
studying these programs in
order tb come up with changes necessary to improve the effectiveness of
the GSCPs.
In order to achieve these objectives. the authors build a
linear programming model incorporating liquidity management constraints.
or Liquidity Specified Linear Programming (LSLP) model.
This model is
based on the Theory of Liquidity Management and is validated using infor-

27
mation gathered on the production.
marketing.
consumption and financing
activities of the farm-firm and househo1~
The tenet of the LSLP model is that in addition to financing his
production needs.
the farmer must meet his household consumption needs.
and cope with unforeseen contingencies in both the firm and the house-
hold.
According to the Theory of Liquidity Management. the farmer re-
sponds to such a situation by maintaining liquidity reserves which are
mainly in the form of cash. and cash substitutes. the latter including
credit under specified conditions.
In the model. this is expressed by
the liquidity management constraints whose right hand sides represent
minimum levels of liquidity.
These levels are subject to increase from
risky activities and decrease from cash and credit reserved.
whose li-
quidity values and the transfer of cash from the cash account rows con-
stitute the obj ective function.
When the debt contracts allow unpaid
debt to exist at the end of the model period.
the objective function
includes carryovers of unpaid debt as well.
The model so obtained generates an optimum of area planted to the
specified crops
and of livestock enterprises.
It also generates.
jointly. an optimum among alternatives for meeting household require-
ments. produces an optimal mix of marketing strategies in accordance with
the alternatives specified.
and gives an optimal level of borrowing at
each lending source.
To validate the model.
its optimum output is com-
pared with observed counterparts of the mode1ed unit.
This process also
includes a search for variations in the values of the liquidity coeffi-
cients in both the liquidity constraints and the objective function until
a suitable "fit" is found.

28
The four studies use the LSLP model in its validated form to simu-
late the behavior of the borrower in each of the following lending poli-
cies resulting from the survey:
(1) changing loan disbursement from kind
to cash. (2) increases in the interest rate. and (3)
increases in the
size of the maximum loan.
Indeed. in the survey. respondents had ob-
jected to higher interest rates.
but they were more than willing to pay
higher interest rates if the loan limit could be raised and if the loans
were disbursed in cash.
Table 2.1 illustrates the results of the LSLP
model applied to BNDA's individual credit program for large farmers [62].
It shows that when loans are disbursed in kind.
and if the interest rate
is increased to 40%. net cash flow and total cash decrease by 9% and 2%.
respectively.
This might lead to a decrease in the performance of the
BND~
When the interest rate equals 40% and loans are disbursed in cash.
both net cash flow and total cash increase but without reaching their
initial values shown in the first column.
If in addition. the credit
limit is increased by 50% over its initial value,
more liquidity is gen-
erated owing to the higher amount of BNDA credit reserve. and total cash
further increases.
Hence.
it might be that BNDA would improve its per-
formance by requiring a higher interest rate if loans are disbursed in
cash and the loan limit increases.
The results of these studies indicate that when a farmer. especially
a small farmer.
perceives the government supported programs as a perma-
nent and reliable source of liquidity. he tends to substitute the credit
for reserved cash and commits more cash to production.
Bhargava and
Baker [16] note that these results do not in themselves produce prescrip-
tions for improvement in the agricultural credit programs,
but they do
suggest payoffs attainable if the programs can be improve~
Although the

29
Table 2.1
Results from Applying the LSLP Model to BNDA's Credit Program
for Individual Large Farmers
Performance Measure
BNDA INTEREST RATE
(CFAF)
11%. and BNDA Loan
40%. and BNDA Loan in
in Kind
Kind
Cash. with
Initial
+50% Increase
Credit
in Credit
Limit
Limit
Credit Reserve at
Moneylender
16.129
24.739
21.892
22.070
Commercial Bank
342.601
346.575
323 .893
324.070
BNDA
o
o
738.653
1.073.274
Cash Reserve
41.438.164
42.374.154
40.520.480
42.072.240
Net Cash Flow
24.151.411
21.964.145
23 .397 .885
23.286.176
Total Cash
65.589.575
64.338.299
63.918.042
65.358.826
Source (62J
authors believe that the combination of the various characteristics of
the government supported credit programs implied by the policy reforms
might bring the borrower to behave more positively toward the lending
institution and thus make an effort to repay his loan whenever he can. it
is suggested that sanctions should be considered to hold default in
check.
Two important features of the LSLP model are that:
(1) it provides
a means to reflect a financial response to the uncertainty inherent in
small scale farming.
and (2) it accounts for the links between the farm
and the household mostly in their share of labor and working capital.
Here the farmer's risk response is reflected in the liquidity management
vectors.
Another feature of the LSLP model is its ability to handle a
large number of interrelated variables such as production and con-

30
sumption. consumption and investment. investment and resource availabili-
ty. all with social and cultural constraints.
One can then conclude that
the LSLP model is well suited to represent the system being studied.
The
model.
however.
does have some weaknesses in dealing with farmers' de-
fault behavior.
It does not account for the default behavior of the
farmer.
Instead it specifies full repayment at maturity of all debt
created.
This is expressed through debt-balance rows specified as
equality constraints with zero right hand sides.
Such a treatment of
default behavior might be acceptable since the model was primarily con-
ceived to analyze the perf ormance of the lending insti tution in which
case assuming full-repayment is the right course.
But when default is so
widespread as a financial behavior. its omission from the model neglects
a highly significant aspect of the farmers' financial behavior.
I f one
is interested in using the liquidity specified linear programming model
to analyze the default problem. it should be modified to include default
activities.
2.2.2
The Simulation Study of Rosegrant and Herdt [54]
The main objective of the study is to simulate the effects of credit
policy and fertilizer subsidies on input use. production and income of
farmers in the Masagana-99 rice production program in the Philippines.
The analysis is based on a mul tiseason decision making model.
It con-
siders a stochastic production technology.
risk neutral and risk averse
decision rules. a short term saving/consumption behavior. and a dual
financial market.
The major findings of this study are:
(a) the combi-
nation of credit policy and fertilizer subsidies cause at most 21 to 30%
increase in yields. (b) benefits are greater on irrigated farms than they

31
are on rainfed farms and Cc) a substantial defaul t rate in the lending
institution reduces credit program benefits.
Rosegrant and Herdt also
predict an average default rate of 10% per year.
This simulation study represents an important improvement over the
studies reviewed earlier in that (1) the model is capable of predicting
an average annual default rate on institutional market loans. and (2) the
model contains enough information to give a good representation of the
farming system. and therefore can help in answering questions such as how
do farmers react to necessary policy measures. why do they react as they
do.
and how they might react in the future.
This model appears suitable
for analyz ing the phenomenon of loan defaul t in developing countries.
However. the way the model treats default should be reconsidered. that is
default occurs when subsistence consumption expenditure exceeds generated
income.
The model predicts an average annual defaul t rate of 10% sup-
porting the predicted increase in farmers' yields. output and incomes.
But according to Pollar and Grewal [48] default rates for the Masagana-99
program were modest at the beginning of the program and increased sub-
stantially in sharp contrast to the 10% rate predicted.
This is also
confirmed in a study by Castillo on Masagana-99 [21].
Pollar and Grewal
attribute such a discrepancy to the fact that Rosegrant and Herdt do not
consider the farmer's willingness to repay in their model.
In so doing
the model overstates farmers' repayment behavior.
The discrepancy be-
tween the model's prediction and the observed value of default may as
well be due to the model's failure to explicitly account for uses of farm
income other than subsistence consumption and purchase of inputs for rice
productio~
For instance.
farm income may be used in reserve to satisfy

32
liquidity requirements.
as shown in the previous studies.
The omission
of these reservations might overestimate the farmers' repayment ability.
Rosegrant and Herdt's model accounts for farmers' risk behavior
which the authors specify following the safety-first decision rule.
In
their model.
the farm unit maximizes the expected value of net returns
that can be obtained with a fixed confidence level.
Their results indi-
cate that the risk neutral specification is more consistent with actual
choices in production than the risk averse specification.
Yet evidence
is that farmers in developing countries. in most cases. behave in ac-
cordance with risk averting behavior. hence should be modeled accord-
ingly.
Risk averse farmers may respond to risk by means of several
alternative responses.
These include production. marketing and financial
responses.
Production responses often are achieved through enterprise
diversification and factor flexibility.
In marketing. risk management is
performed through the use of various forms of forward pricing and
con-
tracting.
Financial responses to risk include insurance and liquidity
management.
However.
the high cost of acquiring and implementing feas-
ible options in production and marketing leaves financial responses.
especially liquidity management.
the most viable means for coping with
risk in developing countries.
As shown in the previous group of policy-
oriented studies. liquidity management is achieved through holdings of
cash. credit and major financial reserves. all expressed in liquidity
requirement constraints and added to a conventional LP model.
Hence.
a
model that not only expresses the equilibrium condition of the farm in
terms of its production. marketing and consumption. but also includes
these financial decisions might be expected to yield estimate of default
behavior more comparable to the actual behavior of the farm-firm and

33
household.
Such an alternative exists in the LSLP model.
We therefore
believe that a default augmented LSLP model is suitable for our study.

34
NOTES
1. Pradham and Sharma study the phenomenon of loan default in a commer-
cial bank rather than a government supported credit program because
commercial banks offer better perspective for improving on the supply
of credit to the agricultural sector in Mysore State.
India:
(l) they
dispose of a large technical personnel.
and (2) they have an adequate
financial base per office (Rs. 82 Crore against Rs. 11000 per Primary
Agricultural Credit Society in the case of the cooperative lending
system) •
2. In their analysis.
Pradhan and Sharma estimated the loan efficiency
using a model which involved a maximum score rate of 9 elements.
These included the following questions:
1. Was the loan sufficient to
meet your r~quirements?i 2. Did you get the loan timelY?i
3. Did the
agricultural field officer visit your farm before loaning?;
4. Did he
visit at sowing time?i
5. Did he visit between sowing and har-
vesting?i
6. Did he visit at harvest time?i
7. Did he give technical
advice?;
8. Did he show concern from crop success and loan repay-
ment?i
9. Was the agricultural officer. in your opinion. efficient?
The score rate for "yes" and "no" answers were 1 and O. respectively
([49].
p.
26).
3. Masagana 99. which means "bountiful harvest" of 99 cavans per hectare.
is a program designed to provide noncollateral.
institutional credit
and extension services on a large-scale to the vast majority of sma1l-
scale rice farmers in the Philippines [14].

35
CHAPTER THREE
METHOD OF ANALYSIS
As stated in the preceding chapter. a default-augmented liquidity
specified linear programming model is used in the current study.
The
model will be validated with information gathered in a survey of farmers
from BNDA's center district. Bouake.
This chapter begins with a descrip-
tion of the Liquidity Specified Linear Programming model.
The default
management procedure is then discussed.
followed by the description of
the dat~
The description of the Bouake Region is presented in the last
section of the chapter.
3.1
Description of the Liguidity Specified Linear
Programming Model
3.1.1
Theoretical Framework
The Liquidity Specified Linear Programming (LSLP) model is based on
the Liquidity Management Theory (LMT). Following Baker [3.
5]
liquidity
may be defined as the capacity to produce cash upon demand and is
measured by the relative cost of doing so.
This capacity to produce cash
upon demand lessens the probability of failing to meet an obligation.
a
fundamental concept of risk.
To better understand the idea of liquidity.
the firlll. as a collection of assets. is identified to have a value that
exceeds the sum of values of its assets considered separately.
That is.
any asset by being in the firm adds a value to it.
Yet in a rationally
organized firm. no asset would be found with an expected sale value
greater than its contribution to the firm's value.
If it were the case.

36
that asset would have been detached and sold.
Consequently. any asset
with a value less than its contribution to the value of the firm remains
in the firm because of the greater value there.
The liquidity of an
asset is thus given by the reduction to the value of the firm if it is
detached and sold.
The less (more) the diminution in the value of the
firm from the sale of an asset. the higher (lower) the liquidity value of
the asset.
Figure 3.1 illustrates the liquidity value of an asset in the
firm.
The horizontal axis measures the expected sale value of the asset.
The asset's contribution to the firm-value is measured on the vertical
axis.
An asset with sales value equal to the diminution in value of the
firm is considered "perfectly liquid-"
The value of such an asset would
lie on OL which portrays a limiting relation between an asset's value and
its contribution to the firm's value.
Indeed. sales of assets in general
are expected to diminish the value of the firm by more than the expected
proceeds of the sales.
That is.
their values would lie on rays above OL
(evg ,
OA).
In general. all balance-sheet assets have liquidity components. with
cash being the most liquid in that its transaction cost is almost zero.
Cash may be used to sustain investment opportunities which would generate
income thus contributing to the value of the firm.
Cash may also be
reserved to add to the firm's capacity to meet unforeseen contingencies.
Cash. however. fails to be a perfectly liquid asset.
This explains its
being in the firm.
As the firm's cash supply increases. the liquidity
value of further additions diminishes.
Moreover.
the greater the number
of non-cash sources of liquidity,
the lower the liquidity value of cas~

37
CFAF
Ql
;j
..-l
ell
>
~'r-l~
0
+J
L
+J
Ql
1lI
1lI
ell
~
0
I::
0
'r-l
+J
;j
,.Q
'r-l
1-1
+J
I::
0
Co)
o
expected sale value of asset
CFAF
Figure 3.1
Liquidity Value of an Asset
Source:
Baker and Bhargava
[8]

38
But liquidity is not restricted to balance sheet assets.
Credit.
defined as the capacity to borrow. can substitute for cash as a source of
liquidity.
since borrowing can be used to acquire cash without actually
disposing of the firm's productive assets.
Credit. therefore. consti-
tutes an important source of liquidity if it is maintained in reserve.
But credit is not a costless form in which to hold liquidity.
Indeed.
borrowing generates debt obligations and thus increases financial risk.
Moreover.
to reserve credit is to forego gains from financial leverage
except at the equilibrium point where marginal gains expected from lev-
erage equal marginal gains expected from liquidity.
However.
credit can
substitute for cash as a source of liquidity and thus reduce the cost of
risk management.
The reliability and the accessibility.
and the flexi-
bility in the use of the loan-proceeds tend to increase the liquidity
value of unused credit.
Liquidity values of cash or credit are expressed by liquidity prices
of unused cash or unused credit.
Indeed. committing cash to use or using
credit through borrowing results in liquidity losses. this is expected to
increase the marginal value product of liquidity to individuals.
Because
liquidity could prove to be a limiting constraint. one might expect a
certain val ue to be placed on it in the form of unused cash or unused
credit.
Moreover. one would expect the liquidity value of unused amounts
to increase with the use of the assets.
This is shown in Figure 3.2a and
3.2b for cash and credit. respectively.
The horizontal axes represent
percentages of cash or credit used (moving left to right) or unused
(moving right to 1 eft).
The 1 iquidi ty val ue of cash or c redi t reserved
is measured on the vertical axes.
The resulting curves are continuous.

39
CFAF
CFAF
<Il
~
<Il
~
<Il
tll
<Il
<Il
tll
l-l
<Il
l-l
I::
OM
I::
OM
oI-J
OM
..c
"Cl
tll
<Il
Cl!
l-l
U
U
4-l
4-l
0
0
<Il
<Il
::l
::l
r l
r l
Cl!
Cl!
>
>
1.00
i
O----Cash u s e d - - -t100%
o
Credit used
-100%
100%-
Cash reserved
0
100%-Credit reserved-O
a. Cash
b. Credit
Figure 3.2
Liquidity Value Curves
Source:
Baker and Bhargava [8]

40
nonlinear and positively sloped to reflect the assumption of diminishing
returns from the use of these assets.
The decision maker1s attitude
toward risk will determine the heights and slopes of these curves.
De-
pending on the degree of risk aversion the curves will be higher or
lower. with steeper or flatter slopes.
As well. the curves will be
affected by the asset structure of the firm.
the cost of acquiring loans
and the amount of credit available to the decision maker.
3.1.2
The Liquidity Specified Linear Programming Model
The Liquidity Specified Linear Programming model constitutes a
linear programming model with added liquidity management constraints.
It
can be summarized as follows.
n
f
1,
Maximiz e z =
~
c'X' +
~
J J
~ cn+j+1,+2Z(k-1)xn+j+1,+21,(k-1) (2.1.a)
j=l
k=1
j=l
subj ect to
n
~
a .. x·
(2.1. b)
1.J J
j=l
i=l ••• • • m
1,+1
~ xn+21,(k-1)+j + ~ am+1+(1,+1)(k-1).pxp
= bm+ 1+(1,+1) (k-1)
j=l
p~
(2.1.c)
1,
~ am+2+(1,+1) (k-1).n+21,(k-1)+jXn+21,(k-1)+j +
j=l
= 0
(2.1. d)
am+1+j+(1,+1) (k-1).n+21,(k-1)+jXn+21,(k-1)+j
- xn+j+1,+2Z(k-1)
= 0
(2.1.e)
j=2 ••••• 1,

41
1,
.~ xn+j+1,+21,(k-1)
J=l
> L
(2.1.f)
>
0
where
k subscript denotes the form of liquid assets. and varies from 1 to
f (equals 4 in our case).
1, subscript denotes the group of activities that allocate the liquid
asset between use and reserve.
or activities that value the per-
centages of liquid assets reserved.
p subscript denotes elements of n that contribute to the supply of
the liquid assets.
q subscript denotes elements of n that use the liquid assets.
r subscript denotes elements of n that either demand or supply
liquidity.
z is the objective function.

is an activity alternative in production.
marketing.
consumption
J
and finance.
a·· is an addition to «0) or subtraction from (>0) b:
1.J
am+1+(1,+1) (k-1).p is the amount of liquid asset supplied by 1 unit of
x p• It has a negative value «0).
am+2+( 1,+1) (k-l).n+2Z (k-1)+j is the amount of liquid asset allocated in
the liquid "account."
It has a negative value «0).
am+1+j+(1,+l) (k-1). n+2Z(k-1)+j is the amount of liquid asset allocated
in the liquid asset "reserve" at percentages by 1 unit of Xn+2 1,(k-
l)+j·
Its value is negative «0).

42
am+2+(~+1) (k-1).q
(>0) is the amount of liquid asset used through the
liquid account by 1 unit of x q•
am+f(~+l)+l.r is an addition to «0) or subtraction from (>0) L by 1
uni t of x r'
bi's are the production.
consumption and marketing requirements.
bm+1+(~+1)(k-1)'S are the right hand side values for the liquid assets
supply constraints.
L is the minimum required liquidity to satisfy unexpected demand for
cash from sources which are external to the farming operations.
Equation (2.lob) and the first n terms of Equation (2.1.a) specify
the conventional Linear Programming (LP)
model.
Equation
(2.1.f)
speci-
fies
the liquidity management constraint which transforms the LP model
into a Liquidity Specified Linear Programming (LSLP) model.
Equation
(2.lec) represents the set of liquid asset supply rows; Equation (2.1.d)
defines the set of liquid asset "account" rows and Equation (2.1.e) the
set of liquid asset "reserve" rows.
They are used to allocate units of
liquid asset supplied between liquid "account" where the asset is used in
the farm-firm and household activities. and liquid "reserve" whence it is
valued in percentage increments to satisfy the liquidity reserve require-
ment (Equation (2.1.f)).
These four equations provide for the management
of liquid assets.
here cash and credit.
Figure 3.3 illustrates their LP
Tableau formulation for managing credit.
A similar tableau is con-
strue ted f or cash.
The resul ting segments. together with produc t Lon ,
consumption and marketing activities and requirements specifications. are
used in the final analysis.

Allocate Credit to Credit
Value of Percentage of
Reserve (%)
Credit in Reserve
Gen-
Constraint
erate
Use
Rela-
80
60
40
20
0
100
80
60
40
20
Credit
Credit
tion
Level
Credit
1
1
1
1
1
1
-a
=
b
Credit Account
-.2
-.4
-.6
-.8
-1
j
a
=
0
Credit Reserve:
80%
-.8
1
=
0
60%
-.6
1
=
0
40%
-.4
1
=
0
20%
-.2
1
=
0
Liquidity Reserve
Requirement
1
1
1
1
1
L
Objective Function
.20 .25 .37
.62
1.10
=
M
CFAF
1.20
1.00
.80
.60
.40
.20 ,'"-
...J.
0% - - Used
.100%
1 0 0 % - Reserved -
0%
E;
Figure 3.3
Liquidity Management Vectors for Credit in an LP Model Incorporating Risk Behavior
Source:
Baker [5]

44
In Equation (2.1.f) the value assigned to a m+f( Z+l).r reflects the effect
of activity x r on the risk of the farm-firm and househol~ A relatively
risky activity would have a large negative coefficient.
To activate such
an activity would add a large increment to the liquidity requirement that
must be met.
A small coefficient reflects a relatively small increment
to the liquidity requirement.
All the xn+j+ Z+2Z (k-1) 's have coefficient
one
(1) in the liquidity constraint (2.1.f) showing that the reservation
of liquid asse~s contributes to the satisfaction of the liquidity re-
quirement.
The contribution of liquidity to the objective function is shown in
the last term of z where the cn+j+Z+2Z(k-1)'s denote the liquidity values
of cash and credit reservations.
Al though the values of these coeffi-
cients are chosen somewhat arbitrarily.
they conform to the proposition
of diminishing returns with respect to increasing amounts of liquid asset
reserved.
Recall that the value of liquidity increases with the amount
of liquid asset used for purposes other than reservation.
It reaches a
high level when the firm's liquidity apppoaches zero. and remains low
when the level of reserved liquid asset is high.
In the specific case of
cash. as the level of reservation approaches 100%. the liquidity value
approaches 1.00 monetary unit.
which is the lower limit for the value of
cash in reserve.
The liquidity value of reserved credit is higher. the
fewer the constraints on its use.
and the lower the transaction cost of
borrowing.
Credit at the moneylender better meets these criteria than
government credit which is often administered with costly and time-
consuming access and with restrictions in the use of loan proceeds.
Credit.
however. whatever its source. remains a poor substitute for cash

45
as a source of liquidity.
In addition to liquidity values.
the objective
function includes the transfer of cash from the cash account and the
carry over of unpaid debts.
The liquidity structure of the farm-firm and household is found in
analysis of the value of the objective function which includes liquidity
values as well as net cash flow.
Because the objective function does not
have a counterpart in direct observations. net cash flow and cash in
reserve constitute the most useful objective performance measures to use
in the validation process.
Credit in reserve is another important per-
formance measure.
Cash available is measured by the difference between
the value of the objective function and cash and credit in reserve.
The
difference between cash available and beginning cash measures the net
cash from the optimum plan and constitutes an appropriate "bottom line"
measure with which to validate the model.
The LSLP model. as used in [15. 20.34.45. 62]. is based on the
assumption that any loan is fully repaid by the due date.
This is
reflected in debt-balance rows in which the amount borrowed is set equal
to the amount repaid.
Loans are repaid through a set of repayment acti-
vities which includes advance repayment and repayment at maturity.
There
are as many repayment activities as there are seasons in the debt period.
Table 3.1 displays the LP tableau segment of BNDA debt management in the
LSLP model used by Yabile [62].
In the table. each uni t of loan amoun t
repaid uses (l+i) units of cash either from the cash account row or from
the objective function.
and reduces by 1 unit the debt level through the
debt balance row.
In actuality. however. only certain loans are fully
repaid by their due dates; some are either repaid after the due date in

Table 3.1
BNDA Large Farmer Credit Program:
Debt Management
----.,-----
Borrow From
Reoavment
BNDA
S4
Constraint
Sl Debt
S2 Debt
S3 Debt
Debt
Sl
S2
S3
S4
R12
R13
R14
Rl3
Rl4
RlZ
R34
R3Z
R4Z
Re1ationlLeve1
j
- - - -I
I
Cash Supply ~1
-1
b
S2
-1
o
S3
-1
o
S4
-1
o
Cash Account
S2
( h i
,
o
1)
S3
( h i
(l+i
,
o
1)
n
2)
S4
( h i
(l+i
(l+i
o
1)
2)
3)
BNDA Credit
Account
Sl
1
S2
1
1
-1
S3
1
1
1
-1
-1
-1
S4
1
1
1
1
-1
-1
-1
-1
-1
-1
o
BNDA Debt
Balance
Sl -1
1
1
1
S2
-1
1
1
1
o
S3
-1
1
1
o
S4
-1
1
H
Objective
(l -i" 2)
(l +i~) (l +i 4)
Function
-
-
-
_ '
I
Source: Yabile (62)
r = 11%
The i's are the level of interest rates corresponding to the length of the season beginning the first day the
loan was contracted.
Sj
(j=1.2.3.4)
represent
the seasons.
~
0\\

47
which case loan amounts constitute delinquent amounts.
or they are not
repaid at all in which case the amounts are defaulted.
This is the case
of some of the Ivorian Government Supported Credit Program's (BNDA) loans
which constitute the focus of the current researc~
In order to be able
to use the LSLP model to analyze the default phenomenon. we will extend
it to include default activities. corresponding to these BNDA loans.
3.2
Incorporating Delinquency and Default in the LSLP Model
Default and delinquency influence liquidity as well as being influ-
enced by liquidity.
As shown by Ladman and Tinnermeier [36]. loan delin-
quency and default could provide for mechanisms of income transfers to
the borrower.
In case of delinquency the transfer is temporary and the
borrower gains from improved income or reduced cost resulting from his
control over cash flow.
In case of default.
the income transfer is
permanent and is equivalent to the real value of the loan principal and
the interest charges less any real amount repaid on principal and in-
terest. less the cost of any damage to the borrower's future credit. and
less any other cost to the borrower that the lender recovers from de-
fault. 2
Following the principle of liquidity management. cash unused in
either case contributes either to the cash reserves where it is valued to
satisfy liquidity requirements.
or to the cash accounts whence used for
other activities.
However.
both delinquency and default generate penal-
ties whose costs. often. must be met through cash operations. hence
requiring liquidity.
The farmer's decision to be delinquent or to de-
fault would therefore be determined by his expected gains and the
lender's response that might result in loss of cash as well as credit.

48
More specifically.
default (delinquency) will occur at the point where
the farmer's marginal cost of and marginal revenue from default (delin-
quency) are equal. given that he aims at maximizing profits.
There may be cases of borrowing where the borrower would "take the
money and run" whenever possible.
But when the availability of credit is
vital to the farm-firm and household.
a farmer interested in maximizing
his profit will less likely adopt such a strategy.
Instead. default can
be viewed as a rational strategy for a farmer interested in preserving
his cash flow management ability.
In order to determine such a strategy
delinquency and default variables are introduced in the LSLP model. along
with a constraint which relates the amount defaulted to the amount repaid
after the due date. Le, delinquent amount.
Because the farmer might be
dropped from the credit program at the end of the second year if he has
not repaid the first year loan. the LSLP model is expanded over a two-
year period.
3.2.1
Specification of Delinquency and Default Activities
In the current study. a short delay in the payment of a loan is
considered a delinquency:
otherwise the farmer has defaulted.
More spe-
cifically. payments received within the season following the due date are
considered delinquent.
Any fraction unpaid then is considered in de-
fault •.
In the model. default activities are specified in percentages of
unpaid loan.
The percentages considered are 20. 40. 60. 80 and 100(%).
These correspond to grid points used to approximate the default penalty
cost and default-liquidity-requirement functions.
which are convex

49
functions of default.
But in the model these functions are represented
by line segments whose abscissa are the specified defau1 t levels.
The
convexi ty of these func tions would guarantee that the optimal defau1 t
level is a linear combination of at most two of the specified default
activities.
In the debt-balance row represented by Equation (3.1) below. bor-
rowing activity is set equal to the sum of repayment before and at the
due date. and default activities.
- B + R1 + R2 + ALO + 1/0.2 DEF20 + 1/0.4 DEF40 + 1/0.6 DEF60
+ 1/0.8 DEF80 + DEF100
= 0
(3.1)
Equation (3.2) specifies that the amount of loan defaulted is equal
to the amount of loan which will not be repaid within the season following
the due date.
R3 - ALO - 0.8/0.2 DEF20 - 0.6/0.4 DEF40 - 0.4/0.6 DEF60
- 0.2/0.8 DEF80
= 0
(3.2)
where B is the borrowing activity.
R1 the advance repayment activity.
R2
the repayment at the due date. R3 the repayment past the due date. Le.
delinquent loan.
and DEFa's
[a = 20. 40. 60. 80. 100(%)] the default
activities.
Eq.uations
(3.1)
and
(3.2)
are obtained from the following equations
describing the procedure that allocates each CFAF of loan unpaid before
and/or at the due date between the amount unpaid and the amount repaid
after the due date.

50
- B + R1 + R2 + ALO + AL 20 + AL40 + AL60 + AL80
+ DEF100
= 0 (3.1 )
R3 - ALO - O. 8AL 20
0.6AL40 - 0.4AL60 - 0.2AL80
= 0 (3.2)
- 0.2AL20
+ DEF20
= 0 (3.3)
- 0.4AL40
+ DEF40
= 0 (3.4)
0.60AL60
+ DEF60
= 0 (3.5)
- 0.8AL80
+ DEF80
= 0 (3.6)
where B. R1. R2 and DEFa's are as defined above.
Equation (3.1) rep re-
sents the debt balance account. Equation (3.2) the past due repayment
account and Equations (3.3) to (3.6) the default accounts at 20%. 40%.
60% and 80%. respectively.
The ALa's [a =
O. 20. 40. 60. 80(%)] denote
activities that allocate each CFAF of loan unpaid through R1 and/or R2 at
default and 1-« repaid after the due date.
For instance.
AL20 allocates
units of loan in the debt account to 20% default (in the default account)
and 80% repaid (in the past due repayment account).
The DEFa's allow for
default to reduce credit supply in subsequent seasons and to add to the
1iquidi ty requirements.
The ALa's which are "artificial" variables used to subdivide loans
between the amount paid and the amount defaulted.
can be eliminated as
follows:
Use (3.3). (3.4).
(3.5) and (3.6) to determine
1
AL20
= 0.20
DEF20
1
AL40
=
0.40
DEF40
1
AL60
=
0.60
DEF60
1
AL80
=
0.80
DEF80

51
then substitute these variables in (3.1) and (3.2) to obtain Equations
(3.1)
and (3.2).
respectively.
Because there is no 100% default account.
ALIOO is directly expressed as DEF100 in Equation (3.1) above.
3.2.2
Lenders' Response to Delinquency and Default
Currently. BNDA attempts to control delinquency by charging an addi-
tional 0.7% interest rate per month of overdue on loan principal with a
30 day-grace-period. In the model. this additional interest charge is
used as the delinquency penalty cost.
In the case of default. it is assumed that BNDA responds by reducing
its credit supply in subsequent seasons.
The coefficient of default
activity in the BNDA credit supply row. therefore. represents default
penalty cost.
Because of lack of data on BNDA's quantity-response to
loan default. and to simplify the analysis. we further assumed a constant
marginal cost of default:
that is. each additional unit of defaulted
loan costs the farmer the same amount.
More specifically. it is assumed
that each CFAF of loan amounts defaulted reduces subsequent credit sup-
plies by at least 1 CFAF.
The appropriate marginal cost is then found by
varying the specified level until a default rate comparable to the ob-
served rate is found.
For example.
let BNCSij denote the credit supply in year i
(i=1.2).
season.j (j=1.2.3.4). and DEFij default on i t h year. jth season loan at
a[a = 20. 40. 60. 80. 100) percentage.
If a farmer defaults on the first
season loan of the first year.
the credit supply in the first.
third and
fourth seasons of the second year would be affected as follows:

52
DEFl120 + DEFl140 + DEFl160 + DEFl180 + DEFll100 - TBN1421 + TBN2123
+ BBN21 + O.OlBML21 + O.OlBCB21
s
BNCS21
DEFl120 + DEFl140 + DEFl160 + DEFl180 + DEF11100 - TBN2123 + TBN23 24
+ BBN23 + O. 01BML23 + O.OlBCB23
s,
BNCS23
DEFl120 + DEFl140 + DEFl160 + DEFl180 + DEFll100 - TBN2324
+ BBN24 + O.OlBML24 + O.OlBCB24
s
BNCS24
where BBNij. BMLij and BCBij denote borrowing activities in the i t h year.
jth season fr6m BNDA. moneylender and commercial bank. respectively.
TBN1421 is the activity that transfers BNDA credit from year 1. season 4
to year 2. season 1. TBN2123 transfers BNDA credit in year 2. from season
1 to 3. and TBN2324 transfers BNDA credit in year 2 from season 3 to 4.
3.2.3
Liquidity Effect of Delinquency and Default
In the model.
delinquency is an alternative to
repayment activity.
and as such it is treated as a source of cash liquidity.
Default. how-
ever.
is assumed to require liquidity as a response to default-penalties.
This liquidity requirement could occur either before or when the farmer
actually defaul t s,
But here we assume that the rational farmer who de-
faults does so out of choice.
Such a farmer. knowing the penalty in-
volved. would require liquidity before he actually defaults.
Specifical-
ly.
it is assumed that each unit of loan amounts defaulted adds one unit
to the liquidity requirement to be satisfied in seasons preceding default.
This is reflected in the model by the coefficients -1 of default activi-
ties in the liquidity requirement rows.

53
In the original LSLP model.
the size of the coefficients in the
liquidity requirement rows reflect how risky the activities are:
the
more negative the coefficient.
the riskier an activity.
In the current
specifications. the riskiness of defaul t is reflected in the fact that
defaul t activities corresponding to a given loan appear in several li-
quidity requirement rows.
3.3
The Data
The data used in the study came mainly from a survey of Ivorian
farmers conducted in 1981 by Yabile [62].
The maj or obj ective of the
survey was to gather information which would help evaluate two BNDA
credit programs:
the Pret de Soudure (PS) and the Individual Credit
Program.
Because the primary objective of BNDA in providing these loans
is to increase agricultural production and income.
a survey of farmers
assisted by BNDA was conducted.
The results of the survey were used to
assess the effectiveness of these credit programs.
and thus of BNDA
itself.
Fifty farmers were surveyed in each of four of the BNDA's dis-
tricts:
South.
South-East. East.
and Center.
These farmers were asked
to respond to a set of questions designed to reveal (1) their borrowing
pattern.
(2) their production.
marketing and consumption decisions.
and
(3) how they perceive and value BNDA lending procedures.
The regions covered by the survey were selected on the basis of (1)
the total number of farmers which have used BNDA services. (2) the volume
of loans which has been allocated to the district in the fiscal year
1979-1980. and (3) the default rate.
The South and Center districts were
chosen to analyze the performance of the individual credit program.

54
While large farmers were chosen in the South. medium farmers were se-
lected in the Center.
Default rates were 72% and 15.7% in the South and
Center districts.
respectively.
The East and South-East districts were
selected to evaluate the performance of the PS program whose default
rates were 0.4% and 10%.
respectively.
The current study is restricted to the sample of farmers chosen in
the Center district.
These farmers grow coffee and cocoa for export.
maize. rice. yam. cassava. and plantain for cash and consumption pur-
poses. and the average farm size is 27.33 hectares.
They borrow from
BNDA under the PFVN.
a subprogram of the Individual Credit Program.
Fifty percent of the farmers defaulted on their loans:
7 farmers (14%)
defaulted on the entire amount. and 18 (36%) defaulted on a fraction of
the loan.
Although the data were not primarily collected to ana1yze the
specific problem of farmers' loan defau1 t , they provide. when supple-
mented with data from several BNDA annual reports and statistics of the
Ivorian Ministry of Agriculture. enough information to conduct our study.
The PFVN program.
despite the relatively low default rate shown in
Table 1.3. is chosen (1) because of the number of farmers in the program
(25% of the total number of farmers in the BNDA's credit programs. and
50% of the total number of farmers receiving production loans) shown in
Table 3.2. and (2) because of the increasing volume of loans made as
compared to other programs as illustrated in Table 1.3.
3.4
Characteristics of Bouake Region
When the Bouak a region is enlarged to cover the entire Center re-
gion. it can be viewed as the second largest in Ivory Coast.
in terms of

55
Table
3.2
Number of PFVN Loans as % of (1) BNDA Total Number of Loans
Made and (2) Total Excluding the Pret de Soudure
Fiscal Year
PFVN Loans as % of Total
PFVN Loans as % of Number
Number of Loans
of Loans Excluding PS
1976-1977
23 .07
72.84
1977-1978
29.81
87.49
1978-1979
26.51
67.62
1979-1980
27.89
71.33
1980-1981
28.82
67.31
1981-1982
25.78
45.88
Source:
BNDA Annual Reports
area
(65.490.00
square
kilometers).
and
population
(estimated at
1.679.563.00 inhabitants in 1975).
The economy of the region is still
largely dominated by an agricultural sector based on the production of
coffee. cocoa and cotton.
The region benefits from relatively well-
developed and adequate transportation and communication facilities.
due
mainly to its location in the center of the country. and the city of
Bouake which is the second largest in Ivory Coast.
The population of Bouake region is mainly rural.
In 1975. the rural
population of which 80% lived on agricul t u re, made up about 70% of the
total population.
As shown in Table 3.3 1• these percentages are the same
in 1985.
The climatic dualism due to the fact that the region is covered by
both the forest and the savannah.
provides favorable conditions for the

56
production of forest zone crops such as coffee.
cocoa.
bananas.
cassava
and plantains. as well as savannah crops such as rice.
cotton. maize. yam
and groundnuts.
The forest zone is characterized by two rainy seasons
a1 ternated by two dry seasons.
The savannah subregion is characterized
by two seasons:
rainy and dry seasons.
Table 3.4 illustrates the sea-
sonal specification of the region.
Bouake region has a significant percentage of fertile land.
Average
farm size.
however.
is small due to the high percentage of rural popula-
tion (80% of the population).
Table 3.5 1 shows that average farm size in
the region is the smallest in the country.
Areas planted in export crops
and food crops are nearly equal; so is their annual growth rate.
Land
utilization rate. which is the ratio of the sum of land under cultivation
and fallow land.
and total land available for agriculture.
appears very
high.
Although i t is believed that the country in general. and Bo uak e
region in particular have abundant land.
the 3.4% annual growth of land
utilization rate shown in Table 3.3 might lead to a severe land con-
straint in the near futur~
Indeed. Bouake region. like other regions in
Ivory Coast. is characteriz ed by the tradi tiona1 system of land tenure
which is basically a "freehold" form.
In it. each community owns the
land.
Each household in the community is given a fraction of the land
whose right is passed from generation to generation.
and father to son.
Fallow land is commonly found in the region especially in the southern
subregion which has an important forest area.
Such a pattern of land
utilization.
which explains the high rate of land utilization.
is main-
tained at the expenses of forest areas and forestry resources.

57
Table 3.3
Indicators of the Agricultural Sector:
The Bouake Region
Indicators
1975
1985
Annual
Growth
Rate (%)
Rural Population (thousands)
547.00
547.00
0.00
Agricultural Population (thousands)
489.00
489.00
0.00
Active Rural Population (thousands)
222.00
222.00
0.00
Active Pop/Rural Population (%)
40.60
40.60
0.00
Active Pop/ Agrictil tural Population (%)
45.50
45.50
0.00
Ag. Population/Rural Population (%)
89.40
89.40
0.00
Total Area Cultivated (thousand of
hectares)
224.00
317.00
3.50
- Area on Food Crops
102.00
143.00
3.40
- Area on Export Crops
122.00
174.00
3.60
Area per Productive Ag. Population
(hectare)
1.00
1.43
3.60
- on Food Crops
0.46
0.64
3.40
- on Export Crops
0.54
0.78
3.80
Labor Requirement per Active Rural
Inhabitant and per Year (days of
work)
156.00
217.00
3.40
- Labor Required for Food Crop
Product
128.00
177 .00
3.30
- Labor Required for Export
Crop Product
28.00
40.00
3.60
Land Utilization Rate (%)
58.30
81.70
3.40
Source:
Fran90ise Binet. Bilan National de l'Emploi en Cote d'Ivoire.
Mai 1982 [16]

58
Table 3.4
Seasonal Specification in Bouake Region
SEASONS
Identi-
Specifi-
Length
Production
Labor
fication
cation
Months Days
Characteris-
Operation
Supplieda
Number
tics
(Working
Days)
1
March and April
2
61
Light Rains
Clearing
33.42
the Land
Planting
Fertilizer
2
May and June
2
61
Short Dry Sea-
Weeding
33.42
son
Harvest of
Maize
3
July-October
4
123
Heavy Rains
Pest & Dis-
67.40
ease Control
Beginning
Harvest
4
November-February
4
120
Long Dry Sea-
Harvest
65.75
son
Fermentation
and Dryingb
Marketing
aLabor supplied per season is computed based on a total of 200 working
days a year defined in [16].
bOnl y the technology of cocoa processing requires fermentation and drying
stages:
the cocoa beans must be fermentated for 6 to 7 days then dried
for 8 to 15 days under the sun (the artificial drying technique is much
shorter).
The processing of coffee requires that the beans are only
dried then crackled.

Table 3.5
Farm Characteristics in the Major Regions of Ivory Coast in 1975
Regions
South
South-West
West
Bouake
Center-
Korhogo
North
Whole
North
Country
Number of Farms
318.800
23.500
80.300
78.800
54.400
27.500
27.700
611.000
Average Farm Size
(hectare)
6.77
3.70
4.19
2.84
3.32
4.42
4.41
5.29
Average Family Size
6.60
5.60
6.70
6.20
7.10
7.70
10.10
6.80
Family Workers
3.10
2.70
3.10
2.80
3.70
4.10
5.10
3.20
Average Annual
Income (CFA)
274.000.00 45.500.00
82.000.00
117.000.00
96.700.00 236.900.00 143.900.00
197.400.00
Average Income per
Family Member
41.500.00
8.100.00
12.300.00
18.900.00
13.600.00
30.800.00
14.800.00
29.000.00
Average Income per
Worker
88.400.00 16.800.00
26.500.00
41.800.00
26.100.00
57.800.00
28.200.00
61.700.00
Source:
Franc;oise Binet. Bilan National de l'Emploi en Cote d'Ivoire. Mai 1982 [16].
\\J1
10

60
Labor in the agricultural sector is supplied primarily by the household.
Adult family members and children 7 years old and above make up the on-
farm labor force.
On average. an adul t can provide 200 days of work a
year.
Labor required for agricultural activities.
which was below the
200 working-days supplied in 1975. would necessitate off-farm supplies by
1985 as shown in Table 3.3.
Off-farm labor can be acquired through labor
hiring from families in the northern region of the country.
Because of
the time lag in farming operations in the Bouake region and the North.
off-farm labor is considered available in abundance.

61
NOTES
1. Bouake region.
as presented in both Table 3.3 and Table 3.5.
cor-
responds to the city of Bouake and the surrounding villages.
the
Center region as a whole comprises Bouake. Bouaf1e. Dabaka1a. Dimbok-
ro and Katio1~
2. An example of such costs is the value of any asset pledged as col-
lateral to secure the 10a~
In general.
the lender takes possession
of this asset when default occurs.
Hence.
one would expect a pledge
of a high-valued asset to increase the loss of the borrower from
default.
thus lower the incentive to default.

62
CHAPTER FOUR
MODEL
4.1
An Overview
The model used here is the LSLP model modified to include delinquen-
cy and default activities. and expanded over a period of 2 years in order
to account for the time dependence of the default consequences.
For our study. we have chosen a representative farm-firm and house-
hold in the Center region of Ivory Coast.
The household comprises 8
persons. 5 of which are children 19 years old and under.
The average
farm size is 27.3 hectares.
Farming in the Center region follows a
pattern of four seasons a year.
As shown in Table 3.4.
most of the farm
work is performed during seasons 1. 3 and 4 which correspond to the short
rainy season. the heavy rain season and the long dry season. respective-
ly.
In order to account for these production seasons. we have specified
the model over an 8-season-p1anning period.
The emphasis on studying financial aspects of the farm has limited
detail in production alternatives that can be allowed for in the model.
As presented in the preceding chapter.
the Center region is suitable for
the production of export crops such as coffee and cocoa. and food crops
such as cassava. yam.
plantain.
maize.
rice.
groundnuts and vegetables.
In the .current study. however. we only consider the following crops:
coffee. cocoa. cassava. yam. rice and maize.
But since the model only
spans over a two-year period. coffee and cocoa. which on the average are
25 year-crops.
will be treated as constants throughout and specified at
the values found in the survey.
The right hand sides of the land. 1abor.

63
and cash constraints where these activities normally appear. will be
modified accordingly.
The farmer may borrow from both informal and institutional sources.
BNDA and commercial banks constitute the institutional sources of loans;
the informal sources may range from relatives to local merchants or
moneylenders.
The BNDA credit.
which is primarily for production pur-
poses. is available in seasons 1. 3 and 4 each year.
The commercial bank
and the moneylender offer loans in each of the 4 seasons of the year. and
the farmer is free to use the proceeds as he chooses.
Statistics show
that farmers very often repay this form of loan in full. but only repay a
fraction of the BNDA loans.
In order to account for such a repayment
behav i.or;
we have introduced delinquency and default activities in the
original LSLP model to fit the current analysis.
In the credit supply
constraints.
the coefficients of these default activities are chosen to
reflect BNDA's response to defau1 t , while in the liquidity requirement
constraints.
they are chosen to reflect the farmer's reaction to default
penalties.
We have mode1ed default as a time dependent phenomenon.
But since
the model only covers a short period of time.
time discounting has been
ignored.
Mathematically. the model can be described as follows:
Maximize
t
n
f
t
1,
z =
1:
1: c~x~ +
s
s
1:
~
.~cn+j+1,+21,(k-1)xn+j+1,+21,(k-1)
(4.1.a)
s=l j=l J J
k=l s=l J=l

64
subject to
n-1
I
a~.x~
s b~
(4.Lb)
j=l ~J J
>
~
i=l ••••• m-1
= b:+1+(L+1) (k-1) (4.1.c)
L
s
s
.I am+2+(L+1) (k-1).n+2L(k-1)+jXn+2L(k-1)+j
J=l
(4.Ld)
s
s
am+1+j+(L+1) (k-1).n+2L (k-1)+jXn+2L (k-1)+j
- x
(4. L e)
n+j+L+21 (k-1)
= 0
j=2 ••••• L
L
d
s
.I x n+j +L+2L(k- 1) -
I x n+1+j +2Lf
J=l
j=l
LL
d
~ as+4
s+4
= 0
(4.1.g)
u
m.n+1+j+2Lfx
n+1+j+2Lf
j=l
s=1 ••••• t-4
d
s
s
s
x n+1+2U +
.I am+ f(L+1)+2.n+1+j+2Lfxn+1+j+2Lf
= 0
(4.1. h)
J=l
s=5 ••••• t
where superscript s denotes the seasons. subscript v denotes BNDA loan
repayment activities. d is the number of default activities corresponding
to loan made in the s th season of the first year. t is the total number
of seasons in the two-year planning period. xn is the activity indicating
the amount borrowed irom the BNDA in the first year. x n+1+2Lf is the

65
.
, 2
h
activity representing the de.l Lnquen t amount and the x n+1+j+2Lf s are t e
default activities.
The other subscripts and activities are as defined
in the preceding chapter.
Equations (4.1.a) through (4.1.f) also have
been defined in Chapter Three.
(4.1.g) represents the modified debt-
balance equation and (4.1.h) is the past-due-repayment account for BNDA
loans.
4.2
Objective Function
The obj ective function to be maximized is similar to that of the
original LSLP model.
It combines liquidity values. net cash flow. and
value of debt outstanding at the end of the model period.
Net cash flow
corresponds to the level of the activity that transfers cash from the
last season cash account to the obj ective function.
Liquidity values
represent the total values of cash and credits in reserves.
Although the
model is specified as a multiperiod LP model. the objective function does
not include discounted values mainly due to the relatively short 2-year
planning span.
Debt that is unpaid at the end of the model period. and
must be carried over. is priced at an appropriate interest rate in the
obj ective function.
4.3
Constraints
The constraint sets of the model are summarized in Tables 4.1 and
4.2.
The first set includes the production. marketing and consumption
constraints.
The second set comprises the financial constraints.
Before
describing the constraints in each of these sets. one should note that
the production activities only include the following food crops:
cassa-

66
Table 4.1
Description of Production.
Marketing and Consumption Con-
straints
ROW
CONSTRAINT
Identifi-
b
cation
Description
Relation
Level e
Unit
LANDA
Land available for root crops (by year)
L
(1)
hectare
LANDB
Land available for grain (by year)
L
(2)
hectare
FLAB
Family labor supply (by season)a
LC
(3)
manday
d
YAMI
Yam inventory (by season)
E
(4)
kilogram
d
MZEI
Maize inventory (by season)
E
(5)
kilogram
d
RCEI
Rice inventory (by season)
E
(6)
kilogram
CASSI
Cassava inventory (by season)
Ed
(7)
kilogram
PROTN
Protein requirement (by season)
G
(8)
kilogram
CALCM
Calcium requirement (by season)
G
(9)
gram
IRON
Iron requirement (by season)
G
(10)
gram
THIAM
Thiamine requirement (by season)
G
(11)
miligram
MAXRCE
Maximum rice restraint (by year)
L
o
hectare
aRefers to the 8 seasons of the 2-year planning period; each year com-
prises 4 seasons.
bE is "Equal to".
L is "Less than or Equal to". G is "Greater than or
Eq ual to".
c Th e rows in the first and fourth season each year are "Greater than or
Equal" relations.
dTh e inventory row of the fourth season in each year is a "Greater than
or Equal" relatio~

67
Table 4.1 (continued)
e
SEASONS
1
2
3
4
5
6
7
8
(1)
1.71
1.71
(2)
1.68
1.68
(3)
97.76
56.23
72.29
252.91
106.03
53.20
64.86
271.51
(4)
1996.80
-
0
0
1996.80
1996.80
0
0
1996.80
(5)
181.96
0
0
181.96
181.96
0
0
181.96
(6)
184.28
0
0
184.28
184.28
0
0
184.28
(7)
419.50
0
0
419.50
419.50
0
0
419.50
(8)
16.80
33.60
33.60
33.60
16.80
16.80
33.60
33.60
(9)
211.43
211.43
422.87
422.87
211.43
211.43
422.87 422.87
(10)
11.54
11.54
23.09
23.09
11.43
11.43
23.09
23.09
(11)
378.36
378.36
756.71
756.71
378.36
378.36
756.71 756.71

68
Table 4.2
Description of Financial Constraints
ROW
CON5TRAINT
Identifi-
cation
Description
Re1ationc
Level d
C5H
Cash supply (by season)a
E
(1)
CFAF
CAC
Cash account (by season)
E
(2)
CFAF
CR
Cash reserves: 20.40.60.80(%)
(by season)
E
o
CFAF
BNCL
BNDA credit supply (51.53.54.55.
57.8)b
L
(3)
CFAF
BNCA
BNDA credit account (51.53.54.55.
57.58)
E
o
CFAF
BNCR
BNDA credit reserves: 20.40.60.80
(%)(51.53.54.55.57.58)
E
o
CFAF
MLCL
Moneylender credit supply (by season)
L
(4)
CFAF
MLCA
Moneylender credit account (by
season)
E
o
CFAF
MLCR
Moneylender credit reserves: 20.40.
60.80(%)
(by season)
E
o
CFAF
CBCL
Commercial bank credit supply (by
season)
L
(5 )
CFAF
CBCA
Commercial bank credit account (by
season)
E
o
CFAF
CBCR
Commercial bank credit reserves: 20.
40.60.80(%)
(by season)
E
o
CFAF
LQRR
Liquidity reserve requirement (by
season)
G
(6)
CFAF
BND
Debt: BNDA loans (51.53.54.55.57.58)
E
o
CFAF
MLD
Debt: Moneylender loans (by season)
E
o
CFAF
CBD
Debt: Commercial bank loans (by
season)
E
o
CFAF
BNRACC Past due repayment account: BNDA
loans (55.56.58)
E
o
CFAF
MAXTRA Maximum credit transfer (by source.
by season)
L
o
CFAF
aCorresponds to the 8 seasons of the 2-year planning period.
b5idenotes the i t h season in the planning period.
cE corresponds to "Equal to".
L is ''Less than or Equal to" and G ''Greater
than or Equal to".

Table 4.2 (continued)
d
SEASONS
,
1
2
3
4
5
6
7
8
(1) 000
1.439.557
0
0
1.293.800
0
0
0
1.293.800
(2) 000
98.8767
109.8767
216.1085
230.4209
98.8767
109.8767
216.1085
230.4209
(3) 000
282.2942
-
179.7140
333.3186
313.9393
-
199.8599
370.5368
(4) 000
17.0499
25.5750
42.6250
51.1500
17.4322
26.1484
43.5807
52.2968
(5) 000
24.5138
36.7707
61.2845
73.5414
25.0634
37.5951
62.6585
75.1902
(6) 000
44.3105
106.9556
129.8411
61.8000
44.0337
106.6036
129.2674
61.8000
e1 CFAF = US $0.0025 (in 1985).
(7\\
\\0

70
va, yam,
rice and maize.
Export crops such as coffee and cocoa are
specified at values observed in the survey due to the short period
covered by the model.
The right hand sides of the land, labor, and cash
constraints where these activities actually appear have been modified
accordingly.
4.3.1
Production, Marketing and Consumption Constraints
Land.
In order to allow for maize/rice rotation,
farm land avail-
able for food crop production is organized into two "blocks":
Land A and
Land B.
Land A is used to grow yam or cassava, and Land B to grow maize
and rice.
On each block the land supply is estimated at 1.71 and 1.68
hectares a year, respectively.
This subdivision of land results from the
fact that yam or cassava is grown on the land the whole year, and maize
and rice are rotated on the same land in the year.
Lab or,
There are three main sources of labor:
(l) on-farm adu1 t-
family-members,
(2) on-farm children 7 years old and above, and (3) off-
farm hired 1abor.
These three sources have been assumed homogeneous
throughout.
Hence, 1abor constraints are defined as family 1abor.
Be-
cause off-farm 1abor is relatively abundant in the region,
hiring needs
not be constrained by the supply.
In estimating the 1abor available each
year, we have assumed a 200 working-days year.
We have also assumed that
adult males and females devote all their time to farming,
thus supplying
the entire amount of 200 working-days each year,
and that children only
work a quarter of the time.
The estimated annual 1abor supply has been
then allocated among the four seasons of the year proportionately to the
length of each seaso~

71
Labor supply rows in seasons 1 and 4 were originally specified as
"less than or equal" constraints.
But after specifying cocoa and coffee
at their observed values.
these constraints were converted to "greater
than or equal" constraints in order to have positive right hand sides.
Food-Crop Inventories.
Food-crop inventory
rows are specified for
each season.
Through these rows. the food produced or purchased in a
season is allocated for consumption or marketing.
or is transferred to
the following season.
In the fourth season of each year. food stocks are
bui1 t
in order- to be used as beginning stocks in the following year.
Because a minimum stock level must be met.
the crop inventory row cor-
responding to this season is specified as a "greater than or equal"
constraint.
The other crop inventory rows all are equality constraints.
Dietary Requirements.
The dietary requirements specified in each
season represent the household's demand for some basic nutrients.
The
required levels. which all are minimums. have been computed following the
FAO nutrient recommendations for Afric~
The nutrients considered in the
model are protein. calcium. iron and thiamine. chosen on the basis of
their relatively low level in the household's diet.
Maximum Rice Restraints.
The maximum rice restraints are means of
ensuring that growing-rice follows growing-maize on the land in each
year.
They are "giving and receiving permission" types of constraints
described by Barnard and Nix [9]
to allow for crop rotation on farm 1an~
They all are "less than or equal" constraints. specified in the current
model so that 1.22 hectares of maize is followed by 1 hectare of rice in
each year.

72
4.3.2.
Financial Constraints
The financial constraints make up the majority of the model's con-
straints conforming with the emphasis of the research on studying finan-
cial aspects of the farm-firm and household.
Each of these constraints
may be defined as follows.
Cash Constraints.
Cash is available to the farmer at the beginning
of the planning period.
This amount of cash is estimated from the survey
at 35% of the farmer's off-farm income and the total of his savings.
Cash requirement~ in other seasons are met by between-season cash
transfers. within season borrowing and cash obtained from crop sales.
The right hand sides of these rows normally are zero.
But after speci-
fying coffee and cocoa at their observed values. the right hand side of
the fourth season's cash supply row in each year becomes positive.
cor-
responding to the amount of cash obtained from the sales (which occur
only in the fourth season of the year) of these crops.
In each season. constraints also are specified for cash account and
cash reserve.
Cash accounts provide for cash that can be supplied to
farm activities requiring cas~
They are equality constraints and their
right hand sides represent the amounts of cash required for producing
coffee and cocoa. and for family additional expenses not specified in the
model.
Cash reserves are used in cash-reserve valuatio~
Cash is spe-
cified -to be reservable at values that increase (decrease) with the
proportion of cash allocated to the cash account (cash reserve).
The
trade-off between the reward from using cash in the farm and the reward

73
from holding cash in reserve would determine the proportions of cash in
the account and in reserve.
Credit Constraints.
Credit is available to the modeled borrower at
the BNDA. a commercial bank and a moneylender.
At the BNDA. borrowing
can occur on the first day of each of three seasons:
1. 3 and 4.
Hence.
credit limits at the BNDA are specified in credit supply constraints for
each of these three seasons and f or each year.
At the commercial bank
and at the moneylender. borrowing can occur on the first day of each of
all four seasons every year.
Hence. credit limits at the bank and at the
moneylender are specified in credit supply constraints for the commercial
bank and the moneylender. respectively. for each season.
Only the credit
supply at the moneylender is given by the survey.
At the BNDA and the
commercial bank. data were available only on the amounts borrowed.
Hence.
the credit supply at the BNDA and at the commercial bank has been
estimated assuming that the ratio of credit available and amount borrowed
at these two sources is identical to that at the moneylender.
Their
credit supply. therefore. was obtained through mul tiplying the amount
borrowed by the ratio of credit available and amount borrowed at the
moneylender.
BNDA credit supply was assumed to grow at an annual 11.21%
[18].
Year to year variation in the supply of credit at the commercial
bank and at the moneylender was estimated at 2.67% and 2.75%. respective-
ly.
For each source and for each season.
constraints as well are speci-
fied for credit accounts and credit reserves (those for BNDA will be
introduced only as means for determining the policy reform in Chapter
Six) defined for 20. 40. 60. and 80{%) reservation levels.
The con-

74
straint levels are zero.
Credit accounts are drawn upon by borrowing
activi ties.
Credit reserves are used in credit reserve valuation.
Credit is specified to be reservable at values that increase the larger
(smaller) the proportion of the credi t supply that is allocated to the
credit account (reserve).
The credit reservation values are critical in
that they reflect opportunity costs subject to which credit supply is
allocated to credit account whence it is available to support borrowing
by the borrower.
Thus credit supply is allocated to credit account in
response to rewards from borrowing,
taking interest costs into account.
Credit supply is allocated to credit reserve in response to rewards from
holding credit in reserve - as a source of liquidity.
Maximum Credit-Transfer Constraints.
The maximum credi t-transf er
constraints provide for means of ensuring that the amount of credit
transferred each season corresponds to the amount of credit that has not
been used in borrowing.
They are therefore structured so that the amount
of credit transferred from one season to the next is limited to the
amount borrowed in the season of origi~ At each source of credit,
there
are as many maximum credit-transfer constraints as there are credit-
transfer activities.
They all are specified as "less than or equal"
constraints since the borrower may choose to transfer less credit in each
season.
Li~uidity Reserve Requirements.
In each season,
the farm-firm and
household requires some minimal reserve of cash or liquidity in order to
respond to unexpected demands for cash as these occur.
The required
levels have been estimated from the survey.
Their seasonal variation,
which is as well determined by the survey, reflects the different risk

75
factor in the seasons.
For instance. atmospheric conditions in the
second season of the year. if too dry. may significantly affect coffee
and cassava yields.
Also. too much rain in the third season may favor
the development of various plant pests requiring substantial phytosan-
itary treatments in order to protect the trees.
the beans and the grains.
Consequently. high liquidity levels have been specified for seasons 1 and
3 in each year.
In each season.
the specified level is subj ec t
to
increase from risky ac tivi ties and decrease from (1) reserved cash and
(2) other activities contributing liquidity.
The liquidity reserve re-
quirements are defined as "greater than or equal" constraints.
Debt-Balance Requirements.
The debt-balance rows include the BNDA.
the moneylender and the commercial bank debt-balance accounts.
These
requirements are specified for each borrowing season and expressed as
equality constraints with zero right hand sides.
Each row is structured
so that the amount borrowed is equal to the amount repaid and/or de-
faulted:
the amount borrowed is equal to the amount repaid in case of
full-repayment.
and equal to the amount repaid and defaulted in case of
loan default.
According to the survey. the farmer repays the full amount
of moneylender and commercial bank loans.
but he only repays a fraction
of BNDA loans.
Hence. only the debt-balance rows corresponding to the
BNDA loans reflect a relationship between borrowing.
repayment and de-
fault activities.
Because the majority of the BNDA loans made in the
second year mature after the model-period and because the consequence of
default is accounted for only after the due date of the loans. we have
not specified default activities corresponding to BNDA loans extended in
the second year.
BNDA debt-balance rowS in the second year.
therefore.

76
are similar to the moneylender and the commercial bank debt-balance
accounts.
BNDA Past-Due Repayment Accounts.
The past-due repayment accounts
for BNDA loans ensure that the amount of loan defaulted corresponds to
the amount of loan which has not been repaid a season after the due dat~
These constraints have been specified for loans given by the BNDA in the
first.
third and fourth seasons of the first year.
and expressed as
equality constraints with zero right hand side values.
4.4
Activities
Two sets of activities have been considered in the model.
These
correspond to the production.
marketing and consumption activities.
and
the financial activities displayed in Tables 4.3 and 4.4. respectively.
The activities in each of these classes are explained below.
4.4.1
Production. Consumption and Marketing Activities
Production Activities.
The production activities have been speci-
fied only for cassava. maize. rice and yam which all are cash and con-
sumption crops.
Cocoa and coffee which are two other important produc-
tion alternatives have been accounted for in the model only as constants
owing to their life span (25 years on average) that exceeds the model
period ~2 years).
These have been set at their values as reported in the
survey.
Hence. the hectares of cocoa and coffee actually grown have been
deducted from total farm land.
The 1abor required by cocoa and coffee
production has been estimated and subtracted from the 1abor supplies.
Also their requirements in cash have been estimated and expressed as the

77
Table 4.3
Description of Production. Marketing and Consumption Activities
COLUMN
Identifi-
cation
Description
Activi ty Unit
CASS
Growing cassava (by season)a
hectare
MAIZE
Growing maize (by season)
hectare
RICE
Growing rice (by season)
hectare
YAM
Growing yam (by season)
hectare
HLB
Hiring labor (by season)
kilogram
SLCASS
Selling cassava (S3. S4. S7. S8)b
kilogram
SLMAIZE
Selling maize (S2. S6)
kilogram
SLRCE
Selling rice (S4. S8)
kilogram
SLYAM
Selling yam (S3. S4. S7. S8)
kilogram
PUCASS
Purchasing cassava (Sl. S2. SS. S6)
kilogram
PUMAIZE
Purchas ing maiz e (si , S3. S4. SS. S7. S8)
kilogram
PURCE
Purchasing rice (si , S2. S3. SS. S6. s7)
kilogram
PUYAM
Purchasing yam (Sl. S2. SS. S6)
kilogram
PUFISH
Purchasing fish (by season)
kilogram
TRCASS
Transferring cassava (Sl. S2. S3. SS. S6. S7)
kilogram
TRMAIZE
Transferring maize (si , S2. S3. SS. S6. S7)
kilogram
TRRCE
Transferring rice (si , S2. S3. SS. S6. S7)
kilogram
TRYAM
Transferring yam (si , S2. S3. SS. S6. S7)
kilogram
CCASS
Consuming cassava (by season)
kilogram
CMAIZE
Consuming maize (by season)
kilogram
CRCE
Consuming rice (by season)
kilogram
CYAM
Consuming yam (by season)
kilogram
aCorrespond to the 8 seasons of the 2-year planning period:
each year
comprises 4 seasons.
bSi is the i t h season in the planning period.

78
Table 4.4
Description of Financial Activities
COLUMN
Identifi-
cation
Description
Activity Unit
ACSH
Allocating cash:
0.20.40.60.80Ct)
(by season)a
CFAF
ABNC
Allocating BNDA credit: 0.20.40.60.80(%) (by season)
CFAF
AMLC
Allocating moneylender credit: 0.20.40.60.80(%)
(by season)
CFAF
ACBC
Allocating commercial bank credit: 0.20.40.60.80(%)
(by s.eason)
CFAF
VCSH
Valuating cash reserve: 20.40.60.80.100(%)
(by
season)
CFAF
VBNC
Va1uating BNDA credit reserve: 20.40.60.80.100(%)
(by season)
CFAF
VMLC
Va1uating moneylender credit reserve: 20.40.60.80.
100(%) (by season)
CFAF
VCBC
Va1uating commercial bank credit reserve: 20.40.60.
80.100(%) (by season)
CFAF
TRCSH
Transferring cash (by season)
CFAF
TRBNC
Transferring BNDA credit (Sl. S3. S4. SS. S7)b
CFAF
TRMLC
Transferring moneylender credit (Sl. S2. S3. S4. SS.
S6. S7)
CFAF
TRCBC
Transferring commercial bank credit (Sl. S2. S3. S4.
SS. S6. S7)
CFAF
BBNC
Borrowing from BNDA rsi , S3. S4. SS. S7. S8)
CFAF
BMLC
Borrowing from the moneylender (by season)
CFAF
BCBC
Borrowing from the commercial bank (by season)
CFAF
RBND
Repaying BNDA debt (S2. S3. S4. SS. S6. S7. S8. Zc)
CFAF
RMLD
Repaying moneylender debt (S2. S3. S4. SS. S6. S7. S8.
Zc)
CFAF
RCBD
Repaying commercial bank debt (S2. S3. S4. SS. S6. S7.
S8. Zc)
CFAF
DEF
Default variables: 20.40.60.80.100(%) (SS. S6. S8)
CFAF
aCorrespond to the 8 seasons of the 2-year planning period:
each comprises
4 seasons.
bSi represents the i t h season in the planning period.
cCorrespond to the objective function.

79
right hand side values of cash account rows.
The cash obtained from the
sale (undertaken in the fourth season of each year) of these two crops
has been introduced as the right hand side value of the cash supply rows
in the fourth season of the year.
Growing cassava, maize, rice or yam requires cash and labor, and the
produce contributes to the crop's inventory whence it is either consumed,
sold or transferred to the next season's inventory.
These activities all
are defined in a planted-harvested-hectare unit.
Cassava or yam is grown
on the land the whole year.
Maize and rice,
however,
are rotated on the
same land in the year, with maize followed by rice.
Food-Transfer Activities.
The food-transfer ac tivi ties serve to
pass cassava, maize, rice or yam from the inventory row in a season
through to the inventory row in the next season.
For instance, yam-
transfer activities move quantities of yam between yam inventory rows
each year:
from season 1 to 2, 2 to 3, and 3 to 40
Each of the food-
transfer ac tivi ties,
theref ore,
has a positive coefficient in the
"giving" inventory row (season 1 inventory row for example) and a nega-
tive coefficient in the "receiving" inventory row (season 2 inventory row
for instance).
These activities are assumed to bear zero costs.
The
activity unit is 1 kilogram.
Consumption Activities.
Cassava, rice, maize or yam available in a
season may be used in consumption to satisfy nutrient requirements in
that season.
Hence, consumption draws on crop inventories and con-
tributes to satisfy nutrient requirements that must be met in the season.
This is reflected in the positive coefficients of these activities in the
inventory and nutrient requirements rows.
Corresponding to each con-

80
sumption activity.
the coefficient in each nutrient requirement row
represents the amount of nutrient supplied by 1 kilogram of food intake.
For instance. each kilogram of maize consumed reduces maize inventory by
1 kilogram and provides 0.078 kilogram of protein. 0.070 gram of calcium.
0.050 gram of iron and 0.20 miligram of thiamine.
Selling Activities.
Rice.
maize. yam or cassava produced may also
be sold.
Because either selling or purchasing activity might be redun-
dant if these two are simultaneously specified for the same commodity in
a given season. selling activities have been defined only in seasons
where the commodity is harvested.
The reasoning behind such a specifica-
tion is that the farmer would sell only if there is an excess supply of
the commodity. in which case it would not be necessary to purchase addi-
tional amounts:
this would be the case in the season where the commodity
is being harvested.
Yam and cassava are harvested in seasons 3 and 4
each year. maize in season 2 and rice in the fourth season of the year.
Hence. selling activities have been specified for these seasons.
corres-
ponding to each commodi ty.
In each case. 1 kilogram of commodity sold
reduces the inventory by 1 kilogram and contributes to the cash supply of
the season where the activity is undertaken.
Uncertainties in the prices
of the crops are reflected in negative coefficients of selling activities
in the liquidity requirement rows.
Purchasing Activities.
Additional food and nutrients may be pur-
chased by the household.
The purchasing activities are specified by food
crop and by season.
Because nutrients often are not sold as such. a food
item. here fish.
considered as an important source of the nutrients.
has
been chosen as the off-farm source of nutrients.
While each kilogram of

81
cassava.
maize.
rice or yam purchased contributes 1 kilogram to the crop
inventory whence it is converted through consumption into nutrient equi-
valent to satisfy the nutrient requirement.
each kilogram of fish pur-
chased is directly converted into nutrient equivalents in the nutrient
requirement rows.
That is. each kilogram of maize purchased. for ex-
ample. would add 1 kilogram to the maize inventory whence it would be
consumed.
However.
each kilogram of fish purchased would directly con-
tribute 0.128 kilogram of protein. 0.140 gram of calcium. 0.009 gram of
iron and 0.020 miligram of thiamine.
Labor Hiring Activities.
Labor can be hired in each of the four
seasons of the year.
Hired labor is assumed supplementary to and homo-
geneous with family labor available in each season.
Hence.
each manday
of labor hired contributes 1 manday to the family labor available to
support all the activities requiring labor.
Because of its relative
abundance in the region.
we have not specified labor hiring supply con-
straints.
Thus the amount of labor hired would depend only on the cost
of acquiring it. i.e. the agricultural wage rate prevailing in the re-
giono
4.4.2
Financial Activities
Like the financial constraints.
the financial activities constitute
the la~gest group of activities in the model.
conforming with the focus
of the study.
The activities in the various categories are described as
follows.
Allocation Activities.
Cash or credit may be used either in re-
serves whence valued to satisfy liquidity requirements.
or in accounts to

82
provide for the quantity of cash or credit that can be supplied for farm
activities.
That is. each CFAF of cash or credit supplied is allocated
into reserves and accounts; this is done using the allocation activities.
These activities are specified in percentages of cash or credit supplied.
The percentages considered are O. 20. 40. 60 and 80(%) corresponding each
to an allocation activity.
At a given percentage level. the fractions in
reserve and in use must add up to one. as shown by the coefficient 1 of
these activities in the cash or credit supply rows.
For instance. if
cash is allocated at 20% level. then 20% is reserved and 80% is used
through the cash account row.
In the case of the BNDA credit. allocation (to be considered only in
Chapter Six) can occur in each of three seasons a year:
1. 3. and 4.
For the commercial bank and the moneylender credit. allocation occurs in
each of the four seasons of the year.
Cash also can be allocated between
use and reserve every season.
Valuation Activities.
Credit or cash in reserves is valued at
values that increase the larger (smaller) the proportion of cash or
credit supply that is allocated to the account (reserve).
Hence. the
valuation activities are means for providing values for cash or credit in
reserve and allow for cash or credit reservations to satisfy the liquidi-
ty reqUirement.
The reservation values corresponding to 20.
40. 60. 80
and 100~%) of cash or credit reserve are shown in Table 4.5.
These
values.
which are generated from the model in conformance comparisons.
vary with the percentages. the seasons and source of liquidity.
Cash is
considered as the most valuable source of liquidity.
BNDA credit. as
will be presented in Chapter Six. is modeled as the least valuable source

83
Table 4.5
Reservation Prices of Cash and Credit
SEASONS
"Liquid"
Percent-
Asset
ages
1
2
3
4
5
6
7
8
20
2.50
3.10
3.05
3.00
2.50
3.05
3.10
3.00
40
2.25
2.70
2.65
2.55
2.25
2.65
2.70
2.55
.c
Ul
60
1.80
2.20
2.15
2.10
1.80
2.15
2.20
2.10
III
U
80
1. 70
2.00
1.95
1.75
1. 70
1.95
2.00
1.75
100
1.65
1.95
1.90
1.70
1.65
1.90
1.95
1.70
Qt o~
20
0.90
1.00
1.05
0.95
1.10
1.20
1.25
1.15
oM "tl
40
0.60
0.70
0.75
0.65
0.80
0.90
0.95
0.85
u Q,l
~
~
60
0.25
0.45
0.50
0.30
0.45
0.65
0.70
0.50
Q,lU
80
0.15
0.30
0.40
0.20
0.35
0.55
0.60
0.40
§~
100
0.10
0.25
0.30
0.15
0.30
0.45
0.45
0.35
U
III
~
~
Q,l
20
0.90
1.00
1.05
0.95
1.05
1.15
1.20
1.10
"tl
~
l=: OM
40
0.60
0.70
0.75
0.65
0.75
0.85
0.90
0.80
Q,l"tl
..-I
Q,l
60
0.25
0.45
0.50
0.30
0.40
0.60
0.65
0.45
>.~
Q,lU
80
0.15
0.30
0.40
0.20
0.30
0.50
0.55
0.35
l=:
0
100
0.10
0.25
0.30
0.15
0.25
0.40
0.40
0.30
lE:
o~ *
20
0.80
1.00
0.90
1.00
1.15
1.05
-e
Q,l
40
0.55
0.70
0.60
0.70
0.85
0.75
~
U
60
0.20
0.45
0.25
0.35
0.60
0.40
<
80
0.10
0.35
0.15
0.25
0.50
0.30
f@
100
0.05
0.25
0.10
0.20
0.35
0.25
~
*Credit supply at the BNDA is specified only for seasons I, 3 and 4, in
each of the 2 years.
Consequently activities for credit reservation at
that source are defined only for these 6 seasons in the planning period.
This is shown in the table by the unspecified reservation prices for
BNDA credit in seasons 2 an 6.

84
of liquidity owing to its lengthy lending procedure which either in-
creases the cost of borrowing or inhibits the timing of its loans, and
the restrictions on the use of the loan proceeds.
The seasonal variation of the liquidity values is determined by the
survey and reflects the risk faced by the mode1ed farm-borrower in each
seaso~
One should recall that the choice of these values is critical to
the allocation process.
Indeed,
reservation values in the case of credit
represent opportunity costs subject to which credit supply is allocated
to credit account whence it is available to support borrow ing by the
borrower.
Also, the reservation va1 ues of cash represent opportunity
costs of committing cash to use (that is, allocating cash to the cash
account).
Hence, the value of cash in use must exceed the value of cash
in reserve.
Otherwise, the cash will, in optimum, be held in reserve.
Each CFAF of cash or credit valued draws on the cash or credit
reserve and contributes to the satisfaction of 1 CFAF of liquidity re-
quirement.
This is shown by the positive coefficient 1 of the valuation
activities in the corresponding reserve accounts,
and liquidity require-
ment rows.
It should be recalled that the reservation values are non1inear
convex functions of the reservation levels.
But in the model,
these
functions are approximated by linear segments whose abscissa are the
val.ua t i.on activities.
The convexity of these functions thus guarantees
that the optimal amount of cash or credit reserved is a linear combina-
tion of at most two of the specified valuation activities.
The accuracy
of this approximation is controlled by the number of segments used in the
sense that the more segments there are, the closer the approximatio~

87
repayment activities as there are seasons in the debt-period (8 to 10
months).
For instance. season 1 debt. which matures at the beginning of
season 4. would be serviced through (1) repayment at the beginning of
season 2. (2) repayment at the beginning of season 3. and (3) repayment
at the beginning of season 4.
The coefficients of these activities in
the cash account rows represent the amount of cash used for the repayment
of 1 CFAF of loan principal plus the interest charge then due.
Each CFAF
of loan amount repaid reduces by 1 CFAF the debt account.
contributes to
satisfy 1 CFAF of liquidity requirement to be met in that season. and
restores 1 CFAF of credit amount absorbed through borrowing.
BNDA Loan-Repayment Activities.
The BNDA loan-repayment activities
are basically structured as the moneylender and the bank loan-repayment
activities:
they reduce the debt account. contribute to satisfy liquidi-
ty requirement and restore credit.
But because BNDA loans may be repaid
after the due date. an additional repayment activity is specified corres-
ponding to each loan.
This activity expresses loan delinquency which
represents the amount of loan repaid within the season following the due
date. Because delinquency bears a penalty cost expressed as an additional
interest charge. the coefficient of this activity in the cash row corres-
ponds to the amount of cash used to repay 1 CFAF of loan principal plus
the interest charge and the penalty cost.
Like the other repayment
activit~es. it contributes to satisfy liquidity requirement.
BNDA Loan-Default Activities.
In order to account for partial
default.
these activities have been defined in percentages of loan.
The
percentages considered are 20. 40. 60.
BO and 100(%) representing each a
default activity.
In the model. a loan is considered defaulted if it has

88
not been repaid within the season following the due date.
It was assumed
that the BNDA responds to default by reducing its credit supply for
future borrowing.
This is reflected in the coefficients of default
activities in the BNDA credit supply rows in subsequent seasons.
Each
one of these coefficients expresses a marginal default-penalty cost whose
value is assumed to be at least equal to 1 CFAF specifying that each CFAF
of loan amount defaulted reduces subsequent credit supplies by at least 1
CFAF.
The appropriate value of this coefficient then is found by varying
the specified level until a default rate comparable to the observed value
is foun~
The coefficients -1 of default activities in the liquidity re-
quirement rows reflect how risky default is for the farm-firm.
Here the
riskiness of the activity is reflected in that default activities corres-
ponding to a given loan appear in several liquidity requirement rows.
For example.
to activate one of the default activities corresponding to
BNDA-debt incurred in season 1 would add to the liquidity requirement to
be met in seasons 1. 2. 3. and 4.
Because the maj ori ty of BNDA loans
given in the second year mature after the planning period and because
default is accounted for only after the due date. default activities for
these loans were not specifie~
Finally. it should be noted that. as in
the case of valuation activities. default activities correspond to ab-
scissa of line-segments used to approximate the default-penalty cost and
the def'ault-liquidity-requirement functions.
The convexity of these
functions thus guarantees that the optimal default level is a linear
combination of at most two of the default activities specified for a
given loan.

89
NOTES
1. The coefficients am+f( L+l)+l r'
as noted in Chapter Three.
represent
the modification in the liquidity requirements due to the activities
xr's.
In an empirical analysis.
they could be easily associated with
the variances of the activity levels and estimated as such. as sug-
gested by Baker [5].
But in previous work on liquidity mangement as
well as in the current study. because of lack of adequate information.
they are generated with the model based on conformance comparisons.
That is. initial values are specified for these coefficients in the
first run of a properly formulated modeli these values are then varied
in successive runs until a production level comparable to the observed
output is found.
Here we base our choice of initial values on infor-
mation provided on the rate of spoilation in crop stockpiles at har-
vest and at the various stages of the marketing process.
2. The default variables also have coefficients in the BNDA credit supply
rows for the second year.
But in order to keep the notations simple
we have not shown this relation in the mathematical formulation of our
model.
It is. however. specified in the Appendix Table Al.

90
CHAPTER FIVE
MODEL RESULTS AND VALIDATION
5.1
Model Results
All the numerical experiments in this study were carried out on the
Cyber 175 using the APEX III mathematical programming package;
all the
computations were made in single precision with the machine-epsilon of
approximately 10-14•
Most of the packages available for solving mathe-
matical programming problems provide options for scaling the coefficients
of the matrix if they differ significantly in magnitude.
In our case,
the magnitude of the coefficients of the LP matrix varies from 10 6 to
10-3•
But because the current version of the APEX III package does not
provide an option for scaling the matrix,
we had to scale it separately
prior to solving the problem.
Here, the Harwe11 subroutine MC19A (single
precision version)
[24]
was used.
5.1.1
Basic Optimal Plan
The optimal solution may be described as follows.
As in the case of
the original LSLP model,
the default-augmented LSLP model gives an opti-
mum of areas planted to various crop, here cassava,
rice,
maize and yam.
At the same time it generates jointly,
(i) an optimum among the alterna-
tives for meeting household requirements and produces (ii) an optimal
level of borrowing at each of the three sources of credit, and (iii) an
optimal distribution of liquidity among the different assets specified
for reserves.
Finally, it generates an optimal level of delinquent loan
and an optimum of amount of loan defaulted.
Table 5.1 summarizes some of

91
Table 5.1
Model Output and Observed Values
Activitya
Model
Sample Observation
Output
Mean
Standard Deviation
Value of Objective Func-
tion
7.275.644
n.a.
n.a.
Reserved Cash in Year
one
1. 081. 817
n.a.
n.a.
two
1.678.068
n.a.
n.a.
Reserved Moneylender
Credit in Year one
80.941
n.a.
n.a.
two
50.548
n.a.
n.a.
Reserved Commercial Bank
Credit in Year one
77 .552
n.a.
n.a.
two
57.647
n.a.
n.a.
Net Cash Flow
2.809.513
2.072.560
6.941.040
BNDA Debt in Year one
523.853
562.000
309.765
two
616.210
n.a.
n.a.
Moneylender Debt in
Year one
34.637
96.400
265.240
two
97.389
n.a.
n.a.
Commercial Bank Debt in
Year one
114.367
138.600
167.500
two
134.643
n.a.
n.a.
Unpaid BNDA Debt in Year
one
175.680
173.152
1.435.009
two
n.a.
n.a.
Unpaid Moneylender Debt
in Year one
0
0
two
0
0
Unpaid Commercial Bank
Debt in Year one
0
0
two
0
0
Area Cu1tivatedb on
Cassava in Year one
0.70
0.51
1.12
two
0.93
n , a.
n.a.

92
Table 5.1 (continued)
Activitya
Model
Sample Observation
Output
Mean
Standard Deviation
Maize
in Year one
0.76
0.76
1.17
two
0.76
n.a.
n.a.
Rice
in Year one
0.92
0.92
3.60
two
0.92
n.a.
n.a.
Yam
in Year one
1.01
1.20
1.16
two
0.78
n.a.
n.a.
MarketingC
Cassava
in Year one
0
0
0
two
n , a.
n.a.
n , a.
Maize
in Year one
675.64
306
949.85
two
659.56
n, a.
n.a.
Rice
in Year one
340.11
718
3.333
two
707.76
n.a.
n.a.
Yam
in Year one
1.061. 25
970.00
3.012.58
two
1.473.01
n.a.
n.a.
aThe unit is CFAF unless otherwise specified
bThe unit is hectare
cThe unit is kilogram

93
the organizational optima produced by the model.
Table 5.2. which is a
detailed section of Table 5.1.
gives
the variables describing the
farmer's borrowing and repayment behaviors.
This table shows that at
optimum the farmer borrows from the BNDA in seasons 1. 3. 4. 5. 7 and 8.
and from the moneylender or the commercial bank in each season except the
first.
In conformance with the specifications of the model presented in
the preceding chapter.
delinquency and default occur only for BNDA loans.
Although defau1~ variables have been specified for all BNDA loans made in
the first half of the planning period. default actually occurs only on
seasons 3 and 4 debts:
the farmer fully repays season 1 debt.
This
might be due to the fact that. in the model. it costs more to default on
season 1 debt than on other season debts.
Indeed.
as mentioned earlier.
defau1 t on the first season debt reduces credit supply in seasons 5. 7
and 8: but defau1 t on season 3 debt reduces credit supply in seasons 7
and 8. and default on season 4 debt reduces credit supply in season 8.
Despite the additional interest charge imposed on past due loans. the
farmer chooses to be delinquent on season 4 debts.
A possible explana-
tion is that the effect of the penalty has been crowded out by the fact
that the farmer disposes of a relatively large amount of cash resulting
from the sale of his crop in the eighth season where delinquency occurs.
The model does not allow the farmer to default on moneylender or commer-
cial bank loans.
He also cannot default on BNDA loans given in the
second year of the planning period.
This is shown in the last column of
Table 5.2 by the dash (--).
Hence. while the farmer repays the full
amounts of loan obtained either from BNDA in the second year. or from the

Table 5.2
Borrowing and Repayment Behavior of the Farmer
Source
Seasonsa--Amoun t
Amount Repaido Within
Total
Amount
of
Borrowedd
Amount
Unpaidd
Credit
2
4
6
8
10
12b
Repaidd
months
months
months
months
months
months
1
179,706
179,706
179,706
0
2
--c
3
141,141
73,544
73,544
67,597
tJj
4
203,006
94,922
94,922 108,083
z
5
156,900
156,900
156,900
t:l
~
6
7
282,422
282,422
282,422
8
176,889
176,889
176,889
1
0
0
0
2
14,394
14,394
14,394
~
0
3
3,850
3,850
3,850
='ID
4
16,394
16,394
16,394
'<
I-'
5
7,116
7,116
7,116
ID
='
o,
6
26,064
26,064
26,064
ID
7
23,545
23,545
23,545
11
8
40,663
40,663
40,663
1
0
0
0
C1
2
10,510
10,510
10,510
m
3
23,270
23,270
23,270
ID
11
4
80,586
80,586
80,586
n
....
5
6,411
6,411
6,411
~
6
34,438
34,438
34,438
tJj
7
34,934
34,934
34,934
III
~
8
58,860
58,860
58,860
8seasons 1, 2, 5 and 6 each is 2 month-long
seasons 3, 4, 7 and 8 each is 4 month-long
brepresents the amount of loan in delinquency
\\0
.t-
ccorresponds to unspecified variables in the model
dt he unit is CFAF

95
moneylender or the commercial bank. he only repays 66% of BNDA loans
obtained in the first year.
In the solution.
when the farmer chooses to repay a loan within the
normal term of the debt. he does not wait till the due date:
he repays
either at the end of the season where the debt was contracted or at the
end of the season that immediately follows. as shown in Table 5.2 for all
of the three sources of credit.
This may be due to the fact that the
farmer saves o~ interest payments when he repays promptly.
Also. because
repayment of a debt restores credit supply in subsequent seasons. the
sooner the farmer repays the more he may borrow in the following seasons.
This repayment behavior agrees with reality and hence may be viewed as an
indication that the model performs well.
But a formal way of testing the
robustness of a model is through a validation process and this is what we
consider in the next section.
5.1.2
Marginal Values
The marginal values or shadow prices represent the effects on the
size of the objective function of one-unit changes in the right-hand side
quantities.
The marginal values also correspond to the costs of the
opportunities lost for making more profit.
They are therefore often
referred to as "opportunity costs".
and they resu1 t
from a careful
weighing up of the demands which each activity makes on the scarce re-
sources and contributes in profit in return [60].
In this study. we will
evaluate the effects of one-unit changes in the right-hand side values of
constraints such as land. 1abor. cash, credit and liquidity requirements.
The marginal values of these resources are displayed in Table 5.3.
Here,

96
Table 5.3
Marginal Value of Extra Resources or Requirements
Resource or Requirement
Marginal Value
Description
Level*
Unit
Land available for Root Crops in
Year one
+
1.67
CFAF
Year two
+103.77
CFAF
Land available for Grain in
Year one
+137.67
CFAF
Year two
+109.58
CFAF
Family Labor in
Year one. in
season one
- 49.48
CFAF
season two
+ 37.84
CFAF
season three
+ 50.52
CFAF
season four
- 79.35
CFAF
Year two. in
season one
- 39.46
CFAF
season two
+ 29.45
CFAF
season three
+ 34.78
CFAF
season four
39.07
CFAF
Cash Supply in
Year one. in
season one
+ 30.14
CFAF
season two
+ 24.46
CFAF
season three
+ 35.74
CFAF
season four
+ 41.26
CFAF
Year two. in
season one
+ 27.25
CFAF
season two
+ 22.74
CFAF
season three
+ 40.38
CFAF
season four
+ 43.78
CFAF
Cash Accounting in
Year one. in
season one
-
21.76
CFAF
season two
20.89
CFAF
season three
- 27.39
CFAF
season four
24.95
CFAF
Year two. in
season one
- 19.75
CFAF
season two
- 18.93
CFAF

97
Table 5.3 (continued)
Resource or Requirement
Marginal Value
Description
Level*
Unit
Cash Accounting in
Year two. in
season three
- 24.63
CFAF
season four
17.15
CFAF
Credit Supplied at BNDA in
Year one. in
-
season one
+ 0.01
CFAF
season two
season three
+ 0.14
CFAF
season four
+ 0.11
CFAF
Year two. in
season one
+ 0.16
CFAF
season two
season three
+ 0.36
CFAF
season four
+ 0.60
CFAF
Credit Supplied at Moneylender in
Year one. in
season one
+ 0.62
CFAF
season two
+ 0.38
CFAF
season three
+ 0.40
CFAF
season four
+ 0.45
CFAF
Year two. in
season one
+ 0.93
CFAF
season two
+ 1.31
CFAF
season three
+ 1.46
CFAF
season four
+ 1.50
CFAF
Credit Supplied at Commercial Bank in
Year one. in
season one
+ 0.71
CFAF
season two
+ 0.65
CFAF
season three
+ 0.64
CFAF
season four
+ 0.73
CFAF
Year two. in
season one
+ 1.10
CFAF
season two
+ 1.66
CFAF
season three
+ 1.84
CFAF
season four
+ 1.98
CFAF

98
Table 5.3 (continued)
Resource or Requirement
Marginal Value
Description
Level*
Unit
Moneylender Credit accounting in
Year one. in
season one
+ 0.02
CFAF
season two
- 0.02
CFAF
season three
- 0.01
CFAF
season four
+ 0.06
CFAF
Year two. in
season one
+ 0.09
CFAF
season two
+ 0.31
CFAF
season three
+ 0.26
CFAF
season four
+ 0.29
CFAF
Commercial Bank Credit Accounting in
Year one. in
season one
+ 0.02
CFAF
season two
+ 0.10
CFAF
season three
+ 0.11
CFAF
season four
+ 0.18
CFAF
Year two. in
season one
+ 0.17
CFAF
season two
+ 0.45
CFAF
season three
+ 0.40
CFAF
season four
+ 0.44
CFAF
Minimum Required Liquidity in
Year one. in
season one
- 0.66
CFAF
season two
o
CFAF
season three
o
CFAF
season four
o
CFAF
Year two. in
season one
- 1.34
CFAF
season two
o
CFAF
season three
- 1.21
CFAF
season four
o
CFAF
*The objective function value gets better if the marginal value is posi-
tive and gets worse if the marginal value is negative.

99
the objective function value gets better if the marginal value is posi-
tive and gets worse if the marginal value is negative.
Because the
objective function includes liquidity values. one would expect these
marginal values to be higher than they would have been if the objective
only included net cash flow.
Land.
As shown earlier. land available for food crop production is
organized into two "blocks" in the model:
Land A and Land B.
Land A is
used to grow root crops represented by yam or cassava in the model. and
Land B to grow grains. here maize and rice.
The resu1 ts of the current
analysis indicate that the marginal value of land available for root
crops significantly increases from 1.67 CFAF in the first year to 103.77
CFAF in the second year.
In the case of grains.
the marginal value
decreases over the planning period. but remains above 100 CFAF.
This
discrepancy in the values of the two types of land may be due to the
difference between the price the farmer receives for the crop he produces
(regarded as an unfinished good) and the price he would pay for the same
crop on the market (considered this time as a finished good).
For exam-
ple. the farmer is paid 60 CFAF per kilogram of rice produced. but must
pay 100 CFAF per kilogram of rice purchased.
He also sells maize for 25
CFAF per kilogram. but purchases it at 70 CFAF per kilogram.
In the case
of yam. however. he receives 50 CFAF and pays 75 CFAF per kilogram.
This
represents an average price-difference of -43 CFAF for grains and of -25
CFAF for root crop.
thus requiring the farmer to produce more grain for
on-farm consumption.
We would therefore expect land available for
growing grains to marginally contribute more to the obj ective function
than land for root crops. if additional units could be acquired.
As

100
noted in Chapter Three. land for agricultural production is acquired
through inheritance. hence priced in non-monetary terms.
But if land
were to be purchased.
its unit cost presumably would be higher than the
extra objective function value indicated by its marginal values.
Hence.
it would not be desirable to expand land.
Also. the land tenure pattern
in the region. although it is of a "free hold" farm.
has a built-in
mechanism to limit the amount of land that each household could acquire.
But if the specjfication of land into two blocks is realistic.
then the
farmer must proceed into a reorganization of the land available to him.
As indicated by the marginal values. the farmer will be better off if he
expands land for growing grains by reducing the amount of land available
for root crops.
This provides for an efficient way of increasing the
value of the objective function without facing the costs associated with
acquiring land from external sources.
Lab or,
Each extra unit of labor increases the obj ec tive func tion
value in the second and third season of each year but decreases it in the
first and fourth season.
This is due to the fact that the labor con-
straint is expressed as a minimum Q) in seasons 1 and 4 of each year.
Overall.
family labor appears to be a limiting factor of production for
the modelled farm (except in seasons 1 and 4 of each year) in the sense
that the labor constraints which are specified as maximums. all are
binding.
this tends to confirm the prediction made by Binet [16] that
on-farm labor supply would have to be supplemented by off-farm supplies
by 1985.
Using agricul tural data for 1975 and the maj or indicators of
population growth in the region the author predicts the shortage of on-
farm labor supply in 1985.
To this end she uses the fact that in 1975

101
the difference between the supply of and demand for labor was negligible
while the growth rate in the demand for labor and the population was
3.40% and 0%. respectively (see Table 3.3).
The farmer could benefit from an abundant off-farm labor supply from
other regions in the country. especially from the North.
due to the time
lag in the farming operation among these regions; but he must acquire the
additional labor at a cost given by the prevailing agricultural wage
rate.
In the basic model. the unit cost of labor is estimated at 449.50
CFAF and 409.50 CFAF for the first and the second year. respectively.
But each extra unit of labor on the average contributes 44 CFAF and 32
CFAF to the obj ec tive func tion val ue in the first and second year. re-
spectively.
as indicated by the marginal values.
Hence.
it seems unde-
sirable to increase labor supplies in the planning period.
Cas~
The marginal value of cash supply varies with the seasons and
years:
yearwise it is higher in the first year than in the second. but
seasonwise it is higher in the fourth season of each of the tw 0 years.
This may be due to the fact that demand for and supply of cash differ in
the seasons.
Recall that each unit of cash supplied is allocated between
cash accounts whence used in the farm. and cash reserves where it is
val ued to satisfy the farmer's liq uidi ty requirements.
The trade-off
between the reward from using cash in the farm or holding it in reserve
hence determines the proportions in the account and in the reserv~
The
sum of these two uses of cash should determine the value of cash supply
in each season.
From Table 4.5. it appears that the reservation prices
of cash are the same in corresponding seasons of the two-year planning
period although they vary between seasons.
Also the demand for cash in

102
the farm is maintained constant over the two years as shown in the
Appendix Table Al.
Hence.
the fact that an extra unit of cash supply in
the first year contributes more to the obj ective function value may be
due to the fact that cash supply opportunities are greater in the second
year.
In the model. the right-hand side of the cash account constraints
specify levels of demands for cash.
They increase with the production
and consumpti~n activities; but they decrease when the amount of cash
allocated to the cash accounts increases.
Increasing these levels would
require an increase in the amount of cash allocated to the cash account.
Consequently. the value of the obj ective function. which includes the
liquidity value of cash in reserve. would decrease for each additional
unit of demand for cash as shown by the marginal values.
This results
from the fact that reserved cash. like credit in reserve. has a value
which contributes to the maximum in the objective functio~
Credit.
Credit is another limiting factor in the model.
Recalling
that the farmer may borrow from three sources: the BNDA. the moneylenders
and the commercial banks.
the BNDA is used only for borrowing purposes.
while credit at the moneylender and the commercial bank is allocated
between borrowing and reservation.
In all the three cases. each extra
uni t of credit supplied increases the value of the obj ec tive function;
but the 'marginal value of credit remains low at the BNDA.
This may be
due to the fact that this source only provides credit for borrowing
purposes.
However. comparing with the marginal val ues of the credit
accounts for moneylender or commercial bank which define the use of these
credits for borrowing. credit supplied at the BNDA appears to be as valu-

103
able as the commercial bank credit.
Indeed. each additional unit of
credit supplied for borrowing at the moneylender either reduces the
objective function value (in seasons 2 and 3 of the first year) or
remains below the marginal value of either the BNDA or the commercial
bank credit. when it increases the value of the objective functio~
As mentioned earlier the interest charge on each unit of credit
supply used for borrowing is 0.09 CFAF. 0.50 CFAF and 0.17 CFAF for BNDA.
moneylender ana commercial bank loans.
respectively.
Hence.
given that
on the average each additional unit of credit supplied for borrowing at
each source contributes 0.23 CFAF. 0.12 CFAF and 0.23 CFAF to the objec-
tive function. respectively. it seems desirable to increase the borrowing
funds only at the BNDA and the commercial bank.
The current results shed additional light on the actual interaction
among the three sources of credit specified in the model.
In this study.
credit is estimated based on the assumption that the ratio of amount
borrowed to credit supply is constant for the three sources.
Whenever a
lender deals with a customer he takes into account the amount borrowed
from other sources.
In general he reduces the credit limit by a certain
amount.
But in our study we assume that this amount is negligible.
This
is mainly due to lack of observation on credit limit at the BNDA and the
commercial bank. and also lack of information on how much each lender
reduces its credit supply.
With these assumptions an undisciplined
farmer runs the risk of being overwhelmed with debt since he can borrow
freely from the three sources.
Fortunately the big difference in the
interest rates charged by these lenders (9% for the BNDA. 16.5% for the
commercial bank and 50% for the moneylender) tends to prevent this from

104
happening.
Also. the fact that the model specifies that any amount of
loan obtained from either the moneylender or the commercial bank must be
fully repaid by the due date may further limits the amount borrowed from
these two sources.
This seems to be confirmed by the information pro-
vided by the marginal values.
Liquidity Requirements.
The right-hand side values of the liquidity
requirement constraints represent the response of the farmer to risk from
sources which ~re external to the farming operations.
These values are
subject to increase due to risky activities and decrease due to cash and
credit reserved.
Because these are minimum levels to be satisfied.
increasing them should at least reduce the value of
the objective
function if the constraints are binding.
The constraints are binding
only in the case of the first season in the first year. and the first and
third seasons in the second year.
This indicates that liquidity goals.
as specified by the liquidity requirement constraints. are fulfilled only
in three of the eight seasons specified in the model.
In these seasons.
each additional unit of required liquidity reduces the objective function
value by 0.66 CFAF.
1.34 CFAF and 1.21 CFAF.
respectively.
5.1.3
Effects of Non-Optimal Variables
The reduced costs.
which are provided as subsidiary information to
the basic optimal solution. are essential in ana1yzing the results of the
model.
They are nonzero for nonbasic activities and represent the amount
by which the cost imputed to an activity from using the resources exceeds
the value-contribution of that activity to the objective functio~
These
costs are of special interest in deciding whether or not to "force" an

105
activity into an optimal solution.
Because the solution is optimal
forcing uni ts of nonbasic ac tivi ties in it would reduce the obj ec tive
function value.
as indicated by the unit costs of these activities.
In the current analysis. we will use this additional information to
determine the effects of forcing in the optimal solution activities for
repaying BNDA loans given in the third and fourth seasons of the first
year.
the seasons of default and delinquency.
These are activities that
correspond to Fepaying the loan at or after the due date in the case of
season 3 loan. and before or at the due date in the case of season 4
loan.
In the current analysis this choice is based on the information
provided in Table 5.2.
This table shows that the farmer defaults on 48%
and 53% on his loans in seasons three and four. respectively.
In season
four he is delinquent on 47% of his 10a~
Table 5.4 summarizes the conditions for introducing the repayment
activities in the optimal so.l.ut Lon,
It gives the levels at which these
activities enter the basis. and shows the corresponding variable that
leaves the basis.
In the basic model.
none of these activities appears
in the objective function.
If these activities were to appear in the
objective function with coefficients shown in column 2 of Table 5.4. they
will then enter the optimal solution at levels shown in column 3 of the
same table.
This will in turn force a corresponding number of other
v a r Lab l.es
out
of
the
basis
(column four
in Table
5.4).
Variable
(RBND1422).
which represents the activity for repaying season 4 loan at
its due date. requires that its coefficient in the objective function
increases to 0.99 CFAF.
It also displays a high activity level. 94.797
CFAF. forcing delinquent repayment. variable (RBND1424) to leave the

106
Table 5.4
Conditions for Repaying BNDA Loans. Activity Level and Varia-
bles Leaving the Optimal Solution
Repayment Activity
Variable
Obj ective
Leaving the
Function
Description
Liquidity Valuec
Leveld
Optimal
Va1ueg
Required for the
Solution
Activity to Enter
the Basis
(1)
RBND1321 a
0.27
5953
LQRR14e
7.274.013
(2)
RBND1322 a
0.52
16485
LQRRl4e
7.267.057
(1)
RBND1421b
0.13
18380
LQRR14e
7.273.134
(2)
(4)
RBND1422b
0.99
94797
RBND1424b
7.181.585
(3)
LQRR22 f
RBND1423 b
0.31
41553
7.262.677
aActivities for repaying Year l-Season 3-BNDA loan in
(1) Year 2-Season 1
(2) Year 2-Season 2
bActivities for repaying Year l-Season 4-BNDA loan in
(1) Year 2-Season 1
(2) Year 2-Season 2
(3) Year 2-Season 3
(4) Year 2-Season 4
cThe unit is CFAF per CFAF of loan amount repaid.
These values also
express the objective function value penalties
dThe unit is CFAF
eConstraint specifying Liquidity Reserve Requirement in Year l-Season 4
fConstraint specifying Liquidity Reserve Requirement in Year 2-Season 2
gThese values correspond to the level of the objective function if the
activities are forced into the optimal solution.
They all are lower
than the initial objective function value which is 7.275.644 CFAF

107
basis.
In the case of season 3 loan,
the delinquent repayment variable
(RBND1322) and the repayment at the due date variable (RBND1321) will
enter the basis if they were to appear in the obj ec tive func tion with
coefficients 0.52 CFAF and 0.27 CFAF,
respectively;
the former activity
will enter at a higher level.
Hence, by forcing some activities in the
optimal solution we will lower the value of the objective function;
but
this would allow us to reduce the amount of loan defaulted in season 3
and eliminate delinquency in season 4.
5.2
Validation of the Model
In theory a model would be valid if the following condi tions are
met:
(1) it reflects a logic that conforms with accepted theory of the
system,
(2) its empirical components (data) are in accord with counter-
part observation of the system,
(3) its output conforms with the system
observed values, and (4) it can be used for its intended purpose.
But
in practice it would be too costly to go through all these steps in order
to check if a model is valid.
In the current study we will only consider
(i) checking the correctness of the mathematical formulation of the
model, (ii) analyzing the tolerance of the model to variations in the
data, and (iii) checking how closely the model output conforms with
observations.
5.2.1
Verification of the Mathematical Formulation of the Model
There are no formal procedures for checking the correctness of a
mathematical formulation of a model.
Here we would consider a screening
exercise, checking if the signs and coefficients are properly chosen, and

108
if the constraints are correctly specified.
In general. a properly
formulated model would yield a feasible optimal solutio~
5.2.2
Sensitivity Analysis
The reliability of the model output could be tested further by
ana1yzing how sensitive the model is to small changes in the right-hand
side quantities.
This analysis is done by establishing intervals where
these coefficients may vary without changing the optimal vector. although
some of its entries and consequently the objective function.
may change
values.
The width of these intervals would serve as an indication on how
sensitive the solution is to errors in the right hand coefficients.
Elements for the sensi tivi ty analysis of our basic model are sum-
marized in Table 5.5.
Note that only a few constraints are considered.
They include those constraints which might affect policy decisions.
Recalling that data on the right-hand side of the constraints were ga-
thered in a survey of those farmers who borrow from the BNDA under the
PFVN credit program.
their precision may be questionable.
Indeed. those
questioned in the surveys could either overestimate or underestimate the
values of some important variables for several reasons such as the lack
of adequate means to keep these values or the suspicion that they might
have of the intended use of the survey results.
For example. in the
current case of loan default where the farmer might be penalized if he
had wilfully failed to repay his debt. he might be tempted to under-
estimate the values of the variables related to his repayment ability in
order to lessen the penalty charges.
Inaccuracy in the data may also be

109
Table 5.5
Right-Hand Side Ranges for Selected Constraints in the Basic
Model
Constraint
Description
Right-Hand Side
Right-Hand Side Range
Accuracy in
Value
Lower
Upper
Right-Hand
Side Value*
Land available for
Root Crops in
Year one
1.71
1.69
2.12 +
0.02 hectare
Year two
1. 71
1.57
1.74 +
0.03 hectare
Land available for
Grain in
Year one
1.68
1.61
2.60 +
0.07 hectare
-
Year two
1.68
1.40
5.66 +
0.28 hectare
Family Labor in
Year one. in
season one
97.76
21.15
676.05 +
76.61 mandays
-
season two
56.23
0
61.44 +
5.21 mandays
season three
72.29
0
73.11 +
0.82 mandays
season four
252.91
213.75
548.48 +
39.16 mandays
Year two. in
season one
106.03
77 .36
310.19 +
28.67 mandays
season two
53.20
0
57.78 +
4.58 mandays
season three
64.86
0
70.66 +
5.80 mandays
season four
271.50
0
540.97 :; 269.47 mandays
Cash Supply in
Year one. in
season one
1.439.558
1.179.497 1.474.009 +
34.451 CFAF
-
season two
0
0
34.455
0
season three
0
0
27.534
0
season four
1.293.800
1.127.558 1.315.821 +
22.021 CFAF
Year two. in
season one
0
0
17.604
0
season two
0
0
11.732
0
season three
0
0
7.034
0
season four
1.293.800
1.109.850 1.823.414 + 183.950 CFAF
Cash Account in
Year one. in
season one
98.877
64.442
358.819 +
34.435 CFAF
season two
109.877
82.326
317 .854 :;
27.551 CFAF
season three
216.102
194.083
382.327 :;
22.019 CFAF
season four
230.420
212.820
363.283 :;
17.600 CFAF

110
Table 5.5 (continued)
Constraint
Description
Right-Hand Side
Right-Hand Side Range
Accuracy in
Value
Lower
Upper
Right-Hand
Side Value*
Cash Account in
Year two. in
season one
98.877
87.138
182.481 .:.
11.738 CFAF
season two
109.877
102.839
193.534 .:.
7.038 CFAF
season thre~
216.102
211.080
263.454 .:.
5.021 CFAF
season four
230.420
18.371
340.768 + 110.348 CFAF
Credit Supplied at
BNDA in
Year one. in
season one
282.294
248.406
539.021 +
33.888 CFAF
season two
season three
179.714
146.451
475.880 +
33.262 CFAF
-
season four
333.186
116.707
358.151 +
24.965 CFAF
Year two. in
season one
313.939
294.976
353.457 +
18.963 CFAF
-
season two
season three
199.860
126.414
211.659 +
11.797 CFAF
-
season four
370.537
193.663
905.502 + 176.874 CFAF
Credit Supplied at
Moneylender in
Year one. in
season one
17.050
1.503
51.479 +
15.547 CFAF
-
season two
25.575
1.588
672.839 +
23 .987 CFAF
-
season three
42.625
4.124
719.845 +
38.501 CFAF
-
season four
51.150
31.083
204.209 +
20.067 CFAF
Year two. in
season one
17.432
0
26.690 +
9.258 CFAF
-
season two
26.148
8.897
84.765 +
17.251 CFAF
-
season three
43.581
4.390
57.289 +
13.708 CFAF
-
season four
52.297
1.478
727.709 +
50.819 CFAF
Credit Supplied at
Commercial Bank in
Year one. in
season one
24.514
1.487
58.942 +
23 .027 CFAF
-
season two
36.771
1.735
113.710 +
35.036 CFAF
-
season three
61.284
34.445
99.331 +
26.839 CFAF
-
season four
73.541
61.143
80.980 +
7.439 CFAF

111
Table 5.5 (continued)
Constraint
Description
Right-Hand Side
Right-Hand Side Range
Accuracy in
Value
Lower
Upper
Right-Hand
Side Value*
Credit Supplied at
Commercial Bank in
Year two. in
season one
25 .063
5.419
196.439 +
19.644 CFAF
-
season two
37.595
17.344
99.262 +
20.251 CFAF
season three
62.658
4.518
76.318 +
13.660 CFAF
-
season four
75.190
1.626
747.222 +
73 .564 CFAF
*The values correspond to the smallest deviation of the limits from the
right-hand side coefficient currently specified.

112
caused by other Common practices such as rounding values to their nearest
digit. as it is often the case for variables such as land supplies or
areas planted in crops.
To evaluate the effects of these inaccuracies on
the model.
we will limit ourselves to intervals with small width since
they indicate that small changes in the corresponding data might signifi-
cantly affect the model output.
These include land. family 1abor and
cash supply constraints.
Land.
In J:he survey. land area. which is measured in hectare. was
given with one decimal.
It would not be unreasonable to assume that in
general farmers would tend to round the area of their land to the nearest
quarter (0.25) hectare.
Thus errors in these figures will always lie in
the interval [-0.25. +0.25].
The sensitivity analysis shows that the area of land used for root
crops may vary between 1.69 and 2.12 hectares in the first year.
and 1.57
and 1.74 hectares in the second year. without changing the variables
currently in the optimal solution.
In the model. the area used in both
cases was 1.71 hectares.
This value was obtained as the mean of 50
observations.
Using our assumption on the magnitude of the error. the
actual value will lie in the interval [1.46. 1.96J which can be rewritten
as
[1.5.
2.0J
after rounding the values of the limits to the first
nearest digit after the decimal point.
Similarly the interval obtained
from the model could be rewritten as [1.5. 2.0J which therefore would
contain the true right hand value of land constraint in the case of land
available for root crop.
This analysis also applies for the case of land available for
grains.
Here.
the true right-hand side coefficient will be contained in

113
the interval [1.43. 1.93]
which will be rewritten as [1.5. 2.0] after
rounding the values of the limits to the nearest digit.
The interval of
variation in the model is
[1.5.
2.5]
and
[1.4. 5.5]
in the first and
second year.
respectively.
In both years the observed interval is con-
tained in the interval obtained from the model.
Hence.
the true right-
hand side coefficient of land constraint would lie in the range given by
the model. in the case of land available for grain.
Family Labor.
The range for family labor constraints appears to be
a problem only in the third season of the first year. where the margin of
error is :!:. 0.82 manday.
But given that the supply of labor is measured
in days of effective work.
we would expect the survey results to differ
from the true value in terms of hours of work in a day.
Hence.
the
interval [-0.82. +0.82]
is large enough to contain the actual error.
Cash Supply.
In the model. the right-hand side coefficients for
cash supply constraints are nonzero only in the first and fourth seasons
of the first year. and the fourth season of the second year.
The margin
of error for these coefficients are + 34.451 CFAF. :!:. 22.021 CFAF. and :!:.
183.950 CFAF. respectively.
Given that the right-hand side values are in
the millions of CFAF. one could consider these margins small enoug~
But
it should be noted that in monetary terms. data are often recorded in the
thousands of CFAF in order to facilitate accounting.
We would then
expect the current value of cash supply to deviate from the true value
only in the order of the hundreds of CFAF.
Hence. a margin of :!:. 22.021
CFAF would contain the true error.
We would conclude here that our basic model is not too sensitive to
the right-hand side data and thus we could rely on its solutio~

114
5.2.3
Evaluating the Conformity of the Model Output With the
Observed Values
In this section we will check how the model output conforms with the
observed values.
Defau1 ted Amount of Loan.
The model allows the farmer to defau1 t
only on BNDA loans given in the first year.
In the optimal solution. the
farmer borrows 523.823 CFAF from the BNDA in the first year.
on which he
defaults 175.680 CFAF.
This amount is 1.46% higher than 173.152 CFAF
observed.
Here.
the difference between the model output and the system obser-
vation could be explained by the fact that the coefficients of defau1 t
variables in the credit supply rows and liquidity requirement constraints
are only approximated.
These approximated values fall short of the
observed values. hence do not lead to the exact amount of loan defaulted.
173.152 CFAF.
But given the magnitude of the default variable which is
of order 10 5• the relative error of 0.0146 (or roughly 10- 2) is accept-
able.
Amount Borrowed from the BND~
In the optimal solution.
the farmer
borrows 523.823 CFAF from the BNDA in the first year. amount which is
7.25% lower than the counterpart observed value 562.000 CFAF.
The dif-
ference in the two va1 ues may be due to either error in the right-hand
sides of the BNDA credit supply constraints or the error in the input
coefficients.
From Table 5.5 it appears that the right-hand side of the BNDA
credit supply constraints could be increased on the average by 30.000
CFAF in the first year without changing the basis; however the amount

115
borrowed from the BNDA would increase. hence reducing the difference
between the model output and the observed value.
As
mentioned earlier.
the amount of credit supplied by BNDA was estimated using the re1ation-
CL
ship R = Bwhere B is the amount borrowed from the BNDA and CL the BNDA
credit limit. B was given while the ratio R was assumed constant for the
three sources of credit and computed using credit information from the
moneylender.
This estimated value might actually be smaller than the
true value and _hence the corresponding amount borrowed would be smaller
than the observed value.
The choice of the input coefficients might also influence the ac-
tivity level in the solutio~
For example if they are chosen too large.
they may greatly restrict the activities such as borrowing.
Amount Borrowed from the Commercial Bank.
At optimal solution. the
farmer borrows 114.367 CFAF from the commercial bank in the first year.
This amount is 21.20% lower than the observed value of 138.600 CFAF.
But
given that the commercial bank and the BNDA credit limits were estimated
in a similar way the same arguments made above can be used to explain the
discrepancy between the model result and the system output.
Amount Borrowed from the Moneylender.
In the optimal solution.
the
farmer borrows 34.637 CFAF from the moneylender in the first year. an
amount which is 1.78 times lower than its counterpart observed value of
96.400 CFAF.
Such a large discrepancy requires a re-examination of the
relations defining this variable in the model.
The primary concern
should be on determining the actual reactions of other lenders when the
farmer borrows from this source.
One needs also to correctly estimate
the relationship between the interest rate charged by the moneylender and

116
the liquidity values of its credit reservation.
given that moneylender
credit reservation appears more rewarding for the modelled farmer.
Planted Areas.
The model generates in the first year an optimum of
areas planted to cassava. maize. rice and yam which are 0.70. 0.76. 0.92
and 1.01 hectares.
respectively.
The model's estimation is exactly equal
to the observed sample mean in the cases of maize and ric~
In the case
of cassava.
the model output exceeds the counterpart system observation
of 0.51 hectare. and the estimated area planted to yam is lower than the
observed value of 1.20 hec tares.
But given the magni tude of the devia-
tions. we may accept the model result.
Marketing Activities.
The model gives zero kilogram of cassava sold
in the first year. value which agrees with the observatio~
The amounts
of maize. rice and yam sold in the first year are 676. 340 and 1061 kilo-
grams. respectively.
In the cases of maize and yam.
the model's estima-
tions exceed their observed values of 306 and 970 kilograms.
respective-
ly; it is lower in the case of rice which has an observed value of 718
kilograms.
Here. we would accept the model output in the case of yam
because the magni tude of its error is small enough. But in the case of
rice and maize. we would need to improve on the definition of the vari-
ables in the model.
The results of the above analysis may serve as an indication that
the default-augmented LSLP model has performed fairly well.
Indeed the
model results conform quite well with the values of most of the variables
for which we disposed of the necessary data to perform the evaluation.
Maybe a more sophisticated validation procedure such as the two-stage
problem solving approach proposed by McCarl and Nelson [38J.
or statis-

117
tical hypotheses testing procedure would have given better results.
But
owing to the relative importance. in the model. of the variables which we
have evaluated. we may limit ourselves to the current results.
Hence. we
conclude that the model is valid enough to serve as a basis for carrying
out post-optimal analyses such as evaluating the potential effect of
changes in selected variables.
an analysis which we approach in the next
chapter.

118
CHAPTER SIX
ANALYSIS OF POLICY REFORMS
6.1
Overview
The default-augmented LSLP model described in Chapter Four was
validated in Chapter Five with data gathered in a survey of farmers in
Bouake region.
Ivory Coast.
The farm-plan thus deduced will serve here
as a norm to evaluate the effects of varying policy variables such as the
interest rate. the size of the loan. and penalty costs imposed on the
borrower for delinquency and default.
In general. banks making loans are concerned about the interest rate
they receive on the loan and the riskiness of the loan.
This is re-
flected in the bank's close analysis of the applicant's income prospects.
the amount of loan given out and the amount of collateral it demands of
loan applicants.
The bank may change. whenever possible. the interest
rate. the size of the loan and the amount of the required collateral in
order to change the behavior of the borrower.
However. the effectiveness
of these changes to positively affect the borrower's attitude and thus
improve the performance of the lending institution is controlled by the
size of the change.
Stiglitz and Weiss showed in an article in 1981 [56]
that increasing the interest rate or the collateral requirements (beyond
some points) may decrease the returns to the bank by either decreasing
the average degree of risk aversion of the pool of borrowers or in a
multiperiod model inducing individual investors to undertake riskier
projects.
In another article in 1983
[57]
these two authors also showed
that there may exist a Nash equilibrium in which banks do not lend to

119
borrowers who do not repay their first-period loans.
Hence.
these vari-
ables will be changed for control purposes in the bank but without ex-
ceeding their equilibrium values.
In the current analysis. we will
determine the variations in the rate of interest and the amount of the
loan which are necessary for managing loan delinquency and default in the
Ivorian Government Supported Credit Program's credit activity for indi-
vidual medium-sized farmers described in Chapter One.
In the model.
delinquency-co~t is specified as an additional interest charge;
the cost
of default is expressed as a reduction in credit supply for future bor-
rowing.
Although one could study the effects of these two costs by
analyzing the interest cost and the loan limit. they will be analyzed
separately for purpose of clarity.
Improving the quality of the finan-
cial services that the bank provides may also serve as means for control-
ling how the borrower behaves with respect to the loan contract:
it may
either in itself generate necessary changes which will modify the bor-
rower's behavior or help enhance the effects of changes introduced by
policy variables such as the interest rate and the size of the loan.
In
the current study. we will determine how improving on the financial
services provided by the BNDA would modify the effects of interest rate
and the size of the loan.
6.2
Description of the Policy Variables and the Model Variants
6.2.1
Default-Penalty Cost
In the basic model.
default activities have been specified so that
default and borrowing are competitive with each other with respect to
credit supply.
That is.
an increase in default for example will reduce

120
the amount of credit available subsequently.
In reality. whether the
relationship between these two activities is effective depends upon the
relative scarcity of credit and the value of credit in reserve.
While
the nature of the relationship is reflected in the left hand side of the
credit supply row in the form of credit absorbed. its effectiveness is
implied in the right hand side of this row.
The degree of scarcity of
credit will increase if its supply decreases; the total amount of credit
used to suppor~ an activity will increase if either the level of the
activity or the amount used per unit of activity increases.
For this
analysis we will determine the effects of varying the amount of credit
used per unit of default. which is reflected in the coefficients of
default activities in the credit supply rows.
As discussed earlier.
the coefficients of default activities in
credi t supply rows were chosen to express the BNDA's response to loan
default. or default-penalty costs.
Because of lack of information on the
BNDA's actual quantity-response to default and because an appropriate
proxy for its measure was not available. a constant marginal cost was
assumed in order to simplify the analysis.
More specifically. it was
assumed that each CFAF of loan amount defaulted absorbs at least 1 CFAF
of credit supply.
The appropriate absorption rate was then found by
changing the specified level until a default rate comparable to the
observea value was found.
The absorption rate corresponding to the
solution presented in the preceding chapter is 1.10.
It is proposed here
to determine the effect of varying this coefficient from 1.10 CFAF to
1.65 CFAF (which corresponds to a 50% increase in the initial value).
The following points will be considered:
1.20. 1.35. 1.50 and 1.65.
We

121
are limiting ourselves to these 4 values. as it will be the case for
other input coefficient changes. in order to restrain the cost of the
analysis.
Indeed. the current version of the APEX III package being used
to solve the problem does not allow a parametric analysis with more than
one set of changes in the input coefficients.
Consequently each coeffi-
cient change requires a separate run.
6.2.2
Interest Rate
In developing countries.
GSCPs' rates of interest often are criti-
cized for being much lower than rational economic policy would dictate.
Indeed. interest rates should be high enough to cover lending cost and
maintain the viability of the lending institutio~
But this appears not
to be the case in most developing countries:
on the one hand.
the
lending costs are high due to high default rates. and on the other hand
the local government intervenes to maintain the interest rate at low
levels in order to assist the agricultural sector.
Hence.
the GSCPs
almost always operate at losses.
From an economic standpoint they must
raise the interest rate and/or reduce the rate of default in order to
maintain the viability of their lending institutio~
For example. given
the 18%1 annual default rate in the BNDA in 1982.
the BNDA would have to
charge an interest rate as high as 40% in order to at least break-even.
Although this interest rate appears to be high for subsidized loans. one
needs to consider it in order to prescribe an operating line for the bank
and thus improve the outreach of the program.
Hence. in this study. we
will determine the effects of varying the rate of interest required for
BNDA loans from its current level 9% to 40% corresponding to a 330%

122
increase.
The following rates will be considered:
17%. 25%. 32% and
40%.
In the model.
these changes will be reflected in the coefficients
of BNDA-loan-repayment activities in the cash accounting rows.
Normally one would expec t the farmer to borrow les s when the in-
terest rates are high.
If the interest rates are high and if the penal-
ty cost imposed on the borrower for default--the rate of reduction in
credit supply for future borrowing--remains constant then default becomes
less expensive due to the small amount borrowed.
This will be especially
noticeable in the-case of default on season 4 debt which only reduces the
credit supply in the last season of the planning period.
Such a pheno-
menon might be explained by the fact that when the amount borrowed in the
first year is small.
even a 100% default would not significantly reduce
the loan limit in the second year.
Consequently. default rate would
increase when the interest rate increases.
Even if the unit default cost
becomes low. the total cost in reality may be hig~
Hence. if the farmer
intends to borrow a significant amount in the second year. he should keep
borrowing in the first year to a minimum.
As a result.
borrowing in the
first year will be low when the interest rate increases.
But instead. if
both the interest rate and the cost of default increase the farm-borrower
defaults less and borrows more as will be indicated later.
In this
study. we will determine the effect of varying default cost from its
initial ,value 1.10 CFAF to 3.65 CFAF which represents 330% increase as in
the case of interest rat~
We will consider the following points:
1.75.
2.40. 3.00 and 3.65.

123
6.2.3
Amount of the Loan
Those who responded in the survey have objected to increases in the
rate of interest; but they were more than willing to pay higher rates if
the amount of the loan and the interest rate could be simultaneously
raised.
Hence. one would expect the farmer to be less sensitive to an
interest rate increase if this is accompanied by the bank's increasing
the size of the loan.
To include these in the model will require that
both the interest rate and the amount of credit supplied by the BNDA be
changed.
But here. we will determine (1)
the effects of varying the
amount of the loan. and (2) the effect of a high interest rate when the
amount of the loan varies.
Following (l). the right hand side of BNDA-
credit supply rows will be parametrically increased by 50% of the current
specification. in 5 equal steps.
In (2). we will parametrically increase
the loan size when the interest rate required by BNDA is set equal to
40%.
the break-even interest rate.
6.2.4
Delinquency-Penalty Cost
In the basic model.
delinquency cost is specified as an additional
interest charge computed based on the monthly 0.7% penalty charge imposed
by BNDA on all past-due loans.
We will identify here the effects of the
delinquency cost when the monthly penalty-charge is increased to 1.55%
corresponding to 2 times the (monthly) interest-charge on loan principal.
Because the delinquency cost is reflected in the coefficients of delin-
quent-repayment activity in the cash accounting rows.
only these coeffi-
cients will be changed in the current analysis.

124
6.2.5
Use of BNDA Credit Reserve as a Source of Liquidity
Because of the lengthy procedure in securing a loan from the BNDA
and because the needy farmers often are restricted in the use of loan
proceeds. the basic model did not include the borrower's valuation of
BNDA credit reserves.
Now. assume that the BNDA changes its loan-dis-
bursement policy from kind to all-cash loan disbursement.
Also assume
that the BNDA reduces the uncertainties about its permanence by (i)
requiring the _appropriate rate of interest on its loans. and (ii) low-
ering detau1 t rate so that its net lending cost approaches zero.
BNDA
credit. therefore. becomes not only a source of loans but also it can
provide the farmer with numerous services such as liquidity when it is
maintained in reserve.
Hence. as part of the changes carried out in
order to reflect the modified behavior of the farm-borrower in the loan
contract. it is proposed to model this credit as another source of li-
quidity to be managed to satisfy the farmer's liquidity requirements.
To
this end. the model will be extended to include (1) BNDA credit alloca-
tion and valuation activities. and (2) BNDA credit reserve and credit
accounting rows as described in Chapter Four.
The major change here is
BNDA credit which is
reserved at values
that increase the larger
(smaller) the proportion allocated to the credit account (reserve).
As
in the case of the other three sources. the liquidity values of BNDA
credi t 'shown in Table 4.5 not only vary wi th the proportion in reserve
but also vary with the seasons. We expect a BNDA credit which generates
the borrower's valuation of its reserves to provide for a solution to our

125
current search for a credit program for the group of medium-sized Ivorian
farmers.
The performance measures characterizing the borrower and
the BNDA
for each set of variable changes are summarized in Table 6.1
6.3
Model Variation Results
6.3.1
Effects of Varying the Default-Penalty Cost
Tables 6.2 and 6.3 illustrate the effects of varying the cost of
default.
all other variables held constant.
The default-penalty cost is
allowed to vary from 1.10 CFAF to 1.65 CFAF. taking on the following
values:
1.20. 1.35. 1.50. and 1.65.
The borrower appears to be losing
from the changes in the cost of default as indicated by the reduction in
the cash available to him.
In the first year. reserved cash decreases
while credit reserved remains constant at the moneylender. and increases
at the commercial bank. maybe due to the substitution relationships among
the sources of liquidity.
In the second year. however. both reserved
cash and credit reserved at the moneylender decreases; credit reserved at
the commercial bank remains constant in the first three steps then de-
creases in the last step.
Net cash flow from the optimal plan slightly
decreases when the default-cost increases from 1.10 to 1.20 then in-
creases with further increases in the cost exceeding its initial value.
But cash available to the farmer decreases continuously due to the reduc-
tions in the cash reserved in the second year.
These results are as
expected given that default decreases when its cost increases.
Indeed.
because default is modeled so that it increases the farmer's liquidity
requirements. one would expect some relaxations in the demand for li-

126
Table 6.1
Performance Measures for Evaluating the Model Variations
Performance Measures Characterizing:
Farm-Borrower
Lender: BNDA
Objective Function Valuea
Amount Loaneda:
by year
Reserved Casha:
by year
Delinquent Amount a:
by year
Reserved Credit a:
by source and by year
Default Amount a
Net Cash Flow from the Optimal Plana
Lending Costa
Cash Availableab
Net Lending Cost a c
athe unit is (000) CFAF
bgiven as the sum of net cash flow from the optimal plan and cash re-
served in the second year
cis measured as the difference between lending cost and interest cost

127
Table 6.2
Effects of Varying the Default-Penalty Cost on the Farm-
Borrower
PERFORMANCE MEASURES
Description
Values* When Default-Penalty** is at
Initial Cost
Alternative Costs
1.10
1.20
1.35
1.50
1.65
Objective Function Value
7276
7278
7266
7264
7256
Reserved Cash in
Year One
1082
1080
1080
1080
1080
Year Two
1678
1683
1661
1661
1640
Reserved Credit in
Year One. at
BNDA
Moneylender
81
81
81
81
81
Commercial Bank
77
92
92
92
92
Year Two. at
BNDA
Moneylender
50
40
40
40
40
Commercial Bank
58
58
58
58
55
Net Cash Flow from Op-
timal Plan
2809
2804
2815
2822
2827
Cash Available
4487
4487
4476
4473
4467
*The unit is (000) CFAF except otherwise specified
**Rate at which default diminishes credit in subsequent seasons

128
Table 6.3
Effects of Varying the Default-Penalty Cost on BNDA
PERFORMANCE MEASURES
Description
Values * When Default-Penalty ** is at
Initial Cost
Alternative Costs
1.10
1.20
1.35
1.50
1.65
Amount Loaned by BNDA
Year One
524
558
558
558
558
Year Two
616
782
772
759
750
BNDA Loan in Delinquency in
Year One. as
Total
95
152
154
153
156
% of Loan Principal
18
27
28
27
28
Year Two. as
Total
o
o
o
o
o
% of Loan Principal
o
o
o
o
o
BNDA Loan Defaulted in
Year One. as
Total
176
85
83
83
81
% of Loan Principal
33
15
15
15
14
Lending Cost in Year One
383
195
195
195
184
Net Lending Cost in Year One
330
145
145
145
134
*The unit is (000) CFAF except otherwise specified
**Rate at which default diminishes credit in subsequent seasons

129
quidity when default decreases.
Hence. reservation activities would
decrease as illustrated by the current results.
Also because the amount
to be repaid and the rate of repayment increase. more cash is absorbed
resu1 ting in the reduction of reserved cash.
This further results in a
decrease in cash available to the farmer.
The BNDA. however. gains from an increase in the amount loaned.
and
a decrease in the rate of default resulting in a substantial reduction in
net lending cost.
Table 6.3 shows that the amount loaned by BNDA in-
creases in both the first and the second year. remaining around 558 in
the first year.
Default rate. which ini tia11y is 33%. decreases to 15%
in the first step; it remains constant at 15% in the two following steps
and again decreases to 14% in the last step.
Consequently. lending cost
decreases.
Also net lending significantly decreases although its value
remains positive.
From these findings.
we conclude that increasing the
cost of default only benefits the BNDA;
the farmer is allowed to borrow
more since defau1 t decreases. but at the expense of cash available to
him.
6.3.2
Effects of Increasing Interest Rate for BNDA Loans
The effect of increasing BNDA interest rates on a group of farmers
in Ivory Coast has already been ana1yzed by Yabi1e [62].
Using the LSLP
model as described in Chapter Three the author simulates the effects of
increasing the interest rate charge by BNDA from 11% to 40%.
The results
indicate that the farmer borrows about 0.7% more from the BNDA when the
interest rate increases.
In ana1yzing his results. Yabi1e notes that the
amount borrowed should have decreased due to the high interest rat~
But

130
he explains these unexpec ted resu1 ts by the fac t that the farmer would
attempt to offset his financial losses due to the shortage of cash supply
at the end of each season. by borrowing more from the BNDA. which. unlike
the other sources. does not have reservation va1 ues.
A1though such an
argument may be acceptable.
the author does not explain why cash supply
is reduced when the interest rate increases.
From the summary of his re-
sults it appears that it is net cash from the optimal plan that decreases
instead of cash supply.
This should be expected.
Indeed. given the high
rate of interest_required by BNDA. if the amount borrowed from this
source increases.
then more cash should be used for payment of interest
and principal. and thus leave less cash for reservation.
Hence. that the
amount of cash decreases is a consequence rather than the cause of the
increase in the amount borrowed from the BNDA.
Logically. one would expect the farmer to borrow less at higher
interest rates.
We may explain that as follows.
Given the rate of
interest. the optimal amount of loan funds used in borrowing is determined
by equating the cost of borrowing with the marginal value product of loan
funds.
This is shown in Figure 6.1 by the distance OA on the horizontal
axis.
When the interest rate is increased. the marginal value product of
additional assets acquired with borrowed funds must increase in order to
restore the equilibrium.
This results in a smaller amount of loan funds
used for borrowing (distance OB).
Such a result has been pointed out by
Adams [1] an advocate of the concept that interest rate charged by GSCPs
in developing countries should be increased or made more flexible in
order to guarantee an efficient resource allocation.
This author argues
that increasing the
interest
rate
charged by
these
lenders
would

131
CFAF
Value of
Loan Funds
i ' I - - -
il----------------"'"-----..7--.....;::""'-::::,....----!
o
A
0 % - - - - - - - - - - - - - - - - - - - · 100%
Percent of Loan Fund Used in Loans
Figure 6.1
Equilibrium in the Use of Loan Funds
i is the borrowing cost and VL indicates the marginal-value
product curve of loan funds

132
strengthen rather than undermine their financial viability.
He also
points out that with higher interest rates. borrowers. especially current
large borrowers. will borrow less; but the lending institution will
benefit from the high interest rate in that it will be forced to seek
additional business from new and small borrowers.
Also.
high interest
rates are likely to generate prompt repayment.
hence reduce loan delin-
quency and default.
But when one is set up to model a representative
farmer as is the case here. all these effects cannot be observed.
The
only immediate ef~ect that will result from increasing the interest rate
is that borrowing will decreas~
Tables 6.4 and 6.5 display the effects of varying the BNDA interest
rate in the default augmented LSLP model.
Here the interest rate is
allowed to vary from its initial value 9% to 40% taking on the val ues
17%. 25%. 32% and 40%.
As a result. default increases and the borrower
gains from an improved cash position; net lending cost for the BNDA.
however. significantly increases. reaching 4 times the initial val ue ,
when the interest rate equals 40%.
Table 6.4 shows that the objective function value.
cash reserved in
the first year and net cash flow from the optimal plan. all decrease.
This is due to (1)
the reduction in the amount borrowed (Table 6.5)
resul ting in less cash supply especially in the first year. and (2) the
increase in cash absorption because of the high interest rates.
However.
because the borrower only pays a small fraction of the loan. he would
most likely to have more cash available in the second year which he
reserves in significant amount in order to satisfy his increased demand
for liquidity.
Consequently.
cash reserved in the second year and cash

133
Table 6.4
Effects of Varying the Interest Rate for BNDA Loans on the
Farm-Borrower
PERFORMANCE MEASURES
Description
Valuesb When the Interest for BNDA Loans is at
Initial Rate
Alternative Rates
a
9
17a
25a
32 a
40
Objective Function Value
7276
7249
7236
7217
7199
Reserved Cash in
Year One
1082
917
916
916
915
Year Two
1678
1896
1932
1917
1959
Reserved Credit in
Year One. at
BNDA
Moneylender
81
82
82
82
82
Commercial Bank
77
81
93
93
93
Year Two. at
BNDA
Moneylender
50
50
50
50
50
Commercial Bank
58
57
71
71
73
Net Cash Flow from
Optimal Plan
2809
2726
2652
2648
2586
Cash Available
4487
4622
4584
4565
4545
aThe current results were obtained for the values 9.24.
16.93.
24.62 and
32.31.
But because interest rates often are integer numbers. these
values are rounded to the nearest digit in this report.
bThe unit is (000) CFAF except otherwise specified

134
Table 6.5
Effects of Varying the Interest Rate for BNDA Loans on the
BNDA
PERFORMANCE MEASURES
Description
Valuesb When the Interest for BNDA Loans is at
Initial Rate
Alternative Rates
40
Amount Loaned by BNDA in
Year One
524
277
276
276
278
Year Two
616
608
582
582
581
BNDA Loan in Delinquency in
Year One. as
Total
95
34
11
11
5
% of Principal
18
12
4
4
2
Year Two. as
Total
o
o
o
o
o
% of Principal
o
o
o
o
o
BNDA Loan Def aul ted in
Year One. as
Total
176
201
225
225
229
% of Principal
33
73
81
81
82
Lending Cost in Year One
383
892
1739
1739
1849
Net Lending Cost in Year One
330
845
1670
1650
1737
a The current results were obtained for values 9.24.
16.93.
24.62 and
32.31
But because interest rates often are integer numbers.
these
values are rounded to the nearest digit in this report.
bThe unit is (000) CFAF except otherwise specified

135
available from the optimal plan increase when the interest rate varies.
Credit reserved at the commercial bank increases in both the first and
the second year.
The increase is more noticeable in the first year and
when the interest rate increases from 17% to 25%.
Credit reserved at the
moneylender slightly increases in the first year but remains constant at
50 thousand CFAF in the second year.
Overall.
reserved credit increases
in the first year in order to compensate for the decrease in cash reser-
vation.
This may be the result of the substitution relationships between
cash and c r e di.t ,
Normally. reserved credit in the second year should
have decreased given that cash in reserve increases in that year.
In-
stead it increases maybe due to the fact that the additional liquidity
requirements resulting from the high default rate are too large to be
satisfied only by the excess reserved cas~
Table 6.5 indicates that the amount loaned by BNDA in both the first
and the second year decreases when the interest rate increases. as ex-
pecte~
This is more noticeable in the first year when the interest rate
increases up to 25%.
In general.
the effects of increasing the interest
rate charged by a GSCP appear higher in this study than observed in the
studies based on the LSLP model. which we reviewed in Chapter Two.
This
may be explained by the fact that the current model includes default
variables.
Recall that in the default augmented LSLP model the BNDA
responds to default by reducing its credit supply for future borrowing.
Hence. it is possible that the cost of default resulted in an increase in
the cost of borrowing thus giving a much smaller amount of funds used in
loan.
as illustrated by the distance QC in Figure 6.1.
Borrowing in the
first year decreases more than that in the second year maybe to prevent a

136
too high total default-cost which would have eliminated borrowing in the
second year.
Table 6.5 also shows that the amount of loan defaulted
increases with the interest rate.
Consequently. lending cost increases.
Ultimately.
increasing the BNDA's interest rate generates high levels of
net lending cost for the BND~
This seems not to agree with the general
thinking on the effects of high interest rate on default [41]
and lending
cost.
But such a substantial difference may be due to an incomplete
specification of the interest rate variation in the model.
Indeed. if
the interest rate_increases and if the penalty cost imposed on the farmer
for default remains constant. it becomes cheaper to default especially on
season 4 debt.
That is. given the expression of the cost of default in
the model. if borrowing in the first year decreases.
then even a 100%
default would not substantially affect the loan limit in the second year.
Consequently.
default will increase when the rate of interest increases.
Hence.
we must also increase the default-penalty cost in order to fully
account for the effect of increasing the interest rate required by BND~
Tables 6.6 and 6.7 summarize the effects of varying the cost of
default when the interest rate for BNDA loans is set equal to 40%.
As in
the case of default-cost variation.
only the BNDA benefits from the
combination of default-cost changes with increased interest rate:
cash
available to the borrower decreases although the slight increase in its
value i~ the last two steps; net lending cost for the BNDA continuously
decreases becoming negative (corresponds to a profitable activity) when
the cost of default is increased from 1.75 to 2.40 CFAF and remains
negative with further increases in the cost.

137
Table 6.6 shows that the objective function value.
cash reserved in
the second year. reserved credi t and cash available. all dec rease when
the cost of default is increased up to 2.40 CFAF.
This is as expected
given that default decreases when its cost increases.
Recall that demand
for liquidity decreases when default decreases resulting in less cash and
credit reservation.
Also. because the interest rate is high and given
that the rate of repayment increases. more cash is absorbed especially in
the second year giving low levels of cash available.
The slight improve-
ment in the last -two steps in the values of the objective function.
cash
reserved in the second year and cash available may be due to the fact
that default rate drops to zero thus allowing for more borrowing especi-
ally in the second year.
This increases the cash supplied in that year
and also the amount reserve~
Table 6.7 indicates that default rate. which is 82% when the in-
terest rate alone is increased.
decreases reaching 11% when the cost of
default equals 2.40 CFAF and drops to zero with further increases in the
cost.
Delinquency. however. increases from 2% to 70%.
Hence. it appears
that the farmer chooses to be delinquent when the interest rate is high
and default-cost increases.
Nevertheless.
the BNDA should be better off
since the delinquent loan. unlike the amount defaulted. is recovered.
The fact that BNDA is better off is further illustrated by the lending
cost which decrease~ resulting in negative net lending costs or profits
for the BNDA.
It can be concluded that the BNDA will gain from in-
creasing the interest rate it charges if this is accompanied by its
strengthening the penalty cost imposed on the borrower for default.

138
Table 6.6
Effects of Varying Default-Penalty Cost When the Interest Rate
is Set Equal to 40% on the Farm-Borrower
PERFORMANCE MEASURES
Description
Values * When BNDA Interest Rate Equals 40% and
Default-Penalty** is at
Initial Cost
Alternative Costs
1.10
1.75
2.40
3.00
3.65
Objective Function Value
7199
7104
7044
7054
7052
Reserved Cash in
Year One
915
914
928
928
928
Year Two
1959
1869
1750
1784
1784
Reserved Credit in
Year One. at
BNDA
Moneylender
82
82
82
82
82
Commercial Bank
93
93
78
78
78
Year Two. at
BNDA
Moneylender
50
50
37
37
37
Commercial Bank
73
58
45
45
45
Net Cash Flow from
Optimal Plan
2586
2599
2685
2661
2659
Cash Available
4545
4468
4435
4445
4443
*The unit is (000) CFAF except otherwise specified
**Rate at which default diminishes credit in subsequent seasons

139
Table 6.7
Effects of Varying Default-Penalty Cost When the Interest Rate
is Set Equal to 40% on BNDA
PERFORMANCE MEASURES
Description
Values * When BNDA Interest Rate Equals 40% and
Default-Penalty** is at
Initial Cost
Alternative Costs
1.10
1.75
2.40
3.00
3.65
Amount Loaned by BNDA in
Year One
278
301
301
301
301
Year Two
581
660
805
882
883
BNDA Loan in Delinquency in
Year One. as
Total
5
84
179
211
211
% of Loan Principal
2
28
59
70
70
Year Two. as
Total
0
0
0
0
0
% of Loan Principal
0
0
0
0
0
BNDA Loan Defaulted in
Year One. as
Total
229
127
323
0
0
% of Loan Principal
82
42
11
0
0
Lending Cost in Year One
1849
295
87
45
45
Net Lending Cost in Year
One
1737
175
(33)
(75)
(75)
*Th
.
e um.t; is (000) CFAF except otherwise specified
**Rate at which default diminishes credit in subsequent seasons

140
In the current analysis.
increasing the default-penalty appears to
be an easy task.
This is mainly due to the fact that the model only
considers a simplified form of the cost of default.
In reality. policies
or sanctions that could be easily implemented in order to achieve similar
results might not be available to the lender.
Therefore. a lender is
likely to reduce his performance if he adopts a policy of high interest
rate because of lower loaned amounts and high default rates.
But as-
suming that this lender could identify the right combination of sanctions
to accompany the increase in the interest rate he requires. the amount he
could loan would increase but without reaching its initial value as
illustrated by Table 6.7.
This table shows that the amount loaned by
BNDA increases with the cost of default.
In the first year.
this amount
substantially increases in the first step then remains nearly constant
around 301 thousand CFAF in the remaining steps.
an amount which is 42%
lower than the initial value reported in Table 6.4.
This phenomenon
could be due to the interaction of the farmer's capacity and willingness
to repay his debts.
It might be possible that the BNDA has succeeded to
eliminate the farmer's unwillingness to repay with the first increase in
the cost of default.
But at higher default-penalty cost and given the
high interest charge the farmer borrows only that amount he could finan-
cially manage.
This should benefit the BNDA since it would loan out the
unused loan funds and gain from the high recovery rate of its loans.
Also. it will benefit a larger number of borrowers given the current high
net lending costs that restrict the outreach of the GSCP.

141
6.3.3
Effects of Increasing the Credit Limit for BNDA Loans
The effects of varying the credit limit for BNDA loans are presented
in Tables 6.8 and 6.9.
Here. the amount of credit supplied by the BNDA
is allowed to increase by up to 50% of the initial specification. in 5
equal increments.
Consequently. cash available to the borrower con-
tinuously increases; net lending cost for BNDA decreases when the credit
limit increases up by 30% of its initial value then increases when the
credit limit is further increased.
The objective function value.
reserved cash. net cash flow from the
optimal plan and cash available increase with the credit limit.
Reserved
credit. however. remains constant at both the moneylender and the commer-
cial bank in both the first and the second year.
As shown in Table 6.9.
the farmer borrows more from the BNDA and reduces his rate of defaul t
doubtless from improved performance provided by higher credit limits.
Normally more cash should be committed to use.
Instead reserved cash.
net cash flow and cash available increase.
One could attribute that to
the fact that the effect of the additional cash made available to the
farmer dominates the effect of the increased cash requirement for in-
terest and principal payment.
Default rate. which initially is 33%. decreases down to 21% when the
credit limit is increased to 30% over its initial value.
But when the
credit limit is further increased.
default rate increases again yet
without exceeding its initial level.
Normally one would expect default
to decrease when the credit limit increases since borrowing would in-
crease and raise the cost of defaul t ,
The increase in the cost will be

142
Table 6.8
Effects of Varying BNDA Credit Limit. Other Variables at Their
Initial Values. on the Farm-Borrower
PERFORMANCE MEASURES
Description
Values* When Other Variables are at Their Initial
Values and BNDA Credit Limit Increases at
Percentages
0
10
20
30
40
50
Objective Function Value
7276
7372
7461
7549
7638
7725
Reserved Cash in
-
Year One
1082
1107
1131
1156
1178
1203
Year Two
1678
1686
1693
1712
1734
1770
Reserved Credit in
Year One. at
BNDA
Moneylender
81
81
81
81
81
81
Commercial Bank
77
77
77
77
77
77
Year Two. at
BNDA
Moneylender
50
50
50
50
50
50
Commercial Bank
58
58
58
58
58
58
Net Cash Flow from
Optimal Plan
2809
2873
2931
2976
3019
3047
Cash Available
4487
4559
4624
4688
4753
4817
*The unit is (000) CFAF except otherwise specified

143
Table 6.9
Effects of Varying BNDA Credit Limit. Other Variables at Their
Initial Values. on BNDA
PERFORMANCE MEASURES
Description
Values* When Other Variables are at Their Initial
Values and BNDA Credit Limit Increases at
Percentages
0
10
20
30
40
50
Amount Loaned by BNDA in
Year One
524
581
640
695
724
776
Year Two
616
722
831
936
987
1088
BNDA Loan in Delinquency in
Year One. as
Total
95
128
163
190
191
210
% of Loan Principal
18
22
25
27
26
27
Year Two. as
Total
0
0
0
0
0
0
% of Loan Principal
0
0
0
0
0
0
BNDA Loan Defaulted in
Year One. as
Total
176
164
150
147
198
205
% of Loan Principal
33
28
23
21
27
26
Lending Cost in Year One
382
349
320
313
420
434
Net Lending Cost in Year
One
330
296
262
250
355
365
*The unit is (000) CFAF except otherwise specified

144
greater the higher the rate of increase in the amount borrowed in the
first year relative to the rate of increase in the credit supplied in the
second year.
From Table 6.9 it appears that the amount borrowed in the
first year increases successively by 11%. 10%. 9%. 4% and 7% while credit
limit increases by 10% at each step.
Hence. it might be that default
increases in the last two steps because of the rate of increase in
borrowing that falls far below the rate of increase in credit supply.
Lending cost. like default rate. decreases then increases when the credit
limit is Lnc r e a s e d beyond the 30%.
Finally. net lending cost signifi-
cantly decreases especially in the first three steps:
it then increases
in the last two steps but remains below its initial value (in terms of %
cost).
Overall. the current policy of larger loan size appears to be the
most effective in generating significant changes in both the farmer's
behavior and the BNDA's lending operation.
However.
in order to carry it
out the bank must have access to the additional loan funds.
which might
not always be the case:
what BNDA has access to must depend on repayment
rates.
But the bank may have a better control over an interest rate
variation.
Certainly. combining a policy of larger loan size with a
policy of high interest rate would help identify the changes necessary to
reduce loan default and thus improve on the borrower and the BNDA's
performances.

145
6.3.4
Effects of Varying the Credit Limit for BNDA Loans When
the Interest Rate is at the Break-Even Point
It is often believed that a policy of high interest rate would not
be too damaging to the farmer's welfare if it is combined with a policy
of larger loan size [1. 62].
In section 6.3.2 we pointed out that the
BNDA will efficiently use a policy of high interest rate to control
default if it can simultaneously increase the interest rate and the
penal ty cost imposed on the borrower for defaul t ,
But in reality. the
appropriate sanctions might not be available to the bank. or the bank may
not be given the freedom to implement these sanctions if they were
available.
Also it was indicated in section 6.3.3 that a policy of
larger loan size generates higher responses on the part of the borrower
and that the availability of loan funds could be a serious constraint in
carrying out the policy.
Here. we will determine how (1) a policy of
larger loan size would modify the behavior of the farmer when he must pay
a high interest charge.
and (2) the combination of a high interest rate.
high default-penalty cost and increased loan size policies could help
identify the necessary change on the part of the farmer and in the bank's
lending operatio~
Tables 6.10 and 6.11 illustrate the effects of increasing the credit
supplied by BNDA when the interest rate is set equal to 40%.
Cash
available to the borrower increases from its value obtained under a high
interest rate policy (Table 6.4).
approaching the level given by a larger
loan size policy (Table 6.8).
Net lending cost decreases significantly
relative to the resul ts in Table 6.5. but it remains untenable for the
BNDA.
This is due to the still high rate of defaul t ,
Hence. the bank

146
Table 6.10
Effects of Varying the BNDA Credit Limit When the Interest
Rate Equals 40%. on the Farm-Borrower
PERFORMANCE MEASURES
Description
Values * When Interest Rate Equals 40% and BNDA
Credit Limit Increases at Percentages
o
10
20
30
40
50
Objective Function Value
7199
7244
7308
7372
7435
7499
Reserved Cash in
Year One
915
907
915
923
931
939
Year Two
1959
1905
1939
1972
2005
2038
Reserved Credit in
Year One. at
BNDA
Moneylender
82
110
110
109
109
109
Commercial Bank
93
110
110
110
110
110
Year Two. at
BNDA
Moneylender
50
50
50
50
50
50
Commercial Bank
73
73
73
73
73
73
Net Cash Flow from
Optimal Plan
2586
2649
2672
2695
2718
2741
Cash Available
4545
4554
4611
4667
4723
4779
*The unit is (000) CFAF except otherwise specified

147
Table 6.11
Effects of Varying the BNDA Credit Limit When the Interest
Rate Equals 40%. on BNDA
PERFORMANCE MEASURES
Definition
Values * When Interest Rate Equals 40% and BNDA
Credit Limit Increases at Percentages
o
10
20
30
40
50
Amount Loaned by BNDA in
Year One
278
291
324
358
391
425
Year Two
581
603
702
802
901
1000
BNDA Loan in Delinquency in
Year One. as
Total
5
20
40
61
81
102
% of Loan Principal
2
7
12
17
21
24
Year Two. as
Total
0
0
0
0
0
0
% of Loan Principal
0
0
0
0
0
0
BNDA Loan Defaulted in
Year One. as
Total
229
254
251
248
245
242
% of Loan Principal
82
87
77
69
63
57
Lending Cost in Year One
1849
2278
1293
967
821
710
Net Lending Cost in Year
One
1737
2162
1163
823
665
540
*The unit is (000) CFAF except otherwise specified

148
must resort to default-penalties in order to control default.
We will
therefore modify the specifications of the model so that the control
variable set includes default-penalty cost.
Now.
assuming that the
current cost of defaul t doubles. i.e. increases to 2.20 CFAF per CFAF of
loan amount defaulted; setting the interest rate equal to 40%. we let the
amount of credit supplied by BNDA increase up by 50% of the initial
specificatio~ Cash available to the farmer improves significantly rela-
tive to its value given by a high interest rate policy. as shown in Table
6.12; but because of the increased cost of default it remains below the
level obtained when only the interest rate and the credit limit are
increased (Table 6.10).
Net lending cost for the BNDA reduces substan-
tially although it remains positive as indicated by Table 6.13.
The BNDA
also gains from an increase in the amount it loans.
and still possesses
loan funds for attracting new customers.
Hence. at a high interest rate.
the BNDA would significantly improve on its lending operation if it could
enforce the sanctions it imposes on the defaulter. while it increases its
loan size.
We expect such a policy to force the farmer to reduce his
default rate.
Delinquency rate.
however.
increases requiring that
sanctions be taken in order to hold it in check. as considered in the
next sec t ion.
6.3.5
Effects of Increasing the Delinquency-Penalty Cost
The effects of increasing the cost of delinquency are presented in
Tables 6.14 and 6.15.
It appears that increasing the cost of delinquency
from 0.7% to 1.55% leaves the performances of both the farmer and the
BNDA nearly unchange~
Cash available to the farmer decreases but only

149
Table 6.12
Effects of Varying BNDA Credit Limit. When Interest Rate
Equals 40% and Default-Penalty Cost is 2.20 CFAF.
on the
Farm-Borrower
PERFORMANCE MEASURES
Description
'*
Values
When Interest Rate Equals 40%. Default-
Penalty Cost is 2.20 CFAF. and BNDA Credit
Limit Increases at Percentages
0
10
20
30
40
50
Objective Func~ion Value
7037
7104
7170
7235
7301
7366
Reserved Cash in
Year One
913
921
928
935
945
955
Year Two
1683
1741
1780
1841
1878
1906
Reserved Credit in
Year One. at
BNDA
Moneylender
82
82
82
82
82
82
Commercial Bank
93
93
93
93
93
93
Year Two. at
BNDA
Moneylender
40
40
40
40
40
40
Commercial Bank
58
58
58
58
58
58
Net Cash Flow from
Optimal Plan
2728
2729
2749
2746
2766
2793
Cash Available
4411
4470
4529
4587
4644
4699
'*The unit is (000) CFAF except otherwise specified

150
Table 6.13
Effects of Varying BNDA Credit Limit When Interest Rate
Equals 40% and Defaul t-Penal ty Cost is 2.20 CFAF.
on BNDA
PERFORMANCE MEASURES
Description
*
Values
When Interest Rate Equals 40%. Default-
Penalty Cost is 2.20 CFAF. and BNDA Credit
Limit Increases at Percentages
0
10
20
30
40
50
Amount Loaned by BNDA in
Year One
301
331
361
392
422
452
Year Two
636
745
823
932
1007
1067
BNDA Loan in Delinquency in
Year One. as
Total
99
130
146
177
192
200
% of Loan Principal
33
39
40
45
45
44
Year Two. as
Total
0
0
0
0
0
0
% of Loan Principal
0
0
0
0
0
0
BNDA Loan Defaulted in
Year One. as
Total
113
103
107
98
104
117
% of Loan Principal
37
31
30
25
25
26
Lending Cost in Year One
247
218
231
208
224
249
Net Lending Cost in Year
One
126
86
87
51
53
68
*The unit is (000) CFAF except otherwise specified

151
Table 6.14
Effects of Changing the Delinquency-Penalty Cost on the Farm-
Borrower
PERFORMANCE MEASURES
Description
Values* When Delinquency-Penalty** is at
Initial Cost
Alternative Cost
0.7
1.55
Objective Function Value
7276
7272
Reserved Cash ~n
Year One
1082
1081
Year Two
1678
1675
Reserved Credit in
Year One. at
BNDA
Moneylender
81
81
Commercial Bank.
77
77
Year Two. at
BNDA
Moneylender
SO
50
Commercial Bank
58
58
Net Cash Flow from
Optimal Plan
2809
2810
Cash Available
4487
4485
*The unit is (000) CFAF except otherwise specified
**Corresponds to additional interest charge imposed on the borrower for
delinquency

152
Table 6.15
Effects of Changing the Delinquency-Penalty Cost on BNDA
PERFORMANCE MEASURES
Description
Values* When Delinquency-Penalty** is at
Initial Cost
Alternative Cost
0.7
1.55
Amount Loaned by BNDA in
Year One
524
524
Year Two
616
614
-
BNDA Loan Delinqu~ncy in
Year One, as
Total
95
93
% of Loan Principal
18
18
Year Two, as
Total
o
o
% of Loan Principal
o
o
BNDA Loan Defaulted in
Year One, as
Total
176
177
% of Loan Principal
33
34
Lending Cost in Year One
383
388
Net Lending Cost in Year One
330
341
*The unit is (000) CFAF except otherwise specified
** Corresponds to additional interest charge imposed on the borrower for
delinquency

153
by 0.04% of the initial value.
The objective function value and reserved
cash decrease by less than 0.5% and reserved credit remains constant at
both the moneylender and the commercial bank in both the first and the
second year.
The amount loaned by BNDA remains constant and net lending
cost for the BNDA increases by 3% of its initial value.
Overall. the
121% increase in the cost of delinquency considered here.
appears to be
insufficient in bringing about a significant change in the farmer's
behavior and in the lending activity of BND~
Maybe a further increase
in this cost would help generate more noticeable effects.
Also a para-
metric analysis on some nonbasic activities as mentioned in Chapter Five
might be helpful.
But because of fund constraint we limit ourselves to
these results.
Moreover. we will treat the variable represented by
delinquency-penalty cost as a parameter in the rest of the analysis.
6.3.6
Effects of All-Cash Disbursement for BNDA Loans
As pointed out earlier.
credit can be managed to satisfy liquidity
requirements if it has some value; but to have a value. credit must be
accessible.
Given that it is accessible.
its liquidity value is higher
the fewer the restrictions on the use of loan proceeds.
In developing
countries. Government Supported Credit Programs (GSCPs) most often do not
meet these criteria mainly because of the loan proceeds which are re-
stricted to production purposes.
Indeed. in order to make sure that the
borrower exclusively uses the proceeds of the loan for production.
loans
are disbursed in kind in most of the cases.
It is expected that this
will generate enough income for the borrower to repay the loans and still
remain well off.
However.
default rates still remain high.
undermining

154
the viability of these programs.
One either has to improve the existing
credit programs or must replace them so that they become attractive to
the borrower.
As part of the proposals for their improvement.
previous
studies [15. 20. 34. 45. 62] suggest that an all-cash disbursement of
loans plus a net lending cost of zero might lead the borrower to value
the credit reserve of these programs and thus generate higher repayment
rates.
Here.
we will attempt to verify this suggestion for the case of
the "Pret de Faisance Valoir Normalise" (PFVN). a BNDA credit program for
individual medium-sized Ivorian farmers.
The results of the model variation presented above indicate (i) when
the interest rate for BNDA loans is equal to 29%.
(ii) when the unit
defaul t-cost equals 2.40 CFAF and (iii) when the credit limit on BNDA
loans increases by 30% over its initial specification.
net lending cost
drops to zero. as illustrated by Table 6.17.
This is mainly due to the
reduced rate of default resulting from the enforcement of default-penal-
ty. and the interest rate which is high enough to cover the cost of
lending.
Despite the high cost imposed on him for def aul t and des pi te
that he must pay a higher interest charge. the borrower has more cash
than he initially does (Table 6.16). certainly due to improved perf or-
mance provided by the higher credit limit.
We conclude that. by allowing
the BNDA to charge a higher interest rate and because of the substantial
reduction in default rate.
the current specifications contribute to
reduce the uncertainty about the BNDA's credit program.
Given these specifications that lead to a zero net lending cost. and
assuming an all-cash loan disbursement. BNDA credit is modeled as another
source of liquidity.
Tables 6.18 and 6.19 summarize the results.
In

155
Table 6.16
Effects of Interest Rate.
Default Cost and Credit Limit When
They Equal 29%. 2.40 CFAF and +30% of Initial Value Respec-
tively. on the Farm-Borrower
PERFORMANCE MEASURES
Description
Values * When Interest Rate. Default-Cost and
Credit Limit are
Equal 29%. 2.40 CFAF
and Up By 30% of Ini-
At Initial Values
tial Value Respectively
Objective Function Value
7276
7313
Reserved Cash in
Year One
1082
936
Year Two
1678
1866
Reserved Credit in
Year One. at
BNDA
Moneylender
81
82
Commercial Bank
77
93
Year Two. at
BNDA
Moneylender
50
45
Commercial Bank
58
45
Net Cash Flow from
Optimal Plan
2809
2806
Cash Available
4487
4672
*The unit is (000) CFAF except otherwise specified

156
Table 6.17
Effects of Interest Rate. Default-Cost and Credit Limit When
They Equal 29%. 2.40 CFAF and +30% of Initial Value Respec-
tively. on BNDA
PERFORMANCE MEASURES
Description
Values * When Interest Rate. Default-Cost and
Credit Limit are
Equal 29%. 2.40 CFAF
and Up By 30% of Ini-
At Initial Values
tial Value Respectively
Amount Loaned by BNDA in
Year One
524
392
Year Two
616
953
BNDA Loan in Delinquency in
Year One. as
Total
95
194
% of Loan Principal
18
49
Year Two. as
Total
o
o
% of Loan Principal
o
o
BNDA Loan Defaulted in
Year One. as
Total
176
48
% of Loan Principal
33
12
Lending Cost in Year One
383
118
Net Lending Cost in Year
One
330
o
*The unit is (000) CFAF except otherwise specified

157
Table 6.18
Effects of BNDA Credit Limit Which is Valued in Reserve on
the Farm-Borrower
PERFORMANCE MEASURES
Description
Values b When Interest Rate. Default-Cost. Credit
Limit are
Equal 29%. 2.40 CFAF. Up By
At Initial Value
30% Initial Value
Without LSa
With LSa
Objective Function Value
7276
7313
7599
Reserved Cash in
Year One
1082
936
945
Year Two
1678
1866
1799
Reserved Credit in
Year One. at
BNDA
327
Moneylender
81
82
73
Commercial Bank.
77
93
78
Year Two. at
BNDA
230
Moneylender
50
45
45
Commercial Bank
58
45
45
Net Cash Flow from
Opt imal PI an
2809
2806
2618
Cash Available
4487
4672
4417
~iquidity specification
bThe unit is (000) CFAF except otherwise specified

158
Table 6.19
Effects of BNDA Credit Limit Which is Valued in Reserve on
BNDA
PERFORMANCE MEASURES
Description
Values b When Interest Rate. Default-Cost. Credit
Limit are
Equal 29%. 2.40 CFAF. Up By
At Initial Value
30% Initial Value
Without LSa
With LSa
-
Amount Loaned by ~NDA in
Year One
524
392
471
Year Two
616
953
918
BNDA Loan in Delinquency in
Year One. as
Total
95
194
236
% of Loan Principal
18
49
50
Year Two. as
Total
o
o
o
% of Loan Principal
o
o
o
BNDA Loan Defaulted in
Year One. as
Total
176
48
o
% of Loan Principal
33
12
o
Lending Cost in Year One
383
118
71
Net Lending Cost in Year
One
330
o
(66)
aLiquidity specification
bThe unit is (000) CFAF except otherwise specified

159
this case.
the borrower has less cash than previously.
but he gains from
an increase in liquidity reserves for meeting unexpected demand for cash
as they come due.
These results agree with the study's contention that
when the borrower values the BNDA credit reserve. he substitutes more
credit reserve for cash which he certainly commits to production and for
repaying his loans.
Table 6.19 indicates that default rate drops to zero and net lending
cost becomes ne_gative thus expressing a profitable operation for BNDA.
These results suggest that (i) if BNDA loans were disbursed exclusively
in cash. (ii) if the borrower had access to a larger loan amount. and
(iii) if BNDA imposes a relatively high penalty cost on the borrower for
default.
then the program will generate profits with an interest rate of
29% which is lower than the 40% break-even rate pointed out by [62].
As
a result we would like to suggest the following policies that might
enhance the performance of both the borrower and the lending institution:
(1) BNDA changes its loan disbursement procedure to an all-cash
disbursement
(2) BNDA charges a 29% interest rate on its loans
(3) BNDA increases the loan limit by 30% over the current level. and
(4) BNDA imposes a unit penalty cost of 2.40 CFAF.

160
NOTES
1. 18% represents the rate of default estimated by Yabile [62] for the
BNDA as a whole.
Taking the average of the def aul t rates in each of
its credit programs in 1981-1982 shown in Table 1.3.
this rate should
equal 21%.
But in the current study we use 18% in order to compare
our policy analysis with Yabile's findings.

161
CHAPTER SEVEN
SUMMARY. CONCLUSION.
SUGGESTIONS
7.1
Summary and Conclusion
The economies
of
many
developing countries
are still dependent on
agriculture.
In Ivory Coast for example.
the agricultural sector is the
major supplier of foreign exchange earnings.
Two major characteristics
of this agriculture are (1)
the predominance of small family
size farms
and (2)
the lack of capital investment needed to increase production and
expand marketable surplus.
In order to promote this vital component of
the economy.
the government has developed several programs to assist the
agricultural sector.
An example of such programs is the Banque Nationale
pour le Developpement Agricole (BNDA).
Loans in this credit program are
disbursed mostly in kind in order to prevent the farmers from using the
proceeds
of the loans for purposes other than productio~
The expec ta-
tions are that this will generate enough income to repay the loans and
leave the farmer better off.
However. delinquency and default rates
remain high. hence undermining the viability of the lending institutio~
Previous studies in the area of loan delinquency and default in
developing countries can be grouped into
(1)
behavioral
studies
and
(2)
policy oriented studies.
Studies in the first group attempt
to explain
the phe~omenon of loan delinquency and default by establishing possible
relationships between default or delinquency and selected socio-economic
characteristics of the farmers and the lending institutions;
they do not
suggest steps to be taken to alleviate the problem and their underlying
model tends to be too small to adequately represent the system that

162
generates del inq uency and/ or def aul t ,
Studies in the second group not
only are based on larger models. but also they lead to steps necessary to
circumvent the problem.
One group of such studies was the Liquidity
Specified Linear Programming (LSLP) model which not only provides a means
to reflect a financial response to the uncertainty inherent in small
scale farming but also accounts for the link between the farm and the
household.
Unfortunately.
however.
the LSLP model does not include the
default behavior of the farm-borrower.
In this study.
we have attempted
to define lending. policy reforms which will help in restructuring the
BNDA's credit program so that the borrower can value its credit reserve.
We have gone from the assumption that an all-cash loan disbursement
combined with a zero net lending cost will make the borrower value the
BNDA credit reserve. and have identified lending-policy reforms which
will lead to zero net lending cost.
In order to do so.
we have modified
the LSLP model to include del inq uency and def aul t ac tivi ties. and have
expanded it over a period of 2 years with an eight-season specification
in order to account for the time dependence of the def aul t phenomenon.
The default-augmented LSLP model is validated with information gathered
in a survey of farmers from BNDA's cent er district. Bouake'.
The vali-
dated form of the model is then used to evaluate the effects of varying
(l) the cost of defaul t ,
(2) the interest rate for BNDA loans. (3) the
credit ~imit for BNDA loans and (4) the cost of delinquency. in order to
identify the conditions leading to a zero net lending cost.
The model is
further modified to include liquidity management vectors for BNDA credit
in order to determine the combined effect of a zero net lending cost and
an all-cash BNDA loan disbursement.

163
Contrary to the commonly believed effects of an interest rate in-
Crease on default and on a GSCP.
this study indicates that increasing the
interest rate leads to a higher default rate.
In that case.
cash avail-
able to the borrower increases and the BNDA's performance decreases.
As
expected.
default
decreases when sanctions imposed on the borrower for
default are enforced; consequently the borrower has less cash and the
BNDA's lending cost decreases.
When the credit limit varies. the per-
formance of both the borrower and the BNDA improves;
but because default
rate still remains significant. BNDA performs below the break-even point.
When (1)
the loan limit increases by 30% over its initial level.
(2)
default-penalty cost
equals 2.40 CFAF per CFAF of amount
of loan de-
faulted.
and (3) the interest rate increases from its initial value 9% to
29%. the net lending cost becomes zero.
If in addition to these three
conditions BNDA adopts an all-cash loan disbursement policy. then the
program becomes profitable;
this suggests that the break-even interest
rate may even be smaller than 29% which is far below the 40% actually
computed.
Finally. increasing the cost of delinquency leaves the per-
formance of both the borrower and the BNDA nearly unchanged;
also delin-
quency and default remain unchanged.
From our results. we may suggest the following policies that may
enhance the performance of the borrower and the BNDA:
(1) increase the interest rate for BNDA loans from the current rate
9% to 29%
(2)
increase the loan limit by 30% over its initial value
(3) increase the cost of default from its current value 1.10 to 2.40
CFAF

164
(4) adopt a policy of all-cash loan disbursement.
Adopting the first three policies will lead to a zero net lending
cost. and adding the fourth will make the program profitable and leave
the farmer well-off.
Normally. the bank could be satisfied with the
effect of the first
three conditions since its primary objective is not
to make profit but to assist the agricultural sector which almost exclu-
sively sustains the economy;
operating at the break-even point would be
sufficient to maintain its viability.
But one needs to consider the
fourth policy not only because it allows the bank to make profit but also
because it considers two important elements of the farmer's environment:
his liquidity management and consumption needs.
Indeed. expenses for the
household are as important as production expenses in the decision making
process of the farmer.
and such an importance should also hold in welfare
terms.
Hence. a credi t program to help the farmer should consider the
household and the farm-firm as one unit in order to completely deal with
the farmer's problem.
Cash disbursement has always been neglected in the
GSCPs in order to prevent leakage of loaned funds into nonproductive
expenditures.
But it should not be ignored that
the bank could benefit
from such an operatio~
Baker et al.
have shown that a secure source of
credit.
available for family emergencies and other consumption expenses
as well as for production purposes.
will persuade farmers to reduce the
level of their liquid asset holdings which they will invest in produc-
tion.
This argument is also supported by our findings.
If the bank
should continue with the disbursement of loan in kind.
such a loan has to
be supplemented by some loans in cash in order to allow for the farmer's
consumption and liquidity management requirements.

165
The implications of this study are twofold:
(1) From a methodological standpoint.
the study leads to a decision-
making model for the farm-borrower that reflects important characteris-
tics of his default behavior.
For example it accounts for means to
represent the lender's response to default and the defaulter's reaction
to such a response; it allows one to account for the relationship between
default and liquidity in the management of the farm-firm and household.
and the fact that the various periods in the model are mutually dependent
allows to determine the expected future impacts of decisions taken today
and the current effect of future policy-decisions.
Our model.
therefore.
could serve as a basis for the analysis of default management not only in
the other programs of the BNDA but also in credit programs in other
countries.
(2) In substance.
the study generates a set of policy-decisions that
could be verified by the BNDA.
The BNDA could actually use the values
derived here for the interest rate. the cost of default and the rate of
increase in the credit limit to achieve similar results.
7.2
Suggestions for Further Research
The default-augmented LSLP model and its variants seem to have
performed well especially in bringing about a significant change in the
rate of default.
It allowed to identify some policy changes that may
help to improve the performances of the borrower and the BND&
However.
the basic model needs to be improved in several aspects.
(1) The model was not successful in providing for an improvement in
the rate of delinquency:
increasing delinquency-cost leaves delinquency

166
and other important variables nearly unchanged. and policies that reduce
default increase delinquency.
This may be due to the fact that delin-
quent-repayment activity is specified so that it is directly related to
defaul t activities in the past-due-repayment account rows.
and is only
constrained by cash.
In future work. this activity should be respeci-
f i.e d, taking into account all possible relations that define it.
This
requires the identification of these relations. what can be done by
surveying both the farmer and the lending institutio~
(2) The BNDA,'s response to defaul t and the borrower's reaction to
default-penalties constitute two important elements in the study.
But
because of lack of information on their actual measures.
they have been
assigned some chosen values.
An empirical estimation of these two re-
sponses should help strengthen the model and improve our understanding of
the def aul t phenomenon.
In future research. one could design a survey
technique as the projective method developed by Harris [30]
to determine
the farmer's actual reservation prices on cash and credit.
This method
might be successfully used to estimate these two responses.
(3) The time dependence of the default phenomenon is accounted for
by defining the model over a two-year planning period.
This assumes that
the program terminates at the end of the second year.
But this will not
be true for a farmer who repays his debts as a response to the policies
determined here.
Hence. one needs to expand the model over a longer
planning period.
But because the loans considered here are short term
loans.
the choice of a two year period is appropriate.
In that case.
we
may account for decisions made after the second year by running the
current LP model recursively over the desired planning period.
The

167
recursive program would also allow to incorporate in the study. the fact
that some yields vary with the climatic conditions of alternate years.
For example.
coffee is often characterized by a biennial cycle with poor
years following good years.
Hence. if repayment is linked to its output.
then the response of default to the policies will depend on the a1ter-
nance of the years.
These should be investigated in future work.
(4) In the model.
sanctions imposed on the borrower for default are
expressed by the rate of reduction in credit supply for future borrowing.
These sanctions could also be treated as payoffs to the defau1 ter in a
Game problem.
Indeed. default involves negotiability. enforceability and
adoption of strategies.
elements which are fundamental in Game Theory.
The problem could therefore be transformed into a Game problem and solved
as such.
This might improve the theoretical background of the study.
Here default could be viewed as a rational strategy on the part of the
borrower in the aim of attaining the most beneficial payoff: but before
the borrower adopts it. he is supplied with all the possible consequences
of his actio~
So conceived. the game displays all the essential charac-
teristics of a non-cooperative.
non-zero-sum game such as the Prisoner's
Dilemma that can be played more than once.
This a1 ternative solution
approach could also be considered in future researc~

168
REFERENCES
[1]
Adams.
Dale. "Are the Arguments for Cheap Agricultural Credit
Sound 1" in Limi tations of Cheap Credit in Promoting Rural De-
velopment. Eds. Ad am , D. W•• D. H. Graham and J. D. Von Pd s c hk e ,
Parts A and B. 1983.
[2]
Ames; G. C. W•• Ryot's Reward:
A Study of Production Credit Repay-
ment Problems of Small Farmers in Mysore Sta t e , India. Unpub-
lished Ph.D. Dissertation. University of Tennessee. Knoxville.
1973.
[3]
Baker. C. B•• "Firm Growth. Liquidity Management and Production
Choices." in Production Economics in Agricultural Research.
Proceedings of Conference held at the Universi ty of Illinois.
March_8-10. s: E. 4108. 1966.
[4]
Baker. C. B•• Credit in the Production Organization of the Firm.
American Journal of Agricultural Economics. Vol. 50. Aug. 1968.
[5]
Baker. C. B•• Liquidity Management in Finance and Risk Be h av i.o r ,
Seminar Paper presented at La Trobe University.
Bundoora.
Vic-
toria.
April 22.
1982.
[6]
Baker.
C. B..
Research Methodology in Agricultural Economics.
Un-
published Manusc ript.
[7]
Baker. C. B•• Role of Credit in the Economic Development of Small
Farm Agriculture. AID Spring Review of Small Farmer Credit. Vol.
XIX. June 1973.
[8]
Baker.
C. B..
and V. x, Bhargava,
Financing Small-Farm Development
in India.
Australian Journal of Agricultural Economics.
August
1974.
pp. 101-119.
[9]
Ba rria r d,
C. S•• and J. S. Nix. Farm Planning and Control. Second
Edition. Cambridge University Press.
[10]
Ba r ry , P. J •• and C. B. Baker. Reservation Prices on Credit Use:
A
Measure of Response to Uncertainty. American Journal of Agricul-
tural Economics. Vol. 53. May 1971. pp. 222-227.
[11]
Ba r ry,
P. J ••
C. B. Baker and L. R. Sanint.
Farmers Credi t
Risks
and Liquidity Management. American Journal of Agricultural Econ-
omics. Vol. 63. May 1981. pp. 228-236.
[12]
Ba r ry , P. J •• J. ~ Hopkin and C. B. Baker. Financial Management in
Agriculture. Second Edition. The Interstate Printers.
Inc •• Dan-
v LLl e , IL. 1979.

169
[13]
Beneke,
R.
Raymond,
and Ronald Winterboer.
Linear Programming Ap-
plications to Agriculture.
The Iowa State University Press.
Ames.
1973.
[14]
Best.
B••
Socio-Economic Causes for Low Repayment Rates of Noncol-
lateral Institutional Rice Loans in the Philippines. Unpublished
M.S. Thesis. University of the Philippines at Los Banos. April
1981.
[15]
B'ha r g av a , V. K•• Effects of Publicly Supported Credit Programs on
Economic Growth of Small Farmers in District Budaun.
India.
Unpublished Ph.D. Dissertation. University of Illinois at Ur-
bana-Champaign.
1974.
[16]
Binet.
Er anc o i ae ,
Bilan National de l'Emploi en Cote d'Ivoire.
Republique Fran~aise. Ministere des Relations Exterieures Coop-
eration et Developpement.
Mai 1982. No. 47.
[17]
BNDA:
Annual Reports 1975-1982.
[18]
BNDA:
Seminar Report. Taabo. Ivory Coast. January 1983.
[19]
BNDA Staff Paper.
The Pret de Soudure of the BNDA. AID Spring
Review of Small Farmer Credits. Vol. VI.
Feb. 1973.
[20]
Cas t Ll.Lc, T. E••
Potential Effects of Modifying the Masagana 99
Program of the Philippines.
Unpublished Ph.D. Dissertation.
University of Illinois. Urbana-Champaign. 1982.
[21]
Cheng,
F.
L..
Financial Analysis and Planning.
Theory and Applica-
tion. Addison-Wesley Publishing Company. 1982.
[22]
Cheng,
F.
L..
Financial Analysis and Planning.
Theory and Applica-
tion:
A Book of Readings. Addison-Wesley Publishing Company.
1983.
[23j
Chhi.kara,
Raj.
Theory of Liquidity Management:
An Appraisal.
Un-
published Term Paper. Fall 1982. University of Illinois. Urbana-
Champaign.
[24]
Cur t i s , A. R•• and J. K. Re Ld , On the Automatic Scaling of Matrices
for Gaussian Elimination.
J.
Inst. Maths. Applies. 1972. 10.
pp.
118-124.
[25]
den Tu Lnd e r , Ba s t Lan, Ivory Coast:
The Challenge
of Success. A
World Bank Country Report. The John Hopkins University Press.
1978.
[26J
Gab r Le L, S. C•• and C. B. Baker. Concepts of Business and Financial
Risk.
American Journal of Agricultural Economics.
August 1980.
pp. 560-564.

170
[27]
Gal.
Thomas.
Postoptimal Analyses.
Parametric Programming and Re-
lated Topics. 1979.
McGraw-Hill Inc.
[28]
Gass.
I. Saul. Decision-Aiding Models:
Validation. Assessment. and
Related Issues for Policy Analysis. Operation Research. V 31: 4-
6. 1983. pp. 603-631.
[29]
• and B. W. Thompson. Guidelines for Model Evaluation:
An Abridged Version of
the U.
S.
General Accounting Office
Exposure Draft.
Operation Research.
28. 1980. pp. 431-439.
[30]
Harris.
S. Kim. A Projective Method for Eliciting Central Illinois
Farmers' Reservation Prices for Cash and Intermediate Credit Re-
serve. Unpublished Ph-D. Dissertation. University of Illinois.
Urbana-Champaign.
1985.
[31]
Heady. 0 .... Earl. and Wilfred Cand l e r , Linear Programming Methods.
Iowa Uniyersity Press.
1958.
[32]
Hun t s b e r g e r ,
D.V••
P.
Billingsley and D. J.
Croft.
Statistical
Inference for Management and Economics. 1975. Allyn and Bacon
Inc.
[33]
Kamajou.
F••
Government Financing of the Development of Small Farm
Agriculture in the Center South Province of Cameroon. Unpub-
lished Ph-D.
Dissertation.
University of Illinois.
Urbana-Cham-
paf.gn,
1978.
[34]
• and C. B. Baker. Reforming Cameroon's Government Credit
Program:
Effects on Liquidity Management by Small Farm
Bor-
rowers. American Journal of Agricultural Economics. November
1980. pp. 709-718.
[35]
Kwadno.
Boakye D..
A Review of the Farm Loan Repayment Problem in
Low Income Countries. Savings and Development.
No. 3. 1979.
Ill.
[36]
La dm an,
J. R••
and R. L. Tinnermeier.
The Political Economy of
Agricultural Credit:
The Case of Bolivia.
American Journal of
Agricultural Economics 63:1 (February) 1981.
pp.
66-72.
[37]
McCarl. A. Bzuce, Model Validation:
An Overview with Some Emphasis
on Risk Models.
Review of Marketing and Agricultural Economics.
Vol. 52. No. 3. December 1984. pp. 153-173.
[38]

and Carl H. Nelson.
Multiple Optimal Solutions in
Linear Programming Models:
Comment. American Journal of Ag-
ricultural Economics 65:1 (February) 1983.
pp.
181-183.
[39]
Meir. G. L•• Leading Issues in Development Economics. Oxford Uni-
versity Press. New York. 1964.

171
[40]
Ministere de l'Agricu1ture. Statistiques Agrico1es. Direction des
Statistiques Rurales et des Enq ue t e s Agrico1es.
1981.
[41J
Montie1.
E.
L•• Built-in Default in Agricultural
Credit Programs.
Unpublished P~D. Dissertation. Harvard University.
1983.
[42]
N'i.m a L, S•• An Analytical Approach to Small Farmer Loan Defaults.
Savings and Development. No. 2. 1978 IV.
[43]
Nne b e, S•• Centre National de Promotion des Entreprises Coopera-
tives (CENAPEC).
AID Spring Review of Small Farmer Credit.
Vol.
VI. Feb. 1973.
[44]
Oc t av i o , G. G•• Masagana 99:
Loan Repayment and Technical Assis-
tance. Journal of Agricu1 tura1 Economics and Development. pp.
98-111.
[45]
• -Modifications of Small Farmer Credit in the Maisan 77
Program of the Philippines. Unpublished ph.D. Dissertation.
University of Illinois. Urbana-Champaign. 1982.
[46]
Paris.
Quirino.
Mu1 tip1e Optimal Solutions in Linear Programming
Models. American Journal of Agricultural Economics 63:4 (Novem-
ber) 1981. pp. 724-727.
[47]
Multiple Optimal Solutions in Linear Programming
Models:
Reply. American Journal of Agricultural Economics 65:1
(February)
1983.
pp.
184-186.
[48]
Po Ll.a r , S. K•• and H. S. Grewa1. Simulating the Impacts of Credit
Policy and Fertilizer Subsidy on Central Luzon Rice Farmers. the
Philippines:
Comment.
American Journal
of Agricultural
Economics Vol. 65. May 1983. pp. 349-350.
[49]
Pradhan. J.. and J. S. Sharma. Factors Discriminating the Borrowers
in Crop Loan Repayment of a Branch of Al1ahabad Bank. Financing
Agriculture. Vol. XIII. No. 4. October-December 1981. pp. 24-28.
[50]
Presidential Committee on Agricultural Credit. A Study on the
Nonrepayment of Agricultural Loans in the Philippines.
1978.
[51]
Rice. E. B•• Summary of the Spring Review of Small Farmer Credit.
AID Spring Review of Small Farmer Credit. Vol. 20. 1973.
[52]
Ros e g r an t , Mark. Choice of Technology. Production and Income for
Philippines Rice Farmers:
Agricultural Policy and Farmer Deci-
sion Making.
Unpublished Ph.D.
Dissertation.
University of
Michigan.
1978.
[53J
• and R. He r d t , Simulating the Impact of Credit Policy
and Fertilizer Subsidy on Central Luzon Rice Farms. the Philip-
pines. American Journal of Agricultural Economics. Vol. 63.
1981. pp. 655-665.

172
[54]
Sargent.
R. G•• ''Verification and Validation of Simulation Models"
in Progress in Modelling and Simulation.
Ed.
F. E.
Cellier.
Aca-
demic Press.
1982.
[55]
Stewart. M. T. Lan , Reasoning and Method in Economics:
An Intro-
duction to Economic Methodology. McGraw-Hill Book Company. 1979.
[56]
S't i g Ld t z , J. E••
and A. Weiss.
Credit Rationing in Markets with
Imperfect Information. American Economic Review. June 1981. 71.
pp. 393-411.
[57]

Incentive Effects of Terminations:
Application to the
Credi t and Labor Markets. American Economic Review. December
1983.73. pp. 912-927.
[58]
Tinnermeier. R•• and C. Dowswell. Workshop Report:
Small Farmer
Credit r AID Spring Review of Small Farmer Credit. Vol. 20. 1973.
[59]
Vogel.
R. c.. Rural Financial Market Performance:
Implications of
Low Delinquency Rates. American Journal of Agricultural Eco-
nomics.
Vol. 63.
February 1981.
pp. 58-65.
[60]
Williams.
H. P..
Model Building in Mathematical Programming.
John
WHey & Sons.
1978.
[61]
Yabile.
R. K•• Viability of BNDA's Credit Programs.
Unpublished
M.S. Thesis.
University of Illinois.
Urbana-Champaign.
1979.
[62J

Viability
of Selected Agricultural Credit Programs in
the Ivory Coast. Unpublished Ph.D. Dissertation. University of
Illinois.
Urbana-Champaign.
1982.

173
APPENDIX

I
Table Al.
Linear Programming Model - Activity Vectors
CASSl
IolAIZEl
RICEl
YAMl
1-1.811
1-1.812
1-1.813
1-1.814
LAND1A
1.00
1 .000
LAND18
1.00
1.00
FLJl811
-13 .07
-23.00
-46.733
1.00
FLAB12
6.54
35.00
30.068
-1 .00
FLJl813
20.00
20.00
40.199
-1.00
FLAB14
-30.00
-40.00
-81 .293
1.00
CAC11
-1857.72
-2086.37
-2518.88
-449.50
CAC12
-928.86
-4172.73
-1259.44
-449.50
CAC13
-5051.20
-449.50
CAC14
-2525.60
-449.50
YAMlll
1400.00
YAMI13
-825.00
YAMI14
2475.00
MZE 111
19.00
MZEI12
-1394.74
RCE 113
55.20
RCE 114
1361.00
CASS 113
-836.00
CASSI14
836.00
IolAXRCEl
-1.22
1.00
.....
....,
~

Table Al.
Linear Programming Model - Activity Vectors (continued)
SLRCE14
SLMlE12
SLCASSI3
SLCASSI4
SLYAM13
SLYAM14
RCE 114
-1.00
MlE 112
1.00
CASSI13
1.00
CASS 114
-1.00
YAMI13
1.00
YAMI14
-1.00
CSH12
-25.00
CSHI3
-17.17
-50.00
CSHI4
-65.00
-17.17
-50.00
LQRRI2
-4.50
LQRRI3
-0.11
LQRR14
-10.40
-0.11
TRRCEI112
TRRCEI213
TRRCEI314
TRMAIZE1112
TRMAIZE1213
TRMAIZE1314
TRYAM1112
TRYAMI213
TRYAM1314
TRCASI112
TRCASI213
TRCASI314
RCE III
1.00
RCE 112
-1 .00
1.00
RCE113
-1.00
1.00
RCE 114
1.00
MlE 111
1.00
MlE 112
-1.00
1.00
MlEI13
-1.00
1.00
MlE 114
1.00
YAMlll
1.00
YAMI12
-1.00
1.00
YAMI13
-1 .00
1.00
YAMI14
1.00
CASSIII
1.00
CASS 112
-1.00
1.00
CASSI13
-1.00
1.00
CASS 114
1.00
.........VI

Table Al.
Linear Programming Model - Activity Vectors (continued)
ffiCEll
CRCEI2
ffiCEI3
CRCEI4
(}.lA IlEII
CMAIlEI2
CJ.1AllEl3
CMAIlEI4
RCE 111
1.00
RCE 112
1.00
RCE 113
1.00
RCE 114
-I .00
tJZE 111
1.00
tJZEI12
1.00
tJZEI13
1.00
j
tJZEI14
-I .00
PROTll
0.07
O.OB
PROT12
0.07
O.OB
PROT13
0.07
O.OB
PROT14
0.07
O.OB
CALCll
0.09
0.07
CALC12
0.09
0.07
CALCI3
0.09
0.07
CALC14
0.09
0.07
IRONll
0.02
0.05
IRON12
0.02
0.05
IRON13
0.02
0.05
IRON14
0.02
0.05
THIAMll
0.10
0.20
THIAM12
0.10
0.20
THIAMI3
0.10
0.20
THIAM14
0.10
0.20
....
.......
(J\\

Table Al.
Linear Programming Model - Activity Vectors (continued)
CYAMll
CYAM12
CYAM13
CYAM14
CCASSll
CCASS12
CCASS13
CCASS14
YAMll1
1.00
YAMI12
1.00
YAMI13
1.00
YAMI14
-1 .00
CASSll1
1.00
CASS 112
1.00
CASS 113
1.00
I
CASS 114
-1.00
PROT11
0.02
0.01
PROT12
0.02
0.01
PROT13
0.02
0.01
PROT14
0.02
0.01
CALC11
0.30
0.68
CALC12
0.30
0.68
CALC13
0.30
0.68
CALC14
0.30
0.68
IRONll
0.02
0.02
IRON12
0.02
0.02
IRON13
0.02
0.02
IRON14
0.02
0.02
THIAM11
0.50
0.04
THIAM12
0.50
0.04
THIAM13
0.50
0.04
THIAM14
0.50
0.04
....
"
"

Table AI.
Linear Programming Model - Activity Vectors (continued)
PURCEll
PURCE12
PURCE13
PUf-1AIZE11
PUMAIZE13
PUMAIZE14
PUYA~111
PUYA~112
PUCASSII
PUCASSI2
RCE 111
-1.00
RCEI12
-1.00
RCE 113
-1.00
MZEIII
-1.00
~IZE113
-1.00
~IZE114
1.00
YAMI11
-1.00
YAMI12
-1.00
CflSSll1
I
-1.00
CASSI12
-1.00
CAC11
-100.00
-70.00
-75.00
-53.00
CAC12
-100.00
-75.00
-53.00
CAC13
-100.00
-70.00
CAC14
-70.00
PUF I SH11
PUF I SH12
PUF I SH13
PUF I SH14
PROT11
0.13
PROT12
0.13
PROT13
0.13
PROT14
0.13
CALCll
0.14
CALC12
0.14
CALC13
0.14
CALC14
0.14
IRON11
0.01
IRON12
0.01
IRON13
0.01
IRON14
0.01
THIAMll
0.02
THIAM12
0.02
THIAM13
0.02
THIAM14
0.02
CACll
-100.00
CAC12
-100.00
CAC13
-100.00
CAC14
-100.00
....
.....
(X)

Table Al.
Linear Programming Model - Activity Vectors (continued)
BBNC11
BBNC13
BBNC14
BMLCll
BMLC12
BMLC13
BK.CI4
BCBCll
BCBC12
BCBC13
BCBC14
CSH11
-1 .00
-1.00
-1.00
CSH12
-1.00
-1.00
CSH13
-1 .00
-1.00
-1.00
CSH14
-1 .00
-1.00
-1.00
BNCL11
1.00
0.01
0.01
BNCL13
1.00
1.00
0.01
0.01
0.01
0.01
0.01
0.01
BNCL 14
1.00
1.00
0.01
0.01
0.01
0.01
0.01
0.01
0.01
0.01
K.CAll
0.005
1.00
0.005
~LCAI2
0.005
1.00
1.00
0.005
0.005
K.CAI3
0.005
0.005
1.00
1.00
1.00
0.005
0 1.005
0.005
MLCA14
0.005
0.005
0.005
1.00
1.00
1.00
1.00
0.005
0.005
0.005
0.005
CBCAll
0.005
0.01
1.00
CBCA12
0.005
0.01
0.01
1.00
1.00
CBCA13
0.005
0.005
0.01
0.01
0.01
1.00
1.00
1.00
CBCA14
0.005
0.005
0.005
0.01
0.01
0.01
0.01
1.00
1.00
1.00
1.00
BNDll
-1.00
BND13
-1 .00
BND14
-1 .00
MLDll
-1 .00
MLD12
-1 .00
MLD13
-1.00
MLD14
-1 .00
CBDll
-1 .00
CBD12
-1.00
CBOD
-1.00
CB014
-1 .00
BNCL21
1.00
0.01
0.01
0.01
0.01
0.01
0.01
BNCL21
0.01
0.01
BNCL22
MLCA21
0.005
0.005
1.00
1.00
1.00
0.005
0.005
0.005
MLCA22
0.005
1.00
0.005
MLCA23
0.005
1.00
0.005
CBCA21
0.005
0.005
0.01
0.01
0.01
1.00
1.00
1.00
CBCA22
0.005
0.01
1.00
CBCA23
0.005
0.01
1.00
LQRR11
1.00
1.00
1.00
LQRR12
1.00
1.00
LQRR13
1.00
1.00
1.00
LQRR14
1.00
1.00
1.00
....
......
10

Table Ale
Linear Programming Model - Activity Vectors (continued)
mt()1112
mt()l113
mt()1114
mt()1121-
DEFIIO
DEFI120
DEFI140
DEFI160
DEFI180
DEFIIIOO
!
BI{)II
1.00
1.00
1.00
1.00
5.00
2.50
1.67
1.25
1.00
BNRA(X;21
1.00
-1.00
-4.00
-1.50
-0.67
-0.25
CACI2
-I .015
CAC13
-1.031
CACI4
-1.061
CAC21
-1.113
LQRR11
-1.00
-1.00
-1.00
-1.00
-1.00
LQRRI2
1.00
-I .00
-1.00
-1.00
-1.00
-1.00
LQRRI3
1.00
-1.00
-1.00
-1.00
-1.00
-1.00
LQRR14
1.00
-I .00
-1.00
-1.00
-1.00
-1.00
LQRR21
1.00
BNCL21
1.10
1.10
1 .10
1.10
1.10
BNCL23
1.10
1.10
1.10
1.10
I .10
BNCL24
1.10
1.10
1.10
1.10
I .10
BNCL13
-I .00
-1.00
"'-CA12
-0.005
"'-CAI3
-0.005
-0.005
"'-CA14
-0.005
-0.005
-0.005
CECA12
-0.005
CECAI3
-0.005
-0.005
CECAI4
-0.005
-0.005
-0.005
-Represents dell nquent repayment var I abl e
...ODo

Table AI.
Linear Programming Model - Activity Vectors (continued)
I
If3tll1314
ltitll1321
If3tllI322*
DEF130
DEF1320
DEF1340
DEF1360
DEfI360
DEf13100
Btll13
1.00
1.00
1.00
5.00
2.50
1.67
1.25
1.00
BNRACC23
1.00
-1 .00
-4.00
-1.50
-0.67
-0.25
CAC14
-1.031
CAC21
-1.061
CACl2
-1.099
lQRR13
-1.00
-1.00
-1.00
-1.00
-1.00
LQRR14
1.00
-1 .00
-1.00
-1.00
-1 .00
-1 .00
lQRR21
1.00
-1.00
-1.00
-1.00
-1.00
-1.00
LQRR22
1.00
BNCl23
1.10
1.10
1.10
1.10
1.10
BNC124
1.10
1.10
1.10
1.10
1.10
BNQ.14
-1.00
RCA14
-0.005
M..CA21
-0.005
-0.005
CBCA14
-0.005
CBCA21
-0.005
-0.005
*Represents delinquent repayment variable
....
CO
...

Table Al.
Linear Programming Model - Activity Vectors (continued)
,
IIltlll421
IIltlll422
IIltlll423
1Ilf.V1424*
DEFI40
DEFI420
DEFI440
DEfl460
DEFI460
DEFI4100
BtIll4
1.00
1.00
1.00
1.00
5.00
2.50
1.67
1.25
1.00
BNRACC24
1.00
-1.00
-4.00
-1.50
-0.67
-0.25
CAC21
-1.031
CAC22
-1.046
CAC23
-1.061
CAC24
-1.113
LQRRI4
-1.00
-1.00
-1.00
-1.00
-1.00
LQRR21
1.00
-1.00
-1.00
-1.00
-1.00
-1.00
LQRR22
1.00
-1.00
-1.00
-1.00
-1.00
-1.00
LQRR23
1.00
-1.00
-1.00
-1.00
-1.00
-1.00
LQRR24
1.00
BNa..24
1.10
1.10
1.10
1.10
I .10
BNCL21
-I .00
"'-CA21
-0.005
"'-CAl2
-0.005
-0.005
"'-CA23
-0.005
-0.005
-0.005
CB CAlI
-0.005
CBCA22
-0.005
-0.005
CB CAn
-0.005
-0.005
-0.005
-Represents del fnquent repayment varlabl e
....
00
N

Table Al.
Linear Programming Model - Activity Vectors (continued)
RR01112
RR01113
RR01114
RROI213
RR01214
RROl221
RROl314
RROl321
RR01421
RR01422
RR01423
ROll
1.00
1.00
1.00
"'-012
1.00
1.00
1.00
ROl3
1.00
1.00
ROl4
1.00
1.00
1.00
CACI2
-1.083
CACI3
-I .167
-1.083
CAC14
-1.333
-I .250
-1.167
CAC21
1.48
-1.333
-1.167
CAC22
-1.250
CAC23
-1.333
LORR12
1.00
LQRR13
1.00
1.00
LORR14
1.00
1.00
1.00
LQRR21
1.00
1.00
1.00
LORR22
1.00
LQRR23
1.00
BNCLJ3
-0.01
-0.01
BNCL 14
-0.01
-0.01
-0.01
RCA12
-I .00
RCA13
-1 .00
-1.00
RCA14
-1.00
-1.00
-1.00
(BCA12
-0.01
(BCAI3
-0.01
-0.01
(BCA14
-0.01
-0.01
-0.01
....
e

Table Al.
Linear Programming Model - Activity Vectors (continued)
RM..DI213
RM..D1214
RK.D1221
RM..D1314
RM..D1321
RM..D1421
RK.D1422
RK.D1423
BNQ..13
0.01
BNQ..14
0.01
-0.01
BNCL21
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
BNQ..23
-0.01
-0.01
-0.01
N..CA13
-1 .00
N..CA14
-1.00
-1.00
-1.00
N..CA21
-1 .00
-1.00
-1.00
-1.00
-1.00
-1.00
K.CA22
-1 .00
-1.00
N..CA23
-1 .00
-1 .00
-1.00
a1CA13
-0.01
CBCA14
-0.01
-0.01
-0.01
a1CA21
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
a1CA22
-0.01
-0.01
a1CA23
-0.01
-0.01
-0.01
....
~

Table Al.
Linear Programming Model - Activity Vectors (continued)
RCBOl1l2
RCBOII13
RCBOll14
RCBOl213
RCBOl214
RCBOl221
RCBOl314
RCBOl321
RCB01421
RCBOl422
RCB01423
,
CBOII
1.00
1.00
1.00
CBOl2
1.00
1.00
1.00
CBOl3
1.00
1.00
CBOl4
1.00
1.00
1.00
CACI2
-1.027
CACI3
-I .055
-1.027
CACI4
-1.110
-1.082
-1 .055
GAelI
-I .137
-1.110
-1.055
CAC22
-1 .082
CAC23
-1.110
lQRRI2
-I .00
lQRRI3
-I .00
-1.00
LQRRI4
-1.00
-1 .00
-I .00
lQRR21
-1.00
-1.00
-1.00
LQRR22
-I .00
lQRR23
-I .00
BNC1I3
-0.01
-0.01
-0.01
BNQI4
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
BNCl21
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
BNQ23
-0.01
-0.01
-0.01
K.CAI2
-0.005
K.GA13
-0.005
-0.005
-0.005
K.CAI4
-0.005
-0.005
-0.005
-0.005
-0.005
-0.005
K.CA21
-0.005
-0.005
-0.005
-0.005
-0.005
-0.005
K.CA22
-0.005
-0.005
K.CA23
-0.005
-0.005
-0.005
rnCAI2
-1.00
CB GAl 3
-I .00
-I .00
-I .00
CBCAI4
-I .00
-I .00
-1.00
-1.00
-I .00
-I .00
CBCA21
-1.00
-1.00
-1.00
-1.00
-1.00
-1.00
CBCA22
-1.00
-I .00
CBCA23
-I .00
-1.00
-1.00
....
~

Table Al.
Linear Programming Model - Activity Vectors (continued)
lBNI I 13
lBN1314
lBN1421
lBN2123
lBN2324
BaNCI 1
BBNC13
BBNC14
BNo.l1
1.00
BNCl13
-1.00
1.00
BNo.I4
-1.00
1.00
BNCl21
-1 .00
1.00
BNo.23
-1 .00
1.00
BNCL24
-I .00
MAXBNTll13
1.00
-0.571
WoXBN11314
1.00
-1 .00
MAXBNTI421
1.00
-1.00
WoXBNT2123
1.00
WoXBNT2324
1.00
OCl112
OC1213
OC1314
OCl421
T/oL2122
T/oL2223
OC2324
B/oLCI I
B/oLCI2
B/oLCI3
BKL14
/oLo.ll
1.00
/oLCLl2
-1 .00
1.00
/oLo.l3
-1 .00
1.00
/oLCLl4
-1 .00
1.00
/oLo.21
-1 .00
1.00
/oLCl22
-1 .00
1.00
/oLo.23
-1.00
1.00
/oLCL24
-1 .00
MAXMl Tll 12
1.00
-1.00
MAXM.. Tl213
1.00
-1.00
MAXMl11314
1.00
-1.00
MAX/oL Tl421
1.00
-1.00
MAXMl T21 22
1.00
MAX/oL T2223
1.00
MAXMl T2324
1.00
~
~

Table Al.
Linear Programming Model - Activity Vectors (continued)
TCBII12
TCB1213
TCBI314
TCB1421
TCB2122
TCB2223
TCB2324
tiCBCII
BCBCI2
BCBCI3
BCBCI4
I
<IlClII
1.00
CB a. 12
-1.00
1.00
CBCLl3
-1.00
1.00
rna.14
-1.00
1.00
<IlCL21
-I .00
1.00
CBa.22
-1.00
1.00
CBCL23
-1.00
1.00
<Il a. 24
-1.00
MIIX<IlT1112
1.00
-1.00
MIIXCBT1213
1.00
-1.00
MIIXCB T1314
1.00
-1.00
MIIXCBT1421
1.00
-I .00
MAX<IlT2122
1.00
MA XCBT2223
1.00
MAX<IlT2324
1.00
TCSHI 112
TCSH1213
TCSH1314
TCSHI421
TCSH2122
TCSH2223
TCSH2324
TCSH24Z
GAC11
-I .00
CSH12
-1.00
CACI2
-I .00
CSH13
-1.00
CACI3
-1.00
CSHI4
-1.00
CACI4
-1 .00
CSH21
-1.00
CAC21
-1.00
CSH22
-1.00
CAC22
-1.00
CSH23
-1.00
CAC23
-1.00
CSH24
-1.00
CAC24
-I .00
OBJECTIVE
1.00
....
ClO
...,

Table Al.
Linear Programming Model - Activity Vectors (continued)
ACSHIIO
ACSHI120
ACSHI140
ACSHI160
ACSHII BO
ACSHI20
ACSHI220
ACSHI240
ACSHI260
ACSHI280
I
CSHII
1.00
1.00
1.00
1.00
1.00
GACII
1.00
0.80
0.60
0.40
0.20
CRI120
-0.20
CRII 40
-0.40
CRI 160
-0.60
CRIIBO
-O.BO
CSHI2
1.00
1.00
1.00
1.00
1.00
GACI2
1.00
0.80
0.60
0.40
0.20
CRI220
-0.20
CRI240
-0.40
CRI260
-0.60
CRI2BO
-O.BO
ACSHI30
ACSHI320
ACSH1340
ACSHI36Q
ACSH1380
ACSHI40
ACSHI420
ACSHI440
ACSHI460
ACSHI480
CSHI3
1.00
1.00
1.00
1.00
1.00
CAC13
1.00
O.BO
0.60
0.40
0.20
CR1320
-0.20
CRI340
-0.40
CRI360
-0.60
CRI380
-O.BO
CSHI4
1.00
1.00
1.00
1.00
1.00
CACI4
1.00
O.BO
0.60
0.40
0.20
CRI 420
-0.20
CRI 440
-0.40
CRI460
-0.60
CRI 480
-O.BO
....
OD
OD

Table Al.
Linear Programming Model - Activity Vectors (continued)

ARCll0
AM..Cl120
AM..Cl140
AM..Cl160
ARC1180
ARC120
AM..C1220
AMCL1240
AM..C1260
ARC1280
M..CLll
1.00 .
1.00
1.00
1.00
1.00
M..CAll
-1.00
-0.80
-0.60
-0.40
-0.20
M..CRl120
-0.20
M.. CRI 140
-0.40
M..CRl160
-0.60
M..CRl180
-0.80
M..C112
1.00
1.00
1.00
1.00
1.00
M..CA12
-1.00
-0.80
-0.60
-0.40
-0.20
M..CR1220
-0.20
M..CR1240
-0.40
M..CR1260
-0.60
M..CR1280
-0.80
AM..C130
AM..C1320
AM..C1340
AM..C1360
AM..C1380
AM..C140
AM..C1420
AM..C1440
AM..C1460
AM..C1480
M..C113
1.00
1.00
1.00
1.00
1.00
M..CA13
-1 .00
-0.80
-0.60
-0.40
-0.20
M..CR1320
-0.20
M..CR1340
-0.40
M..CR1360
-0.60
M..CR1380
-0.80
M..CL14
1.00
1.00
1.00
1.00
1.00
M..CA14
-1.00
-0.80
-0.60
-0.40
-0.20
M..CR1420
-0.20
M..CR1440
-0.40
M..CRl460
-0.60
M..CR1480
-0.80
....
co
\\0

Table Al.
Linear Programming Model - Activity Vectors (continued)
ACBCIIO
AmCI120
ACBCI140
AmCI160
AmCI180
ACBCI20
AmCI220
ACBCI240
AmCI260
ACBCI260
j
cau.u
1.00
1.00
1.00
1.00
1.00
mCAII
-I .00
-0.80
-0.60
-0.40
-0.20
CBrnl120
-0.20
m CRI I 40
-0.40
mrn1160
-0.60
CB CRI 180
-0.80
cao.iz
1.00
1.00
1.00
1.00
1.00
mCAI2
-1.00
-0.80
-0.60
-0.40
-0.20
csouzao
-0.20
mCRI240
-0.40
csonzso
-0.60
mCRI280
-0.80
AmCI30
AmCI320
AmCI340
AmCI360
ACBCI380
ACBCI40
ACBCI420
ACBCI440
ACBCI460
ACBCI480
-
mCL13
1.00
1.00
1.00
1.00
1.00
mCAI3
-1.00
-0.80
-0.60
-0.40
-0.20
mCRI320
-0.20
mrn1340
-0.40
mCRI360
-0.60
CBrn1380
-0.60
mCL14
1.00
1.00
1.00
1.00
1.00
mCAI4
-1.00
-0.60
-0.60
-0.40
-0.20
CB CRI 420
-0.20
mrnl440
-0.40
m CRI 460
-0.60
CBrnl460
-0.80
....
10
o

Table Al.
Linear Programming Model - Activity Vectors (continued)
VCSHI120
VCSHI140
VCSHI160
VCSH11 BD
VCSHlll00
VCSHI220
VCSH1240
VCSH1260
VCSH12BO
VCSH12100
CSH11
1.00
CRI 120
1.00
(RI 140
1.00
CR1160
1.00
(RllBO
1.00
LQRRII
1.00
1.00
1.00
1.00
1.00
CSHI2
1.00
CRI220
1.00
(R1240
1.00
CR1260
1.00
(R12BO
1.00
LQRRI2
1.00
1.00
1.00
1.00
1.00
OOJECTIVE
2.50
2.25
I.BO
1.70
1.65
3.10
2.70
2.20
2.00
1.95
VCSHI320
VCSH1340
VCSH1360
VCSH13BO
VCSH13100
VCSH1420
VCSH1440
VCSH1460
VCSH14BO
VCSH14100
CSHI3
1.00
CR1320
1.00
(R1340
1.00
(RI 360
1.00
(R13BO
1.00
LQRRI3
1.00
1.00
1.00
1.00
1.00
CSHI4
1.00
CRI420
1.00
(RI 440
1.00
CRI460
1.00
(R14BO
1.00
LQRRI4
1.00
1.00
1.00
1.00
1.00
OOJECT IVE
.3 .05
2.65
2.15
1.95
1.90
3.00
2.55
2.10
I. 75
I. 70
....
\\0
....

Table Al.
Linear Programming Model - Activity Vectors (continued)
VM.. Cl 120
VIoUI140
VM.. Cl 160
VM..C1180
VM..Cl1100
VM.. Cl 220
VM.. Cl 240
VM..C1260
VM.. Cl 280
VM..C12100
M..Q.ll
1.00
M.. CRI 120
1.00
M..ffi1140
1.00
M.. CRI 160
1.00
M..ffi1180
1.00
LQRRII
1.00
1.00
1.00
1.00
1.00
M..Q.12
1.00
M.. CRI220
1.00
M..ffi1240
1.00
M..CR1260
1.00
M..ffi1280
1.00
LQRRI2
1.00
1.00
1.00
1.00
1.00
OOJECTlVE
0.90
0.60
0.25
0.15
0.10
1.00
0.70
0.45
0.30
0.25
VM..C1320
VM..C1340
VM..C1360
VM..C1380
VM..C13100
VM..C1420
VM.. Cl 440
VM..CI460
VM.. Cl 480
VM..C141 00
M..Q.13
1.00
.
M.. CRI320
1.00
M..ffi1340
1.00
M..CR1360
1.00
M.. ffi 1380
1.00
LQRRI3
1.00
1.00
1.00
1.00
1.00
M..Q.14
1.00
M.. CRI 420
1.00
M..ffi1440
1.00
M..CR1460
1.00
M..ffi1480
1.00
LQRR14
1.00
1.00
1.00
1.00
1.00
OOJECTlVE
1.05
0.75
0.50
0.40
0.25
0.95
0.65
0.30
0.20
0.15
....
\\0
N

Table Al.
Linear Programming Model - Activity Vectors (continued)
V03CI120
V03CI140
VCBCI160
VCBC1180
V03CIIIOO
VCBCI220
VCBCI240
VCBCI260
VCllC1280
VCBCI2I00
030.11
1.00
CBCRI120
1.00
CBffill40
1.00
CBCRI160
1.00
cscansc
1.00
LQRRII
1.00
1.00
1.00
1.00
1.00
CBC112
1.00
CBffil220
1.00
CBCRI240
1.00
CBffil260
1.00
CB CRI280
1.00
LQRRI2
1.00
1.00
1.00
1.00
1.00
ffiJECTIVE
0.90
0.60
0.25
0.15
0.10
1.00
0.70
0.45
0.35
0.25
V03CI320
VCBCI340
VCBC1360
VCBCI380
VCBCI3100
VCBCI420
VCBCI440
VCBC1460
VCBCI480
VCBC14100
CB a.I 3
1.00
CBCR1320
1.00
CBffil340
1.00
CBCRI360
, .00
CBffil380
1.00
LQRRI3
1.00
1.00
1.00
1.00
1.00
CB a.I 4
1.00
CB CRI 420
1.00
CBffil440
1.00
CBCRI460
1.00
CBffil480
1.00
LQRRI4
1.00
1.00
1.00
1.00
1.00
ffiJECTIVE
1.05
0.75
0.50
0.40
0.25
0.95
0.65
0.30
0.20
0.15
.-
~

Table AI.
Linear Programming Model - Activity Vectors (continued)
CASS2
~IZE2
RIC£2
YAM2
H..B21
H..B22
H..B23
H..B24
lAND2A
1.00
1.00
lNID28
1.00
1.00
FUB21
-13.51
-23.76
-46.32
1.00
FlAB22
6.76
36.19
31.09
-1.00
FlIB23
20.66
20.66
41.57
-1.00
FlAB24
-31.02
-41.36
-64.06
1.00
CAC21
-1657.72
-2066 .37
-2516.66
-409.50
CAC22
-926.66
-4172.73
-1259.44
-409.50
CAC23
-5051 .20
-409.50
CAC24
-2525.60
-409.50
YAMI21
1400.00
YAMI23
-625.00
YAMI24
2475.00
~IZEI21
19.00
MAIZE 122
-1394.74
RCE 123
55.20
RCE 124
1361 .00
CASSI23
-636.00
CASS 124
636.00
MAXRCE2
-1.22
1.00
...\\0,f:o

Table AI.
Linear Programming Model - Activity Vectors (continued)
SLRCE24
SLW.IZE22
SLCASS23
SLCASS24
SLYAM23
SLYAM24
RCEI24
-1.00
MlE 122
1.00
CASS 123
1.00
CASS 124
-1.00
YAMI23
1.00
YAMI24
-I .00
CSH22
-25.00
CSH23
-17.17
-50.00
CSH24
-65.00
-17.17
-50.00
LQRR22
-6.00
LQRR23
-0.13
LQRR24
-9.60
-0.13
TRRCE2122
TRCE2223
mCE2324
TRMA IZE2122
mMAIZE2223
mMA IZE2324
TRYAM2122
mYAM2223
TRYAM2324
TRCASS2122
TRCASS2223
TRCASS2324
RCE 121
1.00
RCE 122
-1.00
1.00
RCE 123
-1.00
1.00
RCE 124
1.00
~IlE 121
1.00
~E122
-I .00
1.00
MZE 123
-I .00
1.00
~E124
1.00
YAMI21
1.00
YAMI22
-I .00
1.00
YAMI23
-1 .00
1.00
YAMI24
1.00
CASS 121
1.00
CASS 122
-1.00
1.00
CASSI23
-I .00
1.00
CASSI24
1.00
...
\\D
'""

Table Al.
Linear Programming Model - Activity Vectors (continued)
I
rna:21
cacezz
rna:23
CRa:24
CMAIZE21
().tA IZE22
Qo4A IZE23
CMAIZE24
Ra: 121
1.00
Ra: 122
1.00
RCE 123
1.00
RCEl24
-1.00
MlEI21
1.00
MlE 122
1.00
MlEI23
1.00
MlEI24
-1.00
ffiOT21
0.07
0.08
PROT22
0.07
0.08
ffiOT23
0.07
0.08
PROT24
0.07
0.06
CAlC21
0.09
0.07
CAlC22
0.09
0.07
CALC23
0.09
0.07
CAlC24
0.09
0.07
IR0N21
0.02
0.05
I R0N22
0.02
0.05
I R0N23
0.02
0.05
I R0N24
0.02
0.05
THIAM21
0.10
0.20
THIAM22
0.10
0.20
THIAM23
0.10
0.20
THIAM24
0.10
0.20
.....
\\0
Q\\

Table Ai.
Linear Programming Model - Activity Vectors (continued)
CYAM21
CYAM22
CYAM23
CYAM24
CCASS21
CCASS22
CCASS23
CCASS24
YAMI21
1.00
YAMI22
1.00
YAMI23
1.00
YAMI24
-1.00
CASSI21
1.00
CASSI22
1.00
CASSI23
1.00
CASSI24
-1.00
PROT21
0.02
0.01
PROT22
0.02
0.01
PROT23
0.02
0.01
PROT24
0.02
0.01
CAlC21
0.30
0.68
CALC22
0.30
0.68
CALC23
0.30
0.68
CALC24
0.30
0.68
IR0N21
0.02
0.02
IR0N22
0.02
0.02
IR0N23
0.02
0.02
IR0N24
0.02
0.02
THIAM21
0.05
0.04
THIAM22
0.05
0.04
THIAM23
0.05
0.04
THIAM24
0.05
0.04
....
ID
......

Table AL
Linear Programming Model - Activity Vectors (continued)
PURCE21
PURC£22
PURCE23
PUWIIZE21
PUWlIZE23
PUWlIZE24
PUYAM21
PUYAM22
PUCASS21
PUCASS22
RCE 121
-1.00
RCE 122
-1.00
RCE 123
-1.00
t-eE121
-1 .00
~IZE123
-1.00
~IZE 124
1.00
YA~1I21
-1.00
YAMI22
-1.00
CASSI21
-1.00
CASSI22
-I .00
CAC21
-100.00
-70.00
-75.00
-53.00
CAC22
-100.00
-75.00
-53.00
CAC23
-100.00
-70.00
CAC24
-70.00
PUF I SHZl
PUFI SH22
PUF I SHZ3
PUF I SH24
PROT21
0.13
PROT22
0.13
PROT23
0.13
PROT24
0.13
CALC21
0.14
CALC22
0.14
CALC23
0.14
CALC24
0.14
IRON21
0.01
IIWN22
0.01
I RON23
0.01
IRON24
0.01
THIMI21
0.02
THIW\\22
0.02
THIAM23
0.02
THIAM24
0.02
CAC21
-100.00
CAC22
-100.00
CAC23
-100.00
CAC24
-100.00
....
ID
OD

Table AL
Linear Programming Model - Activity Vectors (continued)
BBNC21
BBNC23
BBNC24
B"'-C21
B"'-C22
B"'-C23
B"'-C24
BCBC21
BCBC22
BCBC23
BCBC24
CSH21
-I .00
-1.00
-1.00
CSH22
-I .00
-1.00
CSH23
-1.00
-1.00
-1.00
CSH24
-1.00
-I .00
-1.00
BNQ21
1.00
0.01
0.01
j
BNCL23
1.00
1.00
0.01
0.01
0.01
0.01
0.01
0.01
BNQ24
1.00
1.00
1.00
0.01
0.01
0.01
0.01
0.01
0.01
0.01
0.01
"'-CA21
0.005
1.00
0.005
"'-002
0.005
1.00
1.00
0.005
0.005
"'-CA23
0.005
0.005
1.00
1.00
1.00
0.005
0.005
0.005
"'-004
0.005
0.005
0.005
1.00
1.00
1.00
1.00
0.005
0.005
0.005
0.005
CBCA21
0.005
0.005
1.00
CB 002
0.005
0.005
0.005
1.00
1.00
CBCA23
0.005
0.005
0.005
0.005
0.005
1.00
1.00
1.00
CBOO4
0.005
0.005
0.005
0.005
0.005
0.005
0.005
1.00
1.00
1.00
1.00
BN021
-1.00
BN023
-1.00
BN024
-1.00
"'-021
-1.00
"'-022
-1.00
"'-023
-1.00
"'-024
-1.00
W021
-1.00
CB022
-1.00
CB 023
-1.00
CB024
-1.00
LQRR21
-1.00
-1.00
-1.00
LQRR22
-1.00
-1 .00
LQRR23
-1.00
-1.00
-1.00
LQRR24
-1 .00
-1.00
-1 .00
W')(6NT2123
-1.00
W')(6NT2324
-1.00
W\\XMl T2122
-1.00
W'X"'- T2223
-1 .00
W'XMl T2324
-1.00
W'XCBT2122
-1.00
w\\XCBT2223
-1.00
W\\XCBT2324
-1.00
~
\\D
\\0

Table Al.
Linear Programming Model - Activity Vectors (continued)
RNBD2122
If3ND2123
If3ND2124
If3ND2IZ·
RBI'll2314
If3I'lJ23Z
ffil'lJ24Z
BND21
1.00
1.00
1.00
1.00
BND23
1.00
1.00
BND24
1.00
CAC22
1.015
CAC23
1.031
CAC24
1.061
-1.00
BNQ.23
-1.00
-1.00
BNCL24
-1.00
-1.00
-1.00
1ot.CA22
-0.005
Iot.CA23
-0.005
-0.005
1ot.CA24
-0.005
-0.005
-0.005
-0.005
ffiCA22
-0.005
ffiCA23
-0.005
-0.005
ffiCA24
-0.005
-0.005
-0.005
-0.005
LQRR22
1.00
lQRR23
1.00
lQRR24
1.00
1.00
OOJECT lYE
-1.113
-1.061
-1.031
N
o
o

Table Al.
Linear Programming Model - Activity Vectors (continued)
RN..02122
RH...02123
RH...02124
RR02223
RR02224
RN..022Z
RH...02324
RR023Z
RR024Z
N..021
1.00
1.00
1.00
R022
1.00
1.00
1.00
N..023
1.00
1.00
N..024
1.00
CAC22
1.083
CAC23
1.167
1.083
CAC24
1.333
1.250
1.167
BNQ23
-0.01
-0.01
-0.01
BNCL24
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
N..CA22
-1.00
N..CA23
-1 .00
-1.00
-1.00
N..CA24
-1.00
-1 .00
-1 .00
-1.00
-1.00
-1 .00
CBCA22
-0.01
CBCA23
-0.01
-0.01
-0.01
CBCA24
-0.01
-o.ot
-0.01
-0.01
-0.01
-0.01
LQRR22
1.00
LQRR23
1.00
1.00
LQRR24
1.00
1.00
1.00
OOJECT IVE
-1.417
-1.333
-1.167
N
o
~

Table Al.
Linear Programming Model - Activity Vectors (continued)
RCBD2122
RCBD2123
RCBD2124
RCBD2223
RCBD2224
RCllD22Z
RCBD2324
RCBD23Z
RCBD24Z
CllD21
1.00
1.00
1.00
CBD22
1.00
1.00
1.00 j
CBD23
1.00
1.00
CBD24
1.00
CAC22
1.027
CAC23
1.055
1.027
CAC24
1 .110
1.067
1.055
BNQ23
-0.01
-0.01
-0.01
BNCL24
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
H..CA22
-0.005
H..CA23
-0.005
-0.005
-0.005
H..CA24
-0.005
-0.005
-0.005
-0.005
-0.005
-0.005
CllCA22
-1.00
CBCA23
-1.00
-1.00
-1.00
CBCA24
-1.00
-1.00
-1.00
-1.00
-1.00
-1.00
LQRR22
1.00
LQRR23
1.00
1.00
LQRR24
1.00
1.00
1.00
OOJECTlVE
-1.137
-i .t ro
-1.055
N
.0
N

Table Al.
Linear Programming Model - Activity Vectors (continued)
ACSH210
ACSH2120
ACSH2140
ACSH2160
ACSH2180
ACSH220
ACSH2220
ACSH2240
ACSH2260
ACSH2280
,
CSH21
1.00
1.00
1.00
1.00
1.00
GAC21
1.00
0.80
0.60
0.40
0.20
ffi2120
-0.20
CR2140
-0.40
ffi2160
-0.60
CR2180
-0.80
CSH22
1.00
1.00
1.00
1.00
1.00
GAC22
1.00
0.80
0.60
0.40
0.20
ffi2220
-0.20
CR2240
-0.40
ffi226 0
-0.60
00280
-0.80
ACSH230
ACSH2320
ACSH2340
ACSH2360
ACSH2380
ACSH240
ACSH2420
ACSH2440
ACSH2460
ACSH2480
CSH23
1.00
1.00
1.00
1.00
1.00
CAC23
1.00
0.80
0.60
0.40
0.20
CR2320
-0.20
CR2340
-0.40
CR2360
-0.60
CR2380
-0.80
CSH24
1.00
1.00
1.00
1.00
1.00
CAC24
1.00
0.80
0.60
0.40
0.20
00420
-0.20
rn2440
-0.40
00460
-0.60
ffi2480
-0.80
to.)
o
w

Table Al.
Linear Programming Model - Activity Vectors (continued)
ARC210
ARC2120
ARC2140
ARC2160
ARC2180
ARC220
AK.C2220
AK.C2240
AK.C2260
ARC2280
i
K.a.21
1.00
1.00
1.00
1.00
1.00
K.CA21
-I .00
-0.80
-0.60
-0.40
-0.20
K.ffi2120
-0.20
K.ffi2140
-0.40
K.ffi2160
-0.60
K.ffi2180
-0.80
K.a.22
1.00
1.00
1.00
1.00
1.00
K.CA22
-1.00
-0.80
-0.60
-0.40
-0.20
K.ffi2220
-0.20
"1.00240
-0.40
K.ffi2260
-0.60
"1.00280
-0.80
AK.C230
AK.C2320
AK.C2340
AK.C2360
AK.C2380
AK.C240
AK.C2420
AK.C2440
AK.C2460
AK.C2480
K.Cl23
1.00
1.00
1.00
1.00
1.00
K.CA23
-1.00
-0.80
-0.60
-0.40
-0.20
K.CR2320
-0.20
K.ffi2340
-0.40
K.CR2360
-0.60
K.ffi2380
-0.80
K.C124
1.00
1.00
1.00
1.00
1.00
K.CA24
-1.00
-0.80
-0.60
-0.40
-0.20
"1.00420
-0.20
K.ffi2440
-0.40
"1.00460
-0.60
K.ffi2480
-0.80
!'.)
o
~

Table AI.
Linear Programming Model - Activity Vectors (continued)
AffiC210
ACBC2120
ACBC2140
ACBC2160
ACBC2180
ACBC220
ACBC2220
ACBC2240
AffiC2260
ACBC2280
j
CBa.21
1.00
1.00
1.00
1.00
1.00
CBCA2I
-1.00
-0.80
-0.60
-0.40
-0.20
CB<R2120
-0.20
CBCR2140
-0.40
CB<R2160
-0.60
CBCR2180
-0.80
CBa.22
1.00
1.00
1.00
1.00
1.00
CBCA22
-1.00
-0.80
-0.60
-0.40
-0.20
CB<R2220
-0.20
CBCR2240
-0.40
CB <R226 0
-0.60
CBCR2280
-0.80
ACBC230
ACBC2320
ACBC2340
ACBC2360
ACBC2380
ACBC240
ACBC2420
ACBC2440
ACBC2460
ACBC2480
CBCl23
1.00
1.00
1.00
1.00
1.00
CBCA23
-1.00
-0.80
-0.60
-0.40
-0.20
CBCR2320
-0.20
CBCR2340
-0.40
CBCR2360
-0.60
CBCR2380
-0.80
CBC124
1.00
1.00
1.00
1.00
1.00
CBCA24
-1.00
-0.80
-0.60
-0.40
-0.20
CBCR2420
-0.20
CBCR2440
-0.40
CBCR2460
-0.60
CBCR2480
-0.80
to-)
o
Ut

Table Al.
Linear Programming Model - Activity Vectors (continued)
VCSH2120
VCSH2140
VCSH2160
VCSH2180
VCSH21100
VCSH2220
VCSH2240
VCSH2260
VCSH2280
VCSH22100
j
CSH21
1.00
CR2120
1.00
ffi2140
1.00
CR2160
1.00
ffi2180
1.00
LQRR21
1.00
1.00
1.00
1.00
1.00
CSH22
1.00
CR2220
1.00
ffi2240
1.00
CR2260
1.00
ffi2280
1.00
LQRR22
1.00
1.00
1.00
1.00
1.00
<BJECTIVE
2.50
2.25
1.80
1.70
1.65
3.05
2.65
2.15
1.95
1.90
VCSH2320
VCSH2340
VCSH2360
VCSH2380
VCSH23100
VCSH2420
VCSH2440
VCSH246 0
VCSH248.0
VCSH24100
CSH23
1.00
CR2320
1.00
ffi2340
1.00
CR2360
1.00
ffi2380
1.00
LQRR23
1.00
1.00
1.00
1.00
1.00
CSH24
1.00
CR2420
1.00
ffi2440
1.00
CR2460
1.00
ffi2480
1.00
LQRR24
1.00
1.00
1.00
1.00
1.00
(EJECT IV E
3.10
2.70
2.20
2.00
1.95
3.00
2.55
2.10
1. 75
1.70
N
o
0\\

Table Al.
Linear Programming Model - Activity Vectors (continued)
VM..C2120
VM..C2140
VM..C2160
VM..C2180
VM..C21100
VM..C2220
VM..C2240
VM..C2260
VM..C2280
VM..C22100
M..Q.21
1.00
M..CR2120
1.00
M..CR2140
1.00
M.. CR2 I 60
1.00
M..CR2180
1.00
lQRR21
1.00
1.00
1.00
1.00
1.00
M..Q.22
1.00
M..CR2220
1.00
M..CR2240
1.00
M..CR2260
1.00
M..CR2260
1.00
lORR22
1.00
1.00
1.00
1.00
1.00
CBJECTlVE
1.10
0.80
0.45
0.35
0.30
1.20
0.90
0.65
0.55
0.45
VM..C2320
VM..C2340
VM..C2360
VM..C2380
VM..C23100
VM..C2420
VM..C2440
VM.. C246 0
VM..C2480
VM..C24100
M..Q.23
1.00
M..CR2320
1.00
M..ffi2340
1.00
M..CR2360
1.00
M..CR2380
1.00
lQRRZ3
1.00
1.00
1.00
1.00
1.00
M..Q.24
1.00
M..CR2420
1.00
M..ffi2440
1.00
M..CR2460
1.00
M..CR2480
1.00
lQRR24
1.00
1.00
1.00
1.00
1.00
CBJECTlVE
1.25
0.95
0.70
0.60
0.45
1.15
0.85
0.50
0.40
0.35
N
o
......

Table Al.
Linear Programming Model - Activity Vectors (continued)
VCBC2120
VCBC2140
VffiC2160
VCBC2180
VCBC21100
VCBC2220
VCBC2240
VCBC2260
VCBC2280
VCBC22100
CBQ21
1.00
CBCR2120
1.00
CBffi2140
1.00
ffiCR2160
1.00
CBffi2180
1.00
lQRR21
1.00
1.00
1.00
1.00
1.00
CBQ22
1.00
CBCR2220
1.00
CBffi2240
1.00
CBCR2260
1.00
CBffi2280
1.00
LQRR22
1.00
1.00
1.00
1.00
1.00
CBJECTIVE
1.05
0.75
0.40
0.30
0.25
1.15
0.85
0.60
0.50
0.40
VCBC2320
VCBC2340
VCBC2360
VCBC2380
VCBC23100
VCBC2420
VCBC2440
VCBC2460
VCBC2480
VCBC24100
CBQ23
1.00
CBCR2320
1.00
CBffi2340
1.00
CBCR2360
1.00
CBffi2380
1.00
lQRR23
1.00
1.00
1.00
1.00
1.00
ffiQ24
1.00
CBCR2420
1.00
CBffi2440
1.00
CBCR2460
1.00
CBffi2480
1.00
lQRR24
1.00
1.00
1.00
1.00
1.00
OOJECTlVE
1.20
0.90
0.65
0.55
0.40
1.10
0.80
0.45
0.35
0.30
N
o
co

Table Al.
Linear Programming Model - Row Identification Vector of
209
RHS Values
R~
RIGHT HAND SI DE
~r.eER
I DENT I Ft CAT 100
RRAT fON
VALUE
UN IT
Year 1
1
LANDAl
L
1.71
hectares
2
LANDSl
L
1.68
hectares
3
Fli'811
G
97.76
manaays
4
FLA812
L
56.23
manaay s
5
FLPa13
L
72.29
manaays
6
FLPa14
G
252.91
manaays
7
YAMlll
E
1996 .80
k i lograms
8
YAMI12
E
0
k I1 ograms
9
YAMI13
E
0
k i lograms
10
YAMI14
G
1996 .80
k I lograms
11
~Efll
E
181.96
kilograms
12
_ ~E112
E
0
k r lograms
13
~E113
E
0
kit ograms
14
~E114
G
181.96
k r 1ograms
15
RCE 111
E
184.28
kIlograms
16
RCE 112
E
0
k i lograms
17
RCE 113
E
0
k 1I ograms
18
RCE 114
G
184.28
kIf ograms
19
CASSlll
E
419.50
k 11 ograms
20
CASS 112
E
0
ki lograms
21
CASS 113
E
0
k 11 ograms
22
CASS 114
G
419.50
k i lograms
23
PROT11
G
16.80
kilograms
24
PROT12
G
16.80
kit ograms
25
PROT13
G
33.60
k i I ograms
26
PROT14
G
33.60
k i lograms
27
CALCll
G
211.43
grams
28
CALC12
G
211.43
grams
29
CALC13
G
422.87
grams
30
CALC14
G
422.87
fjrams
31
IRON11
G
11.54
grams
32
IRON12
G
11.54
grams
33
IRON13
G
23.09
grams
34
IRON14
G
23.09
grams
35
THIAM11
G
378.35
milligrams
36
THIAM12
G
378.35
mill igrams
37
THIAM13
G
756.71
millIgrams
38
THIAM14
G
756.71
ml I I Igrams
39
MAXRCEl
L
0
hectares
40
CSH11
E
1439557.90
CFAF
41
CSH12
E
0
CFAF
42
CSH13
E
0
CFAF
43
CSH14
E
1293800.00
CFAF
44
CACll
E
98876.67
CFAF
45
CAC12
E
109876.67
CFAF
46
CAC13
E
216102.49
CFAF
47
CAC14
E
230420.94
CFAF
48
CR1120
E
0
CFAF
49
CR1140
E
0
CFAF
50
CRl160
E
0
CFAF
51
CR" 80
E
0
CFAF
52
CRI 220
E
0
CFAF
53
CR1240
E
0
CFAF
54
CR1260
E
0
CFAF
55
CRI 280
E
0
CFAF
56
CR1320
E
0
CFAF
57
CR1340
E
0
CFAF
58
CRI360
E
0
CFAF

210
Table A2
(continued)
ROM
RIGHT HAND S I DE
ttJM3ER
IDENTIFICATION
RELATION
VAJ...UE
UNIT
59
rn1380
E
0
CFAF
60
CR1420
E
0
CFAF
61
CR1440
E
0
CFAF
62
CRI 460
E
0
CFAF
63
CR'480
E
0
CFAF
64
8NQ.ll
L
282294.18
CFAF
65
8NQ.13
L
179714.04
CFAF
66
8NQ.f4
L
333186.65
CFAF
67
MLQ.l1
L
17049.99
CFAF
68
MLQ.12
L
25575.00
CFAF
69
MLQ.13
L
42625.00
CFAF
70
MLQ.14
L
51149.99
CFAF
71
CSQ.ll
L
24513.78
CFAF
72
CSQ.12
L
36770.70
CFAF
73
-CSQ.13
L
61284.49
CFAF
74
CSQ.J 4
L
73541.37
CFAF
75
MLCA11
E
0
CFAF
76
MLCA12
E
0
CFAF
77
MLCA13
E
0
CFAF
78
MLCA14
E
0
CFAF
79
CSCA11
E
0
CFAF
80
CBCA12
E
0
CFAF
81
CBCA13
E
0
CFAF
82
CBCA14
E
0
CFAF
83
r-t.rn1120
E
0
CFAF
84
",-rn114O
E
0
CFAF
85
",-rn1160
E
0
CFAF
86
",-rn1180
E
0
CFAF
87
",-rn1220
E
0
CFAF
88
",-m 1240
E
0
CFAF
89
",-rn1260
E
0
CFAF
90
",-rn1280
E
0
CFAF
91
",-rn,320
E
0
CFAF
92
",-rn134O
E
0
CFAF
93
",-rn1360
E
0
CFAF
94
",-m 1380
E
0
CFAF
95
",-rn1420
E
0
CFAF
96
",-rn1440
E
0
CFAF
97
",-rnl460
E
0
CFAF
98
"'-CR1480
E
0
CFAF
99
CBrn1120
E
0
CFAF
100
cacsn ao
E
0
CFAF
101
CSCR1160
E
0
CFAF
102
csou rao
E
0
CFAF
103
CSCR1220
E
0
CFAF
104
CSCR1240
E
0
CFAF
105
cacnzso
E
0
CFAF
106
CSCR1280
E
0
CFAF
107
CSCR1320
E
0
CFAF
108
csrn1340
E
0
CFAF
109
CSCR1360
E
0
CFAF
110
CSCRI380
E
0
CFAF
111
csrn1420
E
0
CFAF
112
CSCR1440
E
0
CFAF
113
CSCRI460
E
0
CFAF
114
cscuaec
E
0
CFAF
115
8ND11
E
0
CFAF
116
8ND13
E
0
CFAF
117
8ND14
E
0
CFAF
118
"'-011
E
0
CFAF
119
"'-012
E
0
CFAF

211
Table A2
(continued)
R~
RIGHT HAND 5 I DE
f{JIoElER
IDENTIFICATION
RELATION
VALUE
UNIT
120
M...D13
E
0
CFAF
121
M...D14
E
0
CFAF
122
CSC11
E
0
CFAF
123
CSD12
E
0
CFAF
124
CSD13
E
0
CFAF
125
CSD14
E
0
CFAF
126
M'.XTBN1113
L
0
CFAF
127
M'.XTBN1314
L
0
CFAF
128
M'.XTBN1421
L
0
CFAF
129
M'.XTM...1112
L
0
CFAF
130
M'.XTM...1213
L
0
CFAF
131
M'.XTM...1314
L
0
CFAF
132
M'.XTM...1421
L
0
CFAF
133
M'.XTCB 1112
L
0
CFAF
134
M'.XTCB1213
L
0
CFAF
135"
M'.XTCB 1314
L
0
CFAF
136
M'.XTCB1421
L
0
CFAF
137
LQRR11
G
44310.54
CFAF
138
LQRR12
G
106955.60
CFAF
139
LQRR13
G
129841 .08
CFAF
140
LQRR14
G
61800.00
CFAF
Year 2
141
LANDA2
L
1.71
hectares
142
LANDB2
L
1.68
hectares
143
FLAB21
G
106.03
mandays
144
FU622
L
53.20
mandays
145
FL,oI623
L
64.86
mandays
146
FL,oI624
G
271 .50
mancevs
147
YAMI21
E
1996 .80
kilograms
148
YAMI22
E
0
k 11 ograms
149
YAMI23
E
0
k I1 ograms
150
YAMI24
G
1996 .80
kilograms
151
~E121
E
181.96
kilograms
152
~E122
E
0
kilograms
153
~E123
E
0
kilograms
154
~E124
G
181.96
kIlograms
155
RCEI21
E
184.28
kilograms
156
RCEI22
E
0
k 1I ograms
157
RCEI23
E
0
kilograms
158
RCE 124
G
184 .28
kilograms
159
CASSI21
E
419.50
kIlograms
160
CASSI22
E
0
kit ograms
161
CASSI23
E
0
k 1I ograms
162
CASSI24
G
419.50
k 11 ograms
163
PROT21
G
16.80
kilograms
164
PROT22
G
16.80
kilograms
165
PROT23
G
33.60
kilograms
166
PROT24
G
33.60
k 1I ograms
167
CALC21
G
211.43
grams
168
CALC22
G
211.43
grams
169
CALC23
G
422.87
grams
170
CALC24
G
422.87
grams
171
IR0N21
G
11.54
grams
172
IR0N22
G
11.54
grams
173
IRON23
G
23.09
grams
174
IR0N24
G
23.09
grams
175
TH IAM21
G
378.35
milligrams
176
THIAM22
G
378.35
milligrams
177
THIAM23
G
756.71
milligrams

Table A2
(continued)
212
R~
RIGHT HN'lD SI DE
rtJfooeER
IDENTIFICATION
RaATION
v A1.UE
UNIT
178
THJN424
G
756.71
mill rgrams
179
W,XRCE2
L
0
hectares
180
CSH21
E
0
CFAF
181
CSH22
E
0
CFAF
182
CSH23
E
0
CFAF
183
CSH24
E
1293800.00
CFAF
184
CAC21
E
98876.67
CFAF
185
CAC22
E
109876.67
CFAF
186
CAC23·
E
216102.49
CFAF
187
CAC24
E
230420.94
CFAF
188
CR2120
E
0
CFAF
189
CR2140
E
0
CFAF
190
CR2160
E
0
CFAF
191
CR2180
E
0
CFAF
192
CR2220
E
0
CFAF
193
GR2240
E
0
CFAF
194
CR2260
E
0
CFAF
195
CR2i80
E
0
CFAF
196
CR2320
E
0
CFAF
197
CR2340
E
0
CFAF
198
CR2360
E
0
CFAF
199
CR2380
E
0
CFAF
200
00420
E
0
CFAF
201
00440
E
0
CFAF
202
00460
E
0
CFAF
203
00380
E
0
CFAF
204
BNQ.21
L
313939.36
CFAF
205
BNQ.23
L
199859.92
CFAF
206
BNQ.24
L
370536.76
CFAF
207
14.Q.21
L
17432.25
CFAF
208
I4.Q22
L
26148.39
CFAF
209
I4.Q23
L
43580.65
CFAF
210
I4.Q24
L
52296.77
CFAF
211
CBQ21
L
25063.37
CFAF
212
CBQ22
L
37595.09
CFAF
213
CBQ23
L
62658.48
CFAF
214
CBQ24
L
75190.16
CFAF
215
14.001
E
0
CFAF
216
14.002
E
0
CFAF
217
14.CA23
E
0
CFAF
218
14.004
E
0
CFAF
219
CBOO1
E
0
CFAF
220
CB 002
E
0
CFAF
221
CBCA23
E
0
CFAF
222
CB 004
E
0
CFAF
223
14.00120
E
0
CFAF
224
14.00140
E
0
CFAF
225
14.CR2160
E
0
CFAF
226
14.00180
E
0
CFAF
227
14.CR2220
E
0
CFAF
228
14.CR2240
E
0
CFAF
229
14.CR226 0
E
0
CFAF
230
14.CR2280
E
0
CFAF
231
14.CR2320
E
0
CFAF
232
14.CR2340
E
0
CFAF
233
14. CR236 0
E
0
CFAF
234
14.CR2380
E
0
CFAF
235
14.CR2420
E
0
CFAF
236
14.CR2440
E
0
CFAF
237
14.CR2460
E
0
CFAF
238
14.CR2480
E
0
CFAF

213
Table A2
(continued)
RCW
RIGHT HAND 5 I DE
f>UIoeER
ID8HIFICATI~
RELAT ION
VALUE
UNIT
239
caCR2120
E
0
CFAF
240
caOO140
E
0
CFAF
241
caOO160
E
0
CFAF
242
caOO180
E
0
CFAF
243
cs00220
E
0
CFAF
244
caOO24O
E
0
CFAF
245
cs0026 0
E
0
CFAF
246
caOO280
E
0
CFAF
247
caOO320
E
0
CFAF
248
caCR2340
E
0
CFAF
249
cs 0036 0
E
0
CFAF
250
caCR2380
E
0
CFAF
251
cs 00420
E
0
CFAF
252
cs 00440
E
0
CFAF
253
ea0046 0
E
0
CFAF
254
-cs 00480
E
0
CFAF
255
BND21
E
0
CFAF
256
BND23
E
0
CFAF
257
BND24
E
0
CFAF
258
M..D21
E
0
CFAF
259
M.. 022
E
0
CFAF
260
M..023
E
0
CFAF
261
M..024
E
0
CFAF
262
caD21
E
0
CFAF
263
ca022
E
0
CFAF
264
ca023
E
0
CFAF
265
ca024
E
0
CFAF
266
l-'AXTB ti21 23
L
0
CFAF
267
l-'AXTB ti2324
L
0
CFAF
268
fo'AXTM..2122
L
0
CFAF
269
fo'AXT M.. 2223
L
0
CFAF
270
fo'AXTM..2324
L
0
CFAF
271
fo'AXTca2122
L
0
CFAF
272
fo'AXT cs2223
L
0
CFAF
273
fo'AXTca2324
L
0
CFAF
274
LQRR21
G
44033.72
CFAF
275
LQRR22
G
106603.57
CFAF
276
LQRR23
G
129267.43
CFAF
277
LQRR24.
G
61800.00
CFAF
278
BNRACC21
E
0
CFAF
279
BNRACC23
E
0
CFAF
280
BNRACC24
E
0
CFAF

214
VITA
Bernadette Dia was born on March 21. 1955 in Agboville.
Ivory Coast
where she spent the early years of her childhood.
She graduated from Sainte Marie High School in Abidjan. in June.
1973.
She entered the University of Abidj an the same year and obtained
the DUES I and 11 in Agronomy in 1975 and 1976. respectively.
In October. 1976. she was admitted at Ecole Nationale Superieure
Agronomique (E~N. S.A.) d' Abidj an where she received the Diplome d' Agro-
nomie Generale (D.A.G.) in June. 1978.
The same year. she was awarded an
AFGRAD fellowship to pursue studies at the Universi ty of Connecticut.
Storrs.
In December 1980. she completed the requirements for the degree
of Master of Science in Agricultural Economics.
From January to December of 1981 she held the position of Assistante
(Instructor) at E.N.S.A.. teaching Macroeconomics and Mathematical Pro-
gramming.
In January. 1982.
she began work on the Ph.D. degree in the De-
partment of Agricultural Economics at the University of Illinois. Urbana-
Cbampaig~
Course work and preliminary examination requirements for the
degree were completed in March. 1984.
She was elected to Gamma Sigma
Delta. the honor society of agriculture. and to the honor society of Phi
Kappa Phi. in April and November of 1984. respectively.