CRDP SUPPLY RESFWSES UWER UMCERTAINTY IN TU0
SENEGALESE RE6IONS: A CWMTIVE
STUDY
Mamadou Sidibe
A PLAM B PAPER
Submitted to
Michigan State University
in partial
fulfillment
of the requirements
for the degree of
MSTER OF SCIENCE
Department of Agricultural
Economies
1987
-.
_,__. _. .-
-..^.. -...
_
__.
--

---

ANSTRACT
CROP St8PPl.Y RESWES UHDER WCERTAINI-Y
IN TU0 SEHEGALESE REGIONS: A COIIPaRATIVE
!STUDY
Mamadou Sidibe
Farm resource allocation
decision,
particularly
in developing
countiies,
is generally
a risky process. The uncertainty
related to
yield$ and farm product prices suggests that farmers behave in
risk-averse
ways.
Tbis research uses linear programming models under a risk
framefork to investigate
about cropping patterns and technologies
most
profitable
to farmers in two senegalese zones: the Center of the
Pea"u{ Basin and the Upper Casamance.
Normative supply responses are derived for the two agricultural
zones'investigated
under a food security
perspective
by making
assumptions about farm price levels.
In the Central Peanut Basin,
acreage under cultivation
does not respond to price increases;
supply
responses have constant sfopes. In the Upper Casamance zone,
interesting
aspects of land competition
between crops is found. Among
a11 ciops, maize showed higher acreage responses to price increase.

---

S@cia1 thanks go to Dr. Eric W. Crawford my research supervisor,
who patiently
reviewed several drafts and remained steadfast
in his
support.
Thanks also go to Dr. Lester V. Manderscheid my major
profesor and Dr. Ger Schultink
for their helpful
comments. Dr. Stephen
Harsh,deserves
special mention for his helpful
guidance with the
arttficial
intelligence
materiel.
Special thanks also to Frederic
Martin for a11 the support he provided for the realization
of this
work, ;Pam Starr for her valuable assistance during the editing
phase
and a11 the staff of the food security
project
in the Department of
Agricultural
Economies. Im am indebted to the United States Agency for
International
Development (USAID) for its financial
support throughout
my grdduate worki
Fiinally,
1 want to thank my family and the very special friends
who h$ped make my experience at MSlJ both personally
and
profejsionally
rewarding.
1-
--
--

---

.*.....................*...............
iv
l . . . . . . . . . . . . . . . . ..*.......*.**......*.
vii
I-
ipUtOfMTION
1.1 Background
..................
1.2 Justification
..................
:
1.3 Objectives
1.4
..................
Uses
..................
i
4.5 Methodology
4
..................
II- $WULNfMl SYSTEH
$1 Farming system description
. ..*............*.
.l.l Special features
. ..*.......*......
1 .1.2 Cropping patterns
. . ...*.....*...*..
il.2 Zone overview
..................
2.2.1 Description
of the regions
..................
2.2.2 Zone identification
..................
III- L/INEAR
P~~i~
HDDEL
3.1 Typical farm structure
..................
3.1.1 .Land and population size
..................
31.1.2 Cropping pattern
..................
3.1.3 Technical packages
...........
.1
.....
3.1.4 Cropping calendars
..................
31.2 Theoretical
considerations
. . ...*.........**.
q.2.1 States of nature
.*.+..*..*....*.,.
31.2.2 Mathematical
mode1
. ...*.*.*...*..*..
3.3 Mode1 activities
..................
3j.3.1 Crop producing activities
..................
$3.2 Input procurement activities..................
31.3.3 Output selling
activities
l
.
..*..*.*.*.*..**
31.3.4 Cereal buying activities
. . . . . . . . . . . . . . . . . .
31.3.5 Capital transfer
activities..................
31.3.6 Risk transfer
activities
. . . . . . . . . . . . . . . . . .
31.3.7 Dummy activities
.*................
ii
-.
-

---

3 ,4 Objective function values ............ . .....
.4.1 Production coefficients
..................
3.4.2 input coefficients
..................
3.4.3 Cereal transaction
prices
..................
3.4.4 Other coefficients
..................
3.5 Mode1 constraints
..................
3.5.1 Resource use constraints
..................
3.5.2 Chemical imput constraints
..................
3.5.3 Food security
constraints
..................
'3.5.4 Other constraints
..................
IV- pROBLEM SOlUTION AND SENSITIVITY ANALYSIS
4.1 Base run
..................
4.1.1 Cropping intensities
..................
4.1.2 Scarcity
values
..................
4.1.3 Mode1 validation
..................
4.2 Sensitivity
analysis
..................
4.2.1 Resource range variations
..................
4.2.2 Objective
function
ranges
..................
4.3 Knowledge representation
..................
4.3.1 Expert system components
..................
4.3.2 Knowledge base
..................
V-
@CffSSION OF OBJECTIVES AMD SUPPLY RESPDNSES
3.1 Discussion of objectives
..................
5.1.1 Starting
capital
..................
q.1.2 Marginal lands
..................
5.1.3 Impact of population
growth ..................
5.1.4 Food self-sufficiency
rates ..................
51.2 Supply responses
..................
/.2.1 Price assumptions
..................
a.2.2 Normative supply curves
..................
51.3 Expert system functions
..................
q.3.1 Design considerations
..................
5i.3.2 Functions
..................
VI- qMCtus1oxs
6i.l Summary of findings
. . ..**...........*
61.2 Recommendations
. . . . . ..*..........
61.3 Areas for further
research
.,....,**..*..*..*
................................................
...........................................
i ii
----
-*

---

LIST UF TABLES
TABLE;
PAGE
2.1.1
Calculated Area by Major Crops in Senegal
10
2.1.2,
Land Share of Major Crops in Senegal (1970-1983)
12
2.1.3:
Land Share of Food-crops and Cash-crops
12
2.2.1
Population Structure
for 1987
18
2.2.2 1
Land Utilitation
18
2.2.3
Percentage Area Cultivated
for Major Crops in
Peanut Basin and Casamance Regions
20
3.1.1,
Farm Size and Population Composition
25 ’
3.1.2 j
Expected Yields in Central Peanut Basin
25
3.1.3,
Expected Yields in Upper Casamance
25
3.2.1
Definition
of Rainfall
Amount Categories
30
3.2.2
Definition
of Rainfall
Distribution
Categories
30
3.2.3 ’
Expected Yields and States of Nature
32
3.3.1
Mode1 Activities
in Central Peanut Basin
35
3.3.2 <
Mode1 Activities
in Upper Casamance
37
3.5.1 1
Mode1 Constraints
in Central Peanut Basin
41
3.5.2
Mode1 Constraints
in Upper Casamance
42
3.5.3
Structure
of the Food Self-Sufficiency
Constraints
for a Hypothetical
Zone
45
4.1.1
Optimal Plan and Cropping Intensity in Central
Peanut Basin
48
4.1.2
Optimal Real Activities
in Central Peanut Basin
48
4.1.3 :
Optimal Plan and Cropping Intensity in Upper
Casamance
50

---

4.1.4'
Optimal Real Activities
in Upper Casamance~
50
4.1.5
Scarcity
Values of Binding Constraints
in Central
Peanut Basin
53
4.1.6
Scarcity
Values of Binding Constraints
in Upper
Casamance
53
4.1.7
Observed and Calculated Land Shares in Central
Peanut Basin
55
4.1.8:
Observed and Calculated
Land Shares in Upper
Casamance
55
4.2.1
Resource Range Variations
in Central Peanut
Basin
57
4.2.2
Resource Range Variations
in Upper Casamance
57
4.2.3
Objective
Functfon Range in Central Peanut Basin
59
4.2.4
Objective
Function Range in Upper Casamance
60
4.3.1
Variable Attributs
and Optimal LP Tableau
61
4.3.2
Knowledge Base Data Definition
64
5.1.1
Cropping Patterns and Starting
Capital
in
Central Peanut Basin
66
5.1.2
Cropping Patterns and Starting
Capital
in Upper
Casamance
67
5.1.3 j
Cropping Patterns and Marginal Lands in Central
Peanut Dasin
68
5.1.4
Cropping Patterns and Marginal Lands in Upper
Casamance
69
5.1.5
Cropping Patterns and Population Growth in
Central Peanut Basin
71
5.1.6
Cropping Patterns and Population Growth in
Upper Casamance
71
5.1.7
Net Farm Return at Different
Defined Food
Self-Sufficiency
Levels
73
5.2.1
Crop Price Vector in Central Peanut Basin
75
5.2.2
Crop Price Vector in Upper Casamance
75
V
----
.*

---

5.2.3
Price Levels and Cropping Patterns in
Central Peanut Basin
77
5.2.4
Price Levels and Cropping Patterns in
Upper Casamance
78
5.2.5
Supply Responses and Price Effect in Upper
80
Casamance
5.3.1
Expert System Sample Printout
82
./
APPEN@X TABLES
1 :
88
3
10;
4
104
Vi

---

LIS1 OF FIGURES
FIGURt
PAGE
2.1.1
Average Rainfall
in Senegal (1970-1983)
9
2.1.2’
Cultivated
Area and Officia1
Prices of Millet
14
2.1.3
Cultivated
Area and Officia1
Prices of Maize
15
2.1.4
Cultivated
Area and Officia1
Prices of Groundnut
16
2.2.2'
Distribution
of Major Crops by Region
21
3.1.1
Crop Calendar in Central Peanut Basin
27
3.1.2
Crop Calendar in Upper Casamance
28
4.3.1
Components of an Expert System
62
vii.

---

CHAPTER
I
IWTIUH
14 1 bckground
The extent to which farmers in less developed countries
(LDCs)
respond to price changes in agricultural
products has been extensively
debated in recent years. Attempts to estimate trop supply responses to
price 'change have reached'divergent
conclusions.
For example, Hopper
(1965)i, in a study carried out in India, provide empirical
evidence to
support the theory
that farmers in LDCs are remarkably efficient
in
allocdting
the resources at their disposal.
This theory, referred
to
as the 'poor but efficient
' hypothesis,
assumes that farmers in LDCs
are profit
maximizers.
And Yotopoulos and Nugent (1976), following
the
same framework, concluded more specifically
that Indian farmers seem
to be Iremarkably price efficient.
Aljthough those two studies lend support to the "poor but
effici@"
hypothesis,
risk aversion attitudes
typical
to farmers in
subsisience
agriculture
are not included in their framework. Subrata
*
and Kein (1984) argue that peasant farmers operate in an environment
where considerable
uncertainties
exist and where institutional
and
culturpl
constraints
are important.
1
.
----

---

2
Hen@, risk and uncertainty
should be taken into account in any
attempt to analyze farmers' resource allocation
process. Gotsch and
Falcon: (1975), in a study initiated
in the Punjab region of India,
emphasired the fact that farmers are more responsive to a farm-level
net re!venue change than to a relative
price change. The un ertainty
relate/
to meeting
istence
subs
requirements cari offset any price
incentiive.
Thiis research takes risk aversion into account to mode farmers'
behavibr in two Senegalese regions with regard to farm resource
alloca#ion
plans. In the Senegalese farming system context,
risks
confrobting
farmers are related to yield for crops produced for home
consumption and income for crops produced for sale.
l.I(i Justification
Traditionally,
Senegalese agricultural
policies
have promoted food
self-sbfficiency
by focusing on issues such as the expansion of rural
creditj,
rural cooperatives,
and the efficiency
of development
agenciks. This supply side orientation
was largely
based on the
countrk's
comparative advantage in producing cash-crops (groundnut,
cotton$ for export, and on the import of cereals (rice,
wheat) to
supporb urban consumption. Locally produced cereals were used to
secure; the food needs of rural zones.
Rebently, new preoccupations
have emerged. The New Agricultural
Policy; (NAP) emphasizes the limitation
of state intervention
in the
rural economy, the promotion of cereal production
to achieve higher
-- __._ - _.. .__.. -__

---

,
3
food self-sufficiency
rates and the expansion of the peanut processing
industry
to increase exports of peanut products.
In sum, the NAP is
the Senegalese Government's attempt to add a food security
perspective
to previously
more narrow food self-sufficiency
objectives.
Indeed,
the ajteration
of existing
cropping patterns seems necessary to
recondile
both food self-sufficiency
goals and food security
concerns.
But the NAP remains silent
on a crucial
point: better price policies
are a: prerequisite
to the achievement of a better trop mix.
This research is an attempt to measure the impact of government
price \\policy on farmers' resource allocation.
It is part of a larger
agricultural
sector simulation
mode1 developed at Michigan State
University
(MSU) by Martin and Crawford (Martin,
1986a; Martin and
Crawfjrd,
1987).
143 Objectives
Tbis work is based on a set of regional models of the Senegalese
agricultural
sector.
It focuses on the Central Peanut Basin and the
Upper jcasamance zones. The objectives
pursued are to:
1): derive normative supply curves for several crops in each of the
agriclltural
zone considered above, given a trop price vector.
2); determine the impact on farm resource allocation
plans of
requiring
increased cereal food self-sufficiency
rates during bad
rainfaill
years.
3): measure the effects
of population
growth on prevailing
cropping
patterbs.
4)' investigate
the change in cropping patterns when marginal lands
are cujltivated.

---

4
5) study the relationship
between the starting
capital
owned by
farmers and their use of input intensive
technologies.
6) develop a simple expert system to facilitate
interpretation
of
optimdl farm plans derived from the linear programming models.
1;4 Anticipated
uses
The results
derived from this study are primarily
intended to be
used as a diagnostic
tool to aid the regional planning of agriculture
in Sertegal. Several policy alternatives
cari be tested for their
shortjterm
effects
at the regional level.
Knowledge obtained from
produder behavior in response to farm price increases may help
identlfy
where regional comparative advantage lies.
The identification
of binding constraints
creating bottlenecks
for regional development
may clarify
where efforts
should be put in the future.
It should
however be pointed out that mode1 results
are not a Perfect
repredentation
of reality.
The basic assumption underlying
this work is that Senegalese
farmers, as in most less developped countries
(LDCs) try to maximize
expected net revenue subject to ensuring.that
their subsistence
needs
are met under the most adverse state of nature.
This decision
rule is
stated by Low (1974) as the minimum cost of security
criterion.
A typical
farm is modeled in each of the two regions by using a
linear programming approach. Production activities
are broken down
into Pood crops produced for household consumption and food and cash
crops Iproduced for sale. Fifteen states of nature are identified
based

---

5
on two criteria:
the total
annual quantity
of rain and its
distribution
within the year. Uncertainty
is included in the mode1 by
usingtthe minimization
of total absolute deviations
(MOTAD) procedure
as suggested by Hazell and Norton (1986). For the home consumed crops,
deviations
from the average yield expressed in terms of calories
are
calculated
under each state of nature. The same procedure is used for
deviations
from mean income for production
activities
used for sale.
The states of nature showing the highest deviations
from average
yiefdg (in calorie)
for the home consumed crops and average income for
crops Iproduced for sale are included in the constraint
rows of the LP
tableau.
Purchases of farm inputs are addressed by input procurement
activi$ies
(transfer
activities).
The capital
constraint
is divided in
two rows: one where borrowing is allowed and the second where
borro$ing is not allowed. The mode1 also takes into account food
habits prevailing
in each zone studied.
Supply responses are derived by making various assumptions about
price levels for the different
trop producing activities.
Six price
levels are assumed for each cereal produced for sale and for home
consumption, starting
from their weighted mean prices over the fifteen
states of nature incremented by 20% up to 100%. The mode1 is run with
each price vector and supply responses are derived with regard to land
allocqtion
under alternative
technologies.
T-e impact of increasing
food self-sufficiency
rates during bad
rainfall
years is determined by increasing
the initial
rate(30%)
through increments of 10% up to a point of infeasible
solution.
The
procesis requires the calculation
of a new technical
coefficient
at
each step, reflecting
the desire of meeting 80% of total needs in good

---

rainfall
years in association
with the worst years' rates mentioned
above.: This coefficient
is calculated
by taking the ratio bad rainfall
and good rainfall
food self-sufficiency
rates.
It is entered into the
mode1 'in the food self-sufficiency
constraint
under an activity
expreqsing the risk of not meeting food needs.
A:fam
population
increase affects
both consumption and production
sides .of the farm model. From the consumption side, an increase in
household size increases total food needs expressed in terms of
calories
and increases the household food requirements
to be satisfied
by home consumed crops. From the production side, the amount of family
labor iavailable
for field work augments. For example, a Child is
consibered an active agricultural
worker when he enters the age group
startiing
from eight years of age. The mode1 is run with these knew
paramdters.
Ijtroducing
marginal lands to expand the capacity of the farm to
producje more food when producer prices increase involves broadening
the mo/del to include new cereal producing activities
for sale. The LP
is runi with these modifications
and with a new land constraint.
The impact of the starting
capital
is studied by running the mode1
with a: knew startfng
capital
level.
The results
given by the base run
solutibn
are used as guides to formulate assumptions concerning
realisitic
starting
capital
levels.
The knowledge base of the expert system is developed in terms of
relations
(predicates)
and rules. The relations
represent the factual
information
given by the LP optimal solution.
The rules are based on
the interpretation
of the optimal plan parameters. The fifth
generation programtning language Prolog is used to implement the

---

7
systeq. Due to time constraints,
the application
is carried out only
for ttje Central Peanut Basin.
TQe microcomputer package LP88 is used to salve a11 the LP
probldms. The trop budgets set up by Martin (1987a) are the main
source of information
for the technical
coefficients
used in the LP
table+.

---

UWTER II
THE A§ftICULl’URAl SYSTM IR SEREGAL
This chapter contains a description
of the Senegalese agricultural
system. Special features of the farfning system are presented followed
by a description
of cropping patterns at the national
level.
The
rationale
behind the choice of the two regions used in this study is
discujsed.
2.!1 Fanning system description
The purpose of this section is to highlight
special features of
the Senegalese farming system. The prevailing
cropping patterns
are
analyaed from a historical
point of view. The data set used covers the
years il970 to 1983.
2.4.1 Soecial features
Thie Senegalese concept of the farm is based upon the
farm-hbusehold unit. According to Cattin and Faye (see Cattin and
Faye, 1982), the farm-household generally
follows
two levels of
organi?ation.
First,
at the "concession"
level (compound), the farm
unit i:s composed of several households which are under
8

---

9

---

TFwt3LE
221-1-X
r
C3U~T~LVRTEOFIR~~erNRJOR
CROP
c
oc10
Ha,
----------------------------------------------------------------------------------------------------
,. “,.l.
.l!lillu.t
ws.
.“.”
“..
.-
Sorghun
Maire
Paddy
W-s
nutr
Cotton
TOTRL
-------------------------------------------------------------------------------------------------
RrGta
%
Area
%
Rrua
%
Rrea
%
Rrua
%
fhaa
2
-------------------------------------
--------------------------------------------------------------
1’370
‘367
43.2
51
2.3
94
4.2
63
2.8
1049
46.8
14
0.6
27-37
1971
973
43.1
49
2.2
85
3.8
71
3.1
1060
46.9
1B
0.8
2257
1972
936
42.5
32
1.5
54
2.5
86
3.9
1071
48.6
20
0.9
22clu
1973
1103
47.8
35
1.5
fi5
2.8
53
2.3
1025
44.3
2El
1.2
2303
1974
1145
47.1
43
2.0
59
2.4
1052
43.3
30
1.6
2428
1975
965
38.3
58
2.0
94
3:;
62
2.4
1312
52.n
39
1.6
2521
1976
949
3A.
1
89
3.6
1295
52.0
44
1.8
2489
1’377
Y43
40.5
El
2::
63
2.7
5:
25
1161
49.9
47
2.0
2325
lY7B
1OSS
42.7
56
2.3
91
3.7
62
2.5
1154
46.7
48
2.0
2468
1979
969
43.u
6El
3.u
79
3.5
55
2.5
1048
46.6
31
1.4
2248
1980
1117
46.3
7t3
3.2
67
2.8
54
2.2
1066
44.1
30
1.2
2412
1’381
1184
48.4
78
3.2
76
3.1
68
2.8
10lQ
41.2
32
1.3
2449
1982
991
41.6
81
3.6
68
2. Y
48
2.0
1149
48.1
42
1.8
2384
1983
628
38.9
-..-.....- _.-.-.----.*-
-- ------
-----
----
---21-_1:3__-.--_-S”-_--4_______-2__2-9
-_-_
-‘oR1-Sn:U_.-_-___33__IL-
-_._--
f!;-“-
MEAN
1009
43.
Cl
5R
2.5
76
3. 2
62
2.6
1llU
47.3
33
1.4
2347
50
94
3. 3
16
0.7
14
0.5
9
0.5
91
3.1
10
0.4
116
MIN
828
38.1
32
1.S
52
2.4
4R
2.0
101R
41.3
14
0.6
2126
rmli
1184
4E3.4
8G
3.6
94
4.2
86
3.9
1312
52.0
48
“*SI
2521
SlYJt-CX?
:
Nnuvel
le
Pol
itique
flgr-ic:ale,
Uit-mzkion
Genet-ale
de
1s
Product.iat~
Hgricole.
armexes
p rj-lB,
1964.
Stat
i St-. i cs
~18 Icu
1 atud
by
the
authut-

---

11
the direction
of the "chéf de concession"
(head of compound). The
latter,
often the oldest family member, is responsible
forproduction
and cqnsumption decision making processes. Second, the
"farm;concession"
may contain more than one "ménage" (household
sharing the same meals). Decision making processes regarding
produc$ion and consumption are held by the "chef de ménage", hence,
the 'ménage" represents the basic production
and consumption unit,
although it may depends on the "chef de concession" for access to
anima1 traction
equipment.
Ttie association
between climate and the farming system ts
espectally
important in Senegal and should be discussed here. The
marked difference
in rainfall
between Northern and Southern Senegal is
a critical
factor in explaining
differences
in trop calendars between
regions. For example, in Northern Senegal, annual rainfall
is 300 mm,
and in Southern Senegal, it is an average of 1400 mm (Abt Associates,
Inc, p. 26, 1985). Also, from 1970 to 1983, the country experienced
periods of heavy drought particularly
during the years 1972, 1977 and
1983. 'Graph 2.1.1 shows the average rainfall
levels for the years
mentiqned above.
2.J.2 Croooing gatterns
Taible 2.1.1 shows the area cultivated
in the major crops in
Senegal. Several points of interest
are worth mentioning.
Groundnut
and mi?let/sorghum
have the largest shares of total
cultivated
land,
47.3 and 43 % respectively,
during the 1970-83 period. Although
groundhut has a larger share than millet
on the average, the
calculbted
percentages of table 2.1.1 do show a greater share of land
in milnet during the two years following
a period of drought.

---

12
TABLE 2.1.2 : LAND %ARE OF IdA3oR CROPS
IN SENEGAL
Average froc 1870 to 1983
Rice
Ground
Millet
Maize
wW
Cowpeas
nuts
Cotton
Total
43.0
2.5
3.2
2.6
47.3
1.4
100.0
Souvice : Data taken from "Nouvelle Politique
Agricole",
Direction
Generale de la Production Agricole,
annexes p. 6-18, 1984.
TABLE 2.1.3
: LAN0 %ARE OF FOOMROPS ANfl
CASH-Cm IN %M%AL
Average frm 1970 to 1983
Food-crops
Cash-crops
Total
5hares
51.3
48.7
100.0
Source : Data taken from "Nouvelle Politique
Agricole",
Direction
Generale de la Production Agricole,
annexes p. 6-18, 1984.

---

13
It seéms that in years of good rainfall,
groundnuts are grown
predominantly
for income purposes. In years following
poor rainfall,
millet
is grown as an assurance that home consumption needs are met,
first,
Rice occupies about 3 % of cultivated
lands; however, the
calculated
percentages show a decreasing trend ranging from 4.2% in
1970 to 2.4% during 1983. Maize exhibits
a positive
trend; its average
acreage is 3% during the last five years as opposed to about 2% during
1970 fo 1976. Cotton and cowpeas show fairly
constant trends averaging
about 1.4 and 2.6 % of cultivated
land. An illustration
of the
respective
shares of these different
crops is given in table 2.1.2.
Table 2.1.3 illustrates
the different
shares of land of the food-crops
as opiosed to the cash-crops.
Seiveral factors are usually associated with the within-year
variations
of land allocated
to various crops. Among other factors,
producier prices contribute
in a significant
way to explain those
variations.
Amadou Niane (MSU, 1980) used an econometric mode1 to show
that the supply of millet/sorghum
in Senegal is responsive to the
previaus year's officia1
producer price. Although there is evidence
that oinly a small part of the millet/sorghum
produced is marketed
officially
(4 to 5%) (see Ndoye, 1984), officia1
millet
prices may
influejnce farmers'
land allocation
decisions
for millet
thereby
effecting
millet
supply. For example, in the case of groundnuts, the
impact of government groundnut pricing
policies
on land allocation
was
noticeiable during the years 1967 to 1974 (Abt Associates,
1985).
Graphs 2.1.2 to 2.1.4 depict the relationships
between land allocated
to millet,
maize and groundnut with respect to their officia1
prices
from 1970 to 1983. Graph 2.1.2 shows that an increase in millet
price

---

1
14

---

__
_
__
__
^.
.
-
:
.
--
15
-
_..
I 1
I i t i i
1
b 4”
.._-.-
--.--
J

---

4 CIF! lLl Q
< Z cl
d
c3 ci! 0 3 Z cl 7
8) do m
M
--. t I
F? -.
16
il

---

17
is followed by an upward shift
in millet
acreage. This is particularly
visible
during years 1973, 1974 and 1981. Maire acreages respond to a
larger extent to price increases than millet;
this fact is
partiaularly
noticeable
in graph 2.1.3. Groundnut acreages were
responsive to price increases up to 1976. The high prices observed
after ,198O are not followed by acreage increases.
Farmer responsiveness to officia1
price changes appears to be
effective
during the observed years. However, that effect
does differ
depending on the type of trop. While land allocated
to maize shows a
positive
trend during the observed years, groundnuts seem to be losing
popularity
among farmers.
2,2 Zone overviw
This section provides a description
of the relative
importance of
the two regions involved in this work: the Peanut Basin and Casamance.
Comparisons are made between those regions and the rest of the country
in terms of population
structure,
land utilization
and production
patterns.
The rationale
behind the choice of an agricultural
zone
within each region is then discussed.
2.2.1 Description
gf the recrions
Table 2.2.1 summmarizes the population
structure
in each region
and the rest of the country. About 46% of the Senegalese population
and 54% of its rural population
live in the Peanut Basin, Casamance
shelters
14% of the total
population
and accounts for 16% of the rural
population.
-1
--

---

18
TABLE 2.2.1
: POfWATIoW STRUCTURE
(0-W individuah)
Regilons
Total
%
Urban %
Rural
99
Peanut
Basin
2912 (46)
521 (28)
2391 (54)
Casamaince
860 (14)
150 (8)
710 (16)
Senegal
6300 (100)
1890 ( 100)
4410 (100)
Sources : RAPID II population
projection
for 1987
Percentages are calculated
by the author.
TAME 2.2.2
: LAMI UTILfZATICM
uni ts = Ooo Ha
Total
Arable
Crop
Land use
Regions
land %
‘land %
land %
rate
Peanut;
Basin
4442 (21)
2169 (59)
1749 (79)
81
Casamahce
2835 (14)
750 (20)
297 (13)
40
Senega?
20000 (100) 3700 (100) 2220 (100)
60
Sourtes : Data taken from "Nouvelle Politique
Agricole",
Direction
General@ de la Production Agricole,
annexes p. 6-18, 1984.
Percentages are calculated
by the author. Land use
rate is calculated
by dividing
arable land into trop
land and multiplying
the quotient
by 100.

---

19
The land distribution
prevailing
in 1984 is shown in table 2.2.2.
It is interesting
to note that in the Peanut Basin, 81% of the arable
land i:s in use; that figure is far above the 40% land use rate
calculjated
for Casamance.
The cropping patterns existing
in each region are given in Table
2.2.3: Both groundnut and millet/sorghum
are particularly
popular in
the Peanut Basin; each of them uses about 80% of cultivated
land. Rice
and catton are mainly cultivated
in Casamance; their shares in total
cropland used are respectively
of 75 and 45%. Maize is a more
important trop in Casamance than in the Peanut Basin (35% of the total
area cultivated
to maize in Senegal is in the Casamance versus 17% in
the Pelanut Basin). 43% of the land allocated
to cowpeas is in the
Peanut Basin.
2.2.2 &tg
identification
The tables listed
above illustrate
the importance of the
Peanut
Basin. The reasons for its selection
as an area of focus for this
study may be summariied as follows:
- More than 50% of the Senegalese rural population
and 28% of its
urban population
live there.
- More than 80% of the total
area cultivated
to both groundnuts and
millet
in Senegal is in the Peanut Basin.
It should, however, be pointed out that the Peanut Basin shows a
great heterogeneity
in terms of rainfall,
soi1 types and ethnie
groups. Based on those criteria,
four agricultural
zones are
identiified
in this region (Martin,
1987a). This study Will focus on
the Center of the Peanut Basin (CPB) which, in terms of area, covers

---

20
TABLE 2.2.3
: PERCENTME ARfA CULTIVATED
FUR
CRWS
Rice
Ground
Regiorns
Millet
Maire
paddy
Cowpeas
nuts
Cotton
Peanuti
Basin
81.4
17.1
2.5
43.0
86.9
16.8
Casamrynce
9.6
35.4
75.5
2.0
9.9
45.0
Others
9.0* 47.5
22,o
55.0
3.2
38.2
-a-
100 ----
100
Ïoo
-ii
Tôô
-;ôô
Sourice : Data taken from "Nouvelle Politique
Agricole",
Direction
Generale de la Production Agricole,
annexes p. 6-18, 1984.

---

,- - -

---

22
30% of the region's
total
land. 40% of the millet
and 80% of the
cowpeds produced in the region are grown in that zone (SONED, 1981).
On the other hand, the Casamance region is chosen as a second
regiotl of study for these reasons:
- Its average annual rainfall
of 1500 mm (above theWest African
Semiarid Tropics (WASAT) norms) and its land use rate of 40% reveal
important unused agricultural
potential
in that region.
- Rice and cotton, two important cash crops in Senegal are primarily
grown in that region.
Similarly
to the Peanut Basin, this region is divided into three
agricultural
zones on the basis of climate,
rainfall
and ethnie group.
The Upper Casamance (UPC), considered as a high potential
zone for
growing millet
maire and cotton,
is selected as the second
agricultural
zone in this study.

---

CHAF'TER III
This chapter sets out the framework of the LP mode1 used in this
study, It includes a description
of a typical
farm structure,
trop
production
activities,
objective
function
and main constraints
existing
in the studied zones.
321 Typical fana structure
This section describes a typical
farm in each of the two zones
considered in this work. Comparisons are made between those two zones
in terms of land site and family composition.
The technical
packages
and the field
work periods prevailing
in each zone are then discussed.
321.1 m
d
mulation,
size
Tdbfe 3.1.1 summarizes the land size and the demographic
composition of a typical
farm household in both zones. Compared to the
Upper Casamance, farms are typically
larger in the tenter of the
Peanut Sasin in terms of area (6.5 versus 4.5 Ha) but smaller in terras
of famlily size (9.5 versus 16 individuals).
In terms of family
composition,
both zones have identical'percentages
of adult males and
females. It should be noted that the population
categories
in
23

---

24
Table'3.1.1
reflect
the norms being used at the Senegalese Institute
of Agricultural
Research (I$RA). Children younger than seven years of
age are not considered active agricultural
workers.
3.;1.2 Cronoing pattern
Three major rainfed crops are taken into account in the Center of
the Peanut Basin: millet/sorghum,
cowpeas and groundnut. Table 3.1.2
shows'the on-station
expected yields of these crops under three
rainfgll
conditions.
Six major crops are retained in the UpperCasamance:
millet/sorghum,
maize, groundnuts, cotton, rainfed rice and lowland
rice. Table 3.1.3 gives the expected yields of these crops.
A between-zone comparison shows that for millet,
the Upper
Casamqnce has a much better potential
to achieve high yields
than the
Center of the Peanut Basin. This cari be partially
explained by the
more a,equate distribution
and average levels of rainfall
observed in
the Upper Casamance. The same kind of observation
cari be made for
groundjut yields between the two zones.
3.11.3 Jechnicql waaes
Thb expression "technical
packages" refers to a set of
technologies
available
to farmers in a given region for a
partictilar
trop. Overall,
five different
technologies
(modules) each
representing
a combination of inputs and outputs are used in this
study.
Module one corresponds to highly intensive
technologies
based on
forma1 research recommendations. It makes extensive
use of chemical

---

TABLE 3.1.1
: FM SiZE A#0 POWLATIoN
c
ITIQU
AGE CATEGORIES
----_-_-----------------------------------
El
Child
Young
Men
Women
Old
Regiops
size
-8
8 - 14 15 - 59 15 - 59
60t
Peanutj
Basin
6.5
2.3
1.1
2.5
2.5
1.1
Casama/nce 4..5
3.5
2.4
4.2
4.2
1.7
Source : Martin and Sidibé (Martin 1986b; Martin and Sidibé
1987).
TABLE 3.1,2 : EXPECTED YIELDS I# THE
CENTMI. PE&#UT wSI#
. (units - Kg/Ha)
Rainfabl
Millet
Cowpeas
Groundnuts
Bad ye!r
200
450
Averag)e year
$0
700
950
Good ypar
go0
1000
1200
Sourcb:
Martin (1986b)
TABLE 3.1.3 : EXPECTER Y IElRS IN THE
UPPER
E ZO#E
(units = Kg/Ma)
Rainfed
Lowland Ground,
Rainfagl
Millet
Maize Rice
Rice
nuts
Cotton
Bad
700
500
500
650
1100
J$“4”
1500
1200
3000
1500
2000
1500
2000
3000
1200
1800
1500
1800
Source:
Martin and Sidibe (1987)

---

26
input$. Good soi1 types and careful trop management practices
are
required to achieve expected yields.
Modules two and three are also based on research recommendations.
However, they require respectively
less intensification
than module
one arid module three is less intensive
than module two. Those
technologies
are more compatible with farmers'
actual conditions..
Module four represents crops grown on "house fields",
i.e.,
land
next to the house which is put into cultivation
before the first
useful rain. Millet,
maize or vegetables are usually grown in those
fields.
They do not require chemical inputs.
Mgdule five corresponds to crops that were planted late as a
result of labor constraints
or seed problems during the normal work
schedule. Yields are lower than expected in this package as a result
of insufficient
rainfall
or incomplete maturing cycle.
3.;1.4 Cropoii
galendars
THe cropping calendars prevailing
in each zone are largely
determined by the length of the rainy season. Prior to the first
useful rain, a succession of land preparation
operations
must be
carried out by farmers to secure themselves against Sharp yield
losses. Graphs 3.1.1 and 3.1.2 show the different
growing periods used
in this study for the Center of the Peanut Basin and the Upper
Casamance respectively.
The Center of the Peanut Basin has more labor
periodis (6 versus 4) than the Upper Casamance.
3.51 Theoretical
considerations
The purpose of this section is to clarify
some key concepts
underlying
this research. First,
the possible states of nature used to

---

w-l
- -a--
c---.4 -
27

---

w-i
“O.---W fp.-#
-
CI”-*
CL 8L.tL.n Y...nL . -
28

---

29
mode1 risk are discussed.
Second, the theoretical
mode1 representing
a
typical
farm decision context is specified.
3.i2.1 $,&&
cJf pature
It is assumed in this mode1 that farmers maximize expected net
income subject to meeting their subsistence
needs under the most
adverse situations.
This objective
reflects
the fact that Senegalese
farmers are not considered exclusively
as profit
maximizers.
They have
other objectives
expressed in terms of food security
and social
obligations
which are treated as constraints
in this model. Income and
food derived from rainfed crops are strongly
dependent on yields
and
therefore
on rainfall.
The latter
affects
the uncertain
variable
"yields"
in two ways: by the volume of rain observed in a given year,
and its distribution
across the rainy season. Based on historical
data
on rainfall,
the amount of rain'is
categorized
in five groups (see
table '3.2.1) and the distribution
of rainfall
in three classes (see
table 3.2.2).
A crosstabulation
of these two "categorical
variables"
produces fifteen
different
combinat-ions of rainfall
or possible
states
of nature affecting
trop yields.
Table 3.2.3 illustrates
the method
used to calculate
expected yields over the fifteen
states of nature
for diifferent
crops.
3.i2.2 Mathematical mode1
Thie typical
farm is represented in terms of the following
linear
progrming
model:
Maximtize :
R = Zi 'Ci Xi
(i = 1, . . . k activities)
Subjecit to :

---

30
TABLE 3.2.1
DEFINITION OF RAINFALL AMODNT CATE(GORIES
Categories
Ranges
1) Vefiy low
9<
.7 * Q
2) Lo/
,7 * Q <= q < .9 + Q
3) Avejrage
.9 * Q <= q < 1.1 * Q
4) High
1.1 * Q c- q < 1.3 * Q
5) Vefly high
q >= 1.3 * Q
__-_-----------_----I___________________-----------------------
Sourde : flartin
(1987a)
q = Amount of rain during a given year
Q = average rainfall
between 1970-80
TABLE 3.2.2
DEFINITTDN OF RAINFALL DISTRIBUTION CATEGORIES
-_-_--.-3-_-----_-_--_________^__________------------------”--
Categories
Ranges
^_----i--__-_----_--____________________------------*--------
1) Bad
e N E * 1.25
2) Average
E * 1.25 > e >= E * .75
3) eoop
E"
.75 > e
------_--_-_--L------------~-
--------------------------------
Source : Martin (1987a)
e = Sum of deviations
of rainfall
across month
from average rainfall
of that particular
month
across years of observation,
E=(Xe)/n:
n - number of observed years.

---

31
zi,j
aij xi
<- bj
(j = 1, . . . constraints)
W,l X,
>- 0
AU Xm
)s bj
WI X,
>= 0
IN X,
>s bj
.
l

.
MIN1
>= MI1
MAX1
<= MA1
Where :
R
= Net revenue
c
= Net total cost
- Activity
i
X(m) = Activities
for crops produced for home consumption
X(n) = Activities
for crops produced for sale
b(j)
- Resources available
or needs of farm unit
DA
- Vector of worst deviations
from average yield for
crops produced for. home consumption(ca7ories).
AU
- Minimum level of cereal self-sufficiency
DI
= Vector of worst deviations
from average income for
crops produced for sale.
IN
= Minimum level of income.
MIN
= Minimum needs for cereal 1 due to food habits
MAX
= Maximum needs for cereal 1 due to food habits

---

32
TABLE 3.2.3
: EXPECTED YXELDS AND STATES OF NATURE
.
Distr$bution
of
Volume of rain
rainfall
Very low
Low
average
tiigh
Very high
Hillet/m
Bad
B * ,9
(BtA)/2
A
A
B
Averqge
(B+A)/2
A
G
G
A
Good
‘A
G
G”1.2
G*l.2
G
Groundnut, i2luQ!!, Cowaeas
Bad
B* .9
(BtA)/2
A
(BtA)/2
B
Average
(B+A)/L
(AtG)/2
.G
(AtG)/2
A
Good
A
G
G*1.2
G
WW
Bad
B * .7
B
A
B
B * .7
Averqge
B * .9
A
G
A
B
Good
(BtA)/2
(A+GI/2
G*1.2
G
A
!La
Bad
B * .6
B * .8
B
A
G
Aver*e
B * .8
B
.A
G
G * 1.1
Good
'B
A
G
G * 1.1 G * 1.2
Squrce : Martin (1987a)
B = average yield in bad rainfall
years
A = average yield in average rainfall
years
G = average yield in good rainfall
years

---

33
34 Mode1 activities
Thïs section describes the activities
carried out in each of the
agricul'tural
zones involved in this study. For the Central Peanut
Basin, forty-four
activities
are retained while the Upper Casamance
accounts for seventy-two activities.
The following
section discusses
the different
types of activities
in these two zones by type of crops.
3.3.1 QQQ groducinq activities
MiTlet/sorghum
is grown for home consumption and for sale. Home
consumed millet
is cultivated
under the five technical
packages
defined in Chapter Two. Millet
is grown for sale using four
technologies.
Both regions contain nine combinations
of
millet;producing
activities.
MaPze is cultivated
only in the Upper Casamance (UPC) under five
different
technologies,
three for home consumption and two for sale.
Cowpeas is produced in the Central Peanut Basin (CPB) for home
consumption only. Technologies one and two are used for this trop.
Two types of rice are produced in the Upper Casamance: rainfed and
lowland rice. Rainfed rice accounts for six activities:
four modules
for home consumption and two modules for sale. Lowland rice acounts
for ten activities:
five modules for home consumption and five for
sale.
Groundnut is cultivated
in both regions for sale under four
different
technologies.
Cotton is produced in the Upper Casamance under four different
technical
packages. This trop is used only for sale.
Tables 3.3.1 and 3.3.2 summarize the different
types of activities
carried out in both zones.

---

34
3.3.2 Jnaut procurement activiti,es
The tables mentioned in the preceding section show the seven
different
types of input procurement activities
in the two models. The
follotiing
discussion
highlights
special features inherent to each
zone.
Sesd buying activities
are included in both models for groundnut
and corwpeas in the CPB, groundnut and cotton in the UPC. Six
fertil,izer
buying activities
are use! in the Upper Casamance mode1
while the Central Peanut Basin uses only three of these activities.
Insect:icide,
herbicide
and fungicide' buying activities
are used in
both models at varying levels depend~ing on the type of trop grown. The
number of labor hiring activities
in the two zones is different.
Five
growing periods were identified
in the CPB yielding
five labor hiring
activities
while the UPC mode1 includes four hiring
activities.
3.5.3 @&Dut sellm
activities
In the Central Peanut Basin model, three trop selling
activities
are included:
millet/sorghum,
cowpeaS and groundnut. However,
groundnut and cowpea.hay selling
activities
are added to the model,
which yields
five different
activities.
The Upper Casamance mode1 accounts for seven output selling
activities
reflecting
the six crops grown in that zone and an
additional
groundnut hay selling
actjvity.
3.3.4 Çereal buvinq activities
Four different
cereals buying activities
are taken into account by
the mode1 in both zones. Millet/sorghum,
maize, rice and wheat (in
form of bread) may be purchased. For millet/sorghum
and maize, the
mean weighted price over the fifteen
states of nature is used as the

---

35
TAULE 3.3.1
: MOUEL ACTIVITIES IN THE CENTRAL PEANUT BASIN
Activity
Activity
description
Activity
number
name
Al - A5 Millet/sorghum
for home consumption
PMICl to PMIC5
A6 - A9
Millet/sorghum
for sale (ino module 4)
PMIVl to PMIV5
A10 - Al1 Cowpeas
PNIEl to PNIE2
A12 - Al5 Groundnut (no module 4)
PARA1 to PARA5
Input orocukment
Al6
BUY graundnut seed
AS EAR
Al7
BUY NPK (14-7-7) for millet/sorghum
ANPKl
Al8
BUY NPK (6-20-10) for groundnut and cowpeas
ANPK2
A19
BUY urea
AUREE
A20
BUY insecticide
1 for cowpeas
ANINl
A21
BUY insecticide
2 for cowpeas
ANIN
A22
BUY fungicide
for groundnut
ARAFO
A23
Hire labor in period Pl
AM01
A24
Hire labor in period P2
AM02
A25
Hire labor in period P3
AM03
A26
Hire lqbor in period P4
AM04
A27
Hire labor in period P5
AM05
A28
Borrow capital
to buy groundnut seed or food
ACAP
Risk,transfer
rows
A29
Risk transfer
column for food self-sufficiency
RISKA
A30
Risk transfer
column for ininimum income
RISKR

---

36
Actiivity
Activity
description
Activity
number
name
A31
Capital group 1 transfer
column
CAPTl
(borrowing possible)
A32
Capital group 2 transfer
column
CAPT2
(borrowing not possible)
Cereal buvino activities
A33i
Buy millet/sorghum
AMIL
A34
Buy maize
AMAIS
A%
Buy rice
ARIZ
A361
Buy wheat
ABLE
Outout selling
activities
A37
Sel1 millet
for sale
VMIL
A381
Sel1 maize
VMAIS
A39:
Sel1 groundnut shells
VARAG
A401
Sel1 groundnut hay
VARAF
A41
Sel1 cotton
VCOT
&mmv activities
(provisory)
A42
Self-sufficiency
constraint
DUMA
A43
Income constraint
DUMR
Source : Martin (1986b)

---

37
TAM% 3.3.2
: HODEL ACTIVITIES IN THE UPPER CASAMANCE
Activity
Activity
description
Activity
number
name
Production
Al to A5
Millet/sorghum
for home consumption
PMICl to PMIC5
A6 to A9
Millet/sorghum
for sale (no module 4)
PMIVl to PMIVS
Al0 to Al2
Maize for home consumption (no module 3) PMACl to PMACI
Al3 ta Al4
Maize for sale
PMAVl to PMAV2
Al5 to Al7
Rainfed rice for home consumption
PRRCl to PRRC4
Al8 to A19
Rainfed rice for sale
PRRVl to PRRV2
A20 to A24
Lowland rice for home consumption
PLRCl to PLRC5
A25 ta A29
Lowland rice for sale
PLRVl to PLRV5
A30 tq A33
Groundnuts (no module 4)
PARA1 to PARA5
A34 td A37
Cotton (no module 4)
PCOTl to PCOTS
Jnput nrocurement
Buy groundnut seed
ASEAR
Buy cotton seed
ASECO
A46
Buy NPK (14-7-7) for millet/sorghum
ANPKl
A4I
Buy NPK (8-18-27) for maite
ANPK2
A42
Buy NPK (18-46-O) for rice
ANPK3
A4J
Buy NPK (6-20-10) for groundnut
ANPK4
Buy NPK (6-14-35) for cotton
ANPK5
A4%
Buy urea
AUREE
A46
Buy herbicide
for millet/sorghum
and maize
AHEMM
A47
Buy herbicide
for rice
AHERI
A48
Buy fungicide
for groundnut
AFOAR
A44
Buy herbicide
# 1 for cotton
AHECl
A5a
Buy herbicide
# 2 for cotton
AHEC2
A51
Buy insecticide
for cotton
AINC
A52
Hire labor in period Pl
AM01
A53
Hire labor in period P2
AM02
A54
Hire labor in period Rl
AMORl
A55
Hire labor in period R2
AMOR2

---

38
Activity
Activity
description
Activity
number
name
A56
Borrow capital
to buy groundnut seed or food
ACAP
Risk transfer
rows
A57
Risk transfer
column for food self-sufficiency
RISKA
A58
Risk transfer
column for minimum income
RISKR
Çaaital transfer
columns
A59
Capital group 1 transfer
column
CAPTl
(borrow ing possible)
A66
Capital group 2 transfer
column
CAPT2
(burrow ing not possible)
Çereal buvins activities
A61,
Buy millet/sorghum
AMIL
A62
Buy mafze
AMAIS
A63
Buy rice
ARIZ
A64i
Buy wheat
ABLE
Outout selling
activities
Sel1 millet
VMIL
Sel1 maize
VMAIS
Sel1 rice (rainfed
and lowland)
VRIZ
Sel1 groundnut shells
VARAG
Sel1 groundnut hay
VARAF
Sel1 cotton
VCOT
mmv activities
(provisory)
A71
Self-sufficiency
constraint
DUMA
A72
Income constraint
DUMR
Sourcje : Martin and Sidibe (1987)

---

39
market, price. The prices used for the other cereals reflect
those
prevai'ling
in Dakar plus the transportation
cost from Dakar to the
corres/ponding regions.
3.3.5 Çgeital msfer
activities
Tti capital
transfer
activities
are used in each mode1 to
handle
situations
where borrowing is allowed or not allowed. Repayment of
borrowpd capital is handled by a third activity.
3.3.6 &~JIJ transfer: activities
Twn activities
are used to handle risk situations
faced by farmers
relative
to income and subsistence requirements.
In the case of income
requiroments,
the level of the risk transfer
activity
indicates
the
amount by which crops produced for sale must increase to satisfy
the
income constraint.
For the subsistence requirements,
the level of the
risk transfer
activity
indicates
the,amount by which crops produced
for home consumption must increase to meet the minimum food needs.
3.8.7 Dunny activities
Two dummy activities
are used in each mode1 as check activities.
Their presence in the optimal solution
is an indication
of
inconslstencies
in the income or food self sufficiency
deviations
constr@ints.
3.4 Objective
function
values
In this model, the objective
is to minimize costs. For
computational
convenience, the objective
function
coefficients
have
been multiplied
by -1 to convert the problem to one of maximization.
The purpose in this section is to cl'arify
the derivations
of the
coefficients
used in the objective
function.

---

40
3.:4.1 Production coefficients
For the trop producing activities,
the coefficients
used in this
mode1 represent the per hectare cost of inputs used and not specified
in the input procurement activities.
This Will include the variable
cost of using agricultural
equipment and some amount of depreciation
on thijs equipment.
3.4.2 Dput procurement coefficients
The objective
function
coefficients
for the input procurement
activities
are their respective
prices to the producers (1986
figures),
excluding any special short-term
subsidy. A monthly interest
rate of seven percent is estimated for repayment of borrowed capital
to buyi inputs.
3.11.3 Cereal buvinq and outout selling
coefficients
Tha mean weighted prices over the fifteen
states of nature are
used for objective
function
values of the cereal buying and selling
activities,
For rice and wheat, the prices prevailing
in Dakar plus an
inter-region
transportation
cost represent their coefficients.
3.k.4 Qther coefficients
The objective
function‘ values of the risk and capital
transfer
activijties
are equal to zero,
3.6 h-del constraints
This section describes the different
types of constraints
which
the twb regional models include, with a special emphasis on the
food-security
constraints.
The Central Peanut Basin zone has
forty-three
constraints
while the Upper Casamance mode1 accounts for
forty-beven
constraints.

---

41
TABLE 3.5.1
: M#IEL CONSTRAENTS
IN THE
CENTRAL
BASIN
ZONE
Constraint
Description
Units
Sign
RHS
2
PU
<=
Groundnut seeds
0:;
<=
C3
Cowpeas seeds
Kg
<=
c4 - c7
labor in period 1 to 4 Man/day
<=
4E
Labor in period 5
Man/day
<=
137: 0
ci -Cl2
Animal traction
1 to 4 Ani./day
<=
10.0
Cl3
Animal traction
5
Ani./day
<=
64.0
Cl4
NPK for millet
Kg
<=
Cl5
NPK for Cowpeas
Kg
<=
Cl6
Urea for millet
Kg
<=
Cl7
Insecticide
1 cowpeas Treatment
<=
Cl81
Insecticide 2 cowpeas Treatment <=
Cl9
Fungicide groundnut
Treatmen
t <=
c20
R”i %t%
FSS
Calories
>=
c21
Minimum
1 evel of FSS Calories >=
C22
Risk rows for income
CFA
>=
C23
Minimum income
CFA
>=
C24,
Starting
capital
CFA
<=
20000
c25
Capital group 1
CFA
<=
C26
Capital group 2
CFA
<=
:
C27
Repayment of capital
CFA
<=
C28
Millet
sold
Kg
<=
0
C2!J, c30
Cowpea grain, hay sold Kg
<=
c31,
Groundnut unshelled
Kg
<=
ii
c32
Groundnut hay
Kg
<=
C33i
Nutrition
needs
Calories
>=
510:
C34,
Minimum acreage M4
>=
.05
c35:
Maximum acreage M4
Fiil
<=
C36:
Min consumption millet
Kg/Ha
>=
1025 l :
c37
Max consumption millet
Kg/Ha
<=
1710:o
C38:
Min consumption maite
Kg/Ha
>=
c39:
Max consumption maize Kg/Ha
<=
95*D
c40
Min consumption rice WHa
>=
314:o
c41
Max consumption rice
%/Ha
<=
523 .O
C42’
Min consumption wheat Kg/Ha
>=
19.0
c43
Max consumption wheat Kg/Ha
<=
114.0
Sou<te : Martin,
1986b
FSS is defined as food self-sufficiency.

---

42
TABLE 3.5.2
Constraint
Description
Units
Sign
RHS
!!!%?Y”
<=
Land 2
ii
<=
i:6
Groundnut and cotton seed
Kg
<=
Labor in period 1
Man/day
<=
104*;
Labor in periods 2 and 4
Man/day
+
98:0
Labor in period 3
Man/day
126.6
Animal traction
1
Ani./day
1:
20.0
ClO, $1
Kg
<=
Cl2
:
NPK for rice
Kg
<=
Cl3
NPK for groundnut
KS
<=
Cl4
NPK for cotton
Kg
<=
Cl5
Urea for millet
and maize
Kg
Cl6
Herbicide for millet
Treatment
:z
Cl7
Herbicide for rice
Treatment
<=
C18, $9 Herbicide 1, 2 for cotton
Treatment
<=
C20
Insecticide
for cotton
Treatment
<=
C21
Fungicide groundnut
Treatment
<=
c22
"
%!%
FSS
Calories
>=
C23
Minimum level of FSS
Calories
>=
516:: ii
C24
Risk rows for income
CFA
>=
c25
Minimum incarne
CFA
>=
85ii:
C26 ; Starting
capital
CFA
<=
25000
C27, 2!8 Capital groups 1 and 2
CFA
- <=
0
c29
Repayment of capital
CFA
<=
C30, 3il Millet
and maite sold
Kg
<=
:
C32, 3ib Rainfed lowland rice sold
Kg
<=
c34
Groundnut unshelled sold
Kg
<=
0
c35
Groundnut hay sold
Kg
<=
C36
Cotton sold
Kg
<=
ii
c37
Nutrition
needs
Calories
>=
6489
C38
Minimum acreage module 4
Ha
>=
.05
c39
Maximum acreage Module 4
Ha
<=
C40
Min consumption millet
WHa
>=
102i.e
C41
Max consumption millet
M/Ha
<=
171o:o
C42
Min consumption maize
@/Ha
>=
c43
Max consumption maize
RVHa
<=
9R
c44
Min consumption rice
Kg/Ha
B=
314:o
c45
Max consumption rice
Kg/Ha
X=
523.0
C46
Min consumption wheat
@/Ha
>=
19.0
c47
Max consumption wheat
WHa
<=
114.0
Sourice : Martin and Sidibe (1987)
FSS is defined as food self-sufficiency.

---

43
3.5.1 @source use constrahts
These constraints
deal with land, labor, trop seeds and animal
traction
resource restrictions
and availability.
Tables 3.5.1 and
3.5.2 Jay out these constraints
for both zones. The major differences
between the two models are discussed below. The Upper Casamance (UPC)
mode1 has two types of Jand, one for rainfed crops and the other for
lowland rice, yielding
two land constraints.
The Central Peanut Basin
(CPB) mode1 has only one type of land and therefore
one land
constcaint.
Because of its shorter rainy season, the CPB mode1
requises more labor and animal traction
constraints
than its
countdrpart.
3.:5.2 &gmical jnout. constraintq
These constraints
represent the resource restrictions
put on
fertilizer,
herbicide,
insecticide
and fungicide
used by crops under
different
technologies.
Tables 3.5.1 and 3.5.2 show the types of
chemical inputs used in both zones.
3~5.3 Lgpd-securitv
constraints
Food security
constraints
are divided into Food Self- Sufficiency
(FSS) constraints
applying to food crops, and into Income constraints
applying to cash crops. Table 3.5.3 illustrates
the structure
of the
FSS cqnstraints
for a hypothetical
zone. Millet
is used as an example
of a food trop.
The millet activity
is broken down into millet
produced for home
consumption and millet
produced for sale. There are 5 levels of
technelogy for millet
production.
Level of technology 4 is missing in
the millet for sale category because it represents "kitchen garden"
techntplogy which is used only for home consumption.

---

44
in the case of millet
produced for home consumption, there are two
sets of constraints.
The first
set is represented by the most
unfavorable states of nature across levels of technology measured by
the most negative deviations
of mean yields from mean weighted yields.
The second set is represented by the mean weighted yields
across
levels of technology.
The Right-Hand Side (RHS) of this second set of
constraints
is represented by the family food needs in calories
which
must be satisfied
by cereals (65% of total needs).
In the case of millet
produced for sale, there is only one set of
constraints
represented by the weighted yields
of the most unfavorable
statess of nature identified
in the case of millet
produced for home
consun@tion.
Al1 deviations,
weighted yields and mean weighted yields
are
expressed in calories.
The FSS constraints
operate as follows:
if
farmers want to secure their FSS need during the most unfavorable
states of nature, the negative deviations
for a given technology is
compensated by transfering
millet
produced for sale to millet
produced
for home consumption.
If the millet
produced for sale transfered
to
home cpnsumption is not sufficient
to caver the "downside deviations",
the "risk penalty"
of not satifying
the food need arises.
The LP
algoriéhm Will attempt to transfer
the "risk penalty"
through a FSS
coefficient
(see Table 3.5.3 (1)) to increase the level of millet
producod for home consumption. The "downside deviation"
may not be
compensated if a farm is incapable of achieving FSS
objectives.
In this case, an infeasible
solution
is obtained.
The income constraints
applying to cash crops follow basically
the
same structure
and mechanism. However, only the trop for sale category
Y

---

TABLE
3.5.3
:
STRUCTURE
OF THE FOOO
SELF-SUFFICIENCY
CONSTRRINTS
FOR R WI’~=C~THETICAL
ZONE
___------_-____--__-_________I__________-------------------------------------------------------
-‘F-f-BN
N-ILkET
FWF?
AMiE
îTrtxETFuR
i?CTIUITIES
CmWTIUN
SALE
--------------------__l_l
------------1-----
RIS
TRwER
TECHNOLOGIES
1
2
3
4
5
1
2
3
5
McLlNN
RNS
-__-~I-_~_~__-____----~-~~~~~~-~~~~~--~--~~---~-~~~~~--~-~~~-~~~---~~~~~--_~------_-__-----
C3EVfATIONS
OF
tlEBN
YIELDS
FRDM
MERN
bEIFtiTEO
YIELDS
CCslories>
state
1
dl1
612
dl3
dl4
dl5
1111 y 12
y13
y15
1
st.astG?
2
d21
622
423
624
d25
y21
y22
y23
y25
1
.
.
m
.
I
.
.
m
.
I
m
.
.
stat.e
t-l
dt-il
dn2
*
dn3
*
dn4
.
dn5
wl
yn2
yn3
41n5
1
MXNIML1M
LEVE&
UF’
CERERL
FDOO
SELF-SUFF
1 C IENCY
Cc--alut-ies)
c 1 >
Family
FS*5
WY1
WY2
wy3
w-J4
uy5
-I 375
>=f
oad
nowds
------------------------------------------
------------------------------------------------------
d i .j t-epresants
rnsi~
negat
ivu
dsviatinns
ft-am
mean
wsightr.ed
yie lds
fot-
state
af
nistut-e
i
and
f,echtmlcq1~
.j oxpressed
in
tcwms
of
calot-ies.
y i ,.j rupreser1t.s
mi 1 1~4~. we ighCed
yi v lds
for
zst.at.e
of
nat.ut-e
i and
tuchnulc?gy
.j
expt-essud
i t-1 turms
af
ca lat- i es.

---

46
is represented.
Deviations
are of mean income from mean weighted
income (instead of yields).
Mean weighted income appears in the
minimum income constraint
(the counterpart
of the FSS constraint)
across Tevels of technology.
The RHS for the minimum income constraint
is represented by an estimated minimum desired income.
3.5.4 gther çonatraints
Several other constraints
are present in both models. Food habits
constraints
are built
into the models to keep household consumption of
different
cereals within realistic
bounds. Accounting identities
are
used to ensure that quantities
sold for any crops are no more than
quantities
produced. They also provide a means of separating
the
capital
constraint
into two rows: one where borrowing is possible and
the other where borrowing is not allowed. A minimum acreage constraint
is set up for technical
package four to reflect
the importance of the
"kitchkn garden' in farmers' food security.
Appendices Tables 1 and 2 show the initial
LP tableau of the two
zones :involved in this study.

---

SENSITIVITY AfUkLYSIS
This chapter focuses on the analysis of the optimal solutions
of
the two models discussed in the previous chapters. The first
section
addresses the optimal enterprise
mix prevailing
in the two zones
studied.
It is followed by a comparison of cropping intensities
derived in the two models with primary data. A post-optimality
analysis
is then performed on the optimal plans observed in the two
zones kith emphasis on resource and objective
function
ranges. The
last section is devoted to the development of a knowledge base,
supporting the expert system that this study seeks to implement.
4.Z Base run
Two assumptions form the basis of the results
derived from the
base run of the LP models. First,
it is assumed that farmers want to
achieve 80% food self-sufficiency
(FSS) in average years and at least
38% FSS during bad years. Second, minimum income estimates
(see
Chapter 3) in the Central Peanut Basin (CPB) and Upper Casamance (UPC)
zones are used as farmers' objectives
with regard to desired income.
47

---

48
TABLE 4.1.1 : OPTIRAL PLAN A#D CROPPTNG
INTENSITY
IN ME CENTtW. lW#?&lT
EASIN
(Raturn - 171 656 CFA)
TECHNICAL PACKAGES
TOTAL
------_--_-“-_-----c________
-------..---_
Crops
12
3
4
5
Ha
%
Pn anted (Ha)
Millet; for home
1.48
1.42 1.00
3.9
60
consumption
Groundnut
2.60
2.6
40
Source;: LP88 printout.
Technical packages 1 to 5 refered to trop technologies
defined in chapter 3.
TAkLE 4.1.2 : OPTïML REAL ACTIVITIES IN THE
CENTRAL
BASIN ZWE
ActivTty
Level
Unit
Net return
BUY NPK millet
147.7
Kg
-81.5
Buy urea
73.9
Kg
-72.5
Fungiciide groundnut
Treatment
-1000.0
Buy capital
5514t:
CFA
-0.28
FSS riFk
3763:4
Calorie
Income risk
148399.5
CFA
:*Fi
Buy rii=e
95.0
Kg
-163:0
Buy wh'at
19.0
Kg
-278.0
Unshel % ed groundnuts
1896.2
Kg
90.0
Groundnut hay
2290.1
Kg
40.0
Sburce : LP88 printout.
FSS is defined as food self-sufficiency.

---

49
4.‘1.1 m
intensitv
Table 4.1.1 illustrates
the optimal plan and cropping intensity
for the CPB zone. Millet/sorghum
for home consumption is retained by
the mode1 under technologies
one, three and four with a share of 60%
of land use. Intensive technologies
are the most economically
attractive
packages for this trop. Groundnut under package three, the
least intensive
module recommended by research, has a land share of
4oxP of total land use. Cowpeas have not been found to be an
economically
interesting
trop in this zone. The net return associated
with the optimal plan is 171656 CFA. Other activities
in the optimal
solution
are also shown in Table 4.1.2. Labor hiring activities
are
not in this plan, denoting an unused labor capacity
in a11 periods.
This plan requires borrowing of capital
to buy food or cash-trop
inputs to support it. Mode1 results
show that both the minimum food
self-sufficiency
(FSS) and the minimum income risks are in the
solution.
A food deficit
of 3763 units of calories
in association
with
an imcome loss of 148400 CFA francs would have been incurred to
farmer:s if risk has not been taken into account. The cereal buying
activities
show that farmers in this zone are better off by not
selling
millet
and by buying rice and wheat for consumption.
Inputs
must be purchased for both millet
and groundnut crops.
FOP the UPC model, a combination.of
six crops under different
technologies
is included in the optimal enterprise
mix of that zone.
An illustration
of this plan is given in Table 4.1.3. Millet/sorghum
for home consumption is produced in packages one, three and four;it
has a land share of 38%. Maize is cultivated
in package two with

---

50
TABLE 4.1.3 : OPTIMAL
PLAN Al@ CROPPING
INTENSITY
IN THE UPPER
E ZWE
(Return = 85035 WA)
TECHNICAL PACKAGES,
TOTAL
-------------------__c__________
----I----v--
Crops
1
2
3
4
5
Ha % Upland
Millet cons. 0.16
1.58
0.74
1.48
38
Haize !Cons.
0.74
0.74
20
Rainfed rice
0.03
0.03
0.8
Lowl arid rice
0.58
0.02
0.60
Grountfnut
1.30
1.30
3;
cottoni
0.36
0.36
9.2
Sdurce: LP88 printout.
(*) Lowland rice is cultivated
on different
land
therefore,
it does not appear in the percentage
calculation.
TABLE 4.1.4 : OPTIMAL REAL ACTIVITIES IN
ME UPPER
E XI#E
Activity
Level
Unit
Net return
Buy groundnut seeds
153.6
Kg
-110.0
Buy cotton seeds
17.0
Kg
-110.0
Buy NPK millet
15.5
Kg
-90.9
Buy N K Rainfed rice
73.4
Kg
-93.9
Buy N & , lowland rice
153.:
Kg
-115.9
Buy N/ groundnut
Kg.
-88.9
Buy NF cotton
54:o
Kg
-90.9
Buy uqea
101.3
Kg
-81.9
Boy herbicide
millet
Treatment
-9750.0
Buy herbicide
cotton 1
0::
Treatment
-8750.0
Buy herbicide
cotton 2
0.4
Treatment
-10000.0
Buy insecticide
cotton
Treatment
-7350.0
Fungicjide groundnut
:*s
63555: 8
Treatment
-1000.0
Buy caipital
CFA
-0.28
FSS riisk
4718.7
Calorie
Buy mi:llet
182.3
Kg
43*;
'
Sel1 c!otton
613.4
R3
95:o
Sel1 gjround. shell
1810.6
Kg
90.0
Sel1 giround. hay
2947.7
Q3
13.0
Source : LP88 printout.
FSS is defined as food self-sufficiency.
-- -- ---

---

51
20% share of rainfed land. Rainfed rice in package one, groundnut in
package 3 and cotton in package 1 have respective
rainfed land shares
of 0.8, 33 and 9% of total land use. The optimal objective
function
value (net revenue) is 85035 CFA francs. The remaining optimal
activities
are given in Table 4.1.4. The borrowing of capital
to buy
food ar cash-trop inputs at a 28% interest
rate is necessary to carry
out this plan. The FSS risk activity
indicates
that 4719 units of
calortes
are transfered
from crops produced for sale to crops produced
for hqme consumption.
4.1.2 Scarcity
values sf bindinq Con$traints
The implicit
resource values given by the solution
to the dual LP
problem are referred to as the shadow prices of those resources.
This
section discusses the scarcity
values of the binding constraints
existi!ng in the two zones included in this study.
For the CPB model, Table 4.1.5 illustrates
the shadow prices of
the binding constraints
prevailing
in that zone, and highlights
some
interesting
aspects of mode1 results.
Land shows the very high
opportunity
cost of 84686 CFA francs, attesting
that an additional
unit of land would greatly
improve the net revenue of a typical
farm.
Millet
fertilizer
is a key input to this plan: its implicit
price
shows that farmers should be willing
to support more than a 300% price
increaise and still
remain efficient.
The same observation
cari also be
made for urea fertilizer.
The starting
capital
constraint
is strongly
binding to this optimal solution;
each additional
unit of it would
triple
its contribution
to the net return.
The capital
type two
constraint,
closely related to the starting
capital,
follows
the same
behavior. The capital
type one constraint
does show that borrowing is
still
iprofitable
even at a doubled interest
rate.

---

52
Table 4.1.6 illustrates
the shadow prices prevailing
in the UPC
model, Both land types are binding constraints
to this plan. The
implicit
price of the lowland rice land is 50% higher than its
counterpart.
A between-zone comparison shows land is more valuable in
the CPB than in the UPC zone. Al1 fertilizers
and trop protection
products are scarce resources in this model. Labor in period one was
found to be a binding constraint.
However, its contribution
to net
income is only slightly
above its actual value of 500 CFA. Al1 capital
constraints
are binding for this enterprise
mix. Their scarcity
values
are lower in this zone than they were for the CPB.
411.3 j4odel. yg3idation
Tables 4.1.7 and 4.1.8 show the shares of the different
crops in
total land cultivated
for the mode1 results
compared to observed data
in the two agricultural
zones. The following
observations
are worth
mentioning for both zones.
In the CPB, millet
land share was 12% higher in the mode1 results
than in the observed data, groundnut area was 10% smaller in the mode1
and cowpeas are not in the optimal enterprise
mix. These variations
cari be explained in part by the following
facts.
When agricultural
inputs are available
as assumed in this study, there are alterations
in the actual cropping patterns.' There is a shift
away from groundnut
and cowpeas towards millet/sorghum,
yielding
a transfer
of land
traditionally
devoted to the former in favor of millet.
Cowpeab, which
seems to be an important trop at the present time, become economically
unattractive,
Thie UPC zone shows also some interesting
results.
Millet
produced
in the mode1 increases by 1% from observed data. Maize increases by 5%
and rike by 2% indicating
that rice and maize could potentially
be

---

53
TABLE 4.1.5 : SCARCITY VALUES OF BINiXN6 CONSTRAINTS
IN THE CENTRAL
PEANW BASIN ZONE
Constraints
Shadow price
Unit
RHS Value
land
84686.0
CFA/Ha
Groundnut
FertIliizer
seeds
millet
140.8
356.2
CFA/Kg
LE
CFA/Kg
FertETizer
ground.
196.8
CFA/Kg
i:D
Urea
316.8
CFA/Kg
Inseci$cide cotton 1
8640.0
CFA/Treat.
a::
Insecgicide cotton 2
6528.0
CFA/Treat.
0.0
Fungiiqide groundnut
4370.0
CFA/Treat.
Startiing
Capitail
capital
1
3.4
.28
CFA
2ooo::o
Capitdl 2
CFA
o*o
Food self-sufficiency
110::
CFA
4270: 0
Source: LP88 printout.
TfiBLE 4.1.6 : SCARCITY
VALUES
OF BINDING CONSTRAINTS
IN TtfE UPPER
E ZONE
Con$traints
Shadow price
Unit
RHS Value
Rainfe@ land
75223.0
CFA//Ha
Lowlanti land
116183.0
CFA/Ha
0:;
Groundjnut seeds
140.8
CFA/Kg
Cottoni seeds
140.8
CFA/Kg
0::
Fertiliizer
millet
321.4
CFA/Kg
Fertillizer
rice 1
332.0
i:O
Fertiliizer
rice 2
409.8
t$$
Fertiljizer
groundnut
314.3
CFA/Kg
0::
Fertil;izer
cotton
CFA/Kg
Urea
116.4
104.8
CFA,'Kg
0::
Herbic:ide millet
34471.6
CFA/Treat.
Herbic:ide rice
44365.5
CFA/Treat.
Ri
Herbic;ide cotton 1
11200.0
CFA/Treat.
, 010
Herbic"ide cotton 2
12800.0
CFA/Treat.
Fungic$de groundnut
3535.0
CFA/Treat.
ii*;
Labor period 1
513.9
CFA/Day
104:o
Starting
capital
2.5
CFA
25000 ..O
Capitag 1
.28
CFA
0.0
Capita/l 2
3:::
CFA
Food self-sufficiency
CFA
4270: ii
Sourpe : LP88 printout.

---

54
leading crops in the UPC zone. Groundnuts lose some of its
popularity(4X)
and cotton land share remains constant between mode1
results
and observed data.
4$2 Sensitivity
analysis
This section focuses on the analysis of changes in the optimal
solutiions of the LP problems given changes in various coefficients
associ:ated with the problems. The discussions
are centered around the
resource range variations
of the binding constraints
and the objective
functi:on values of the optimal activities.
4.i2.1 eesours
range variations
Table 4.2.1 illustrates
the results derived for the CPB model.
Farm land size cari vary from 6.0 to 9.0 Ha without affecting
the
actual enterprise
mix. This shows that mode1 results
cari tolerate
35%
variation
in land size without any effect on this optimal plan. The
millet
fertilizer
resource range shows that this plan Will be
mainta"i.ned whether farmers purchase at least 205 kgs. of NPK millet
or
hold c,arry-over stock of up to 136 kgs. An amount of 313 kgs. of urea
for mi'llet must also be available
to carry out this plan. The starting
capital
is allowed to vary within the range 3269 to 31071 CFA francs
without any effects
on the actual resource allocation,
The repayment
of the borrowed capital
is not a binding constraint
to this plan; this
point supports the fact that borrowing at 28% interest
rate is still
worthwhile
to farmers.
Mode1 results
are sensitive
to the minimum food self-sufficiency
constraint.
An 11% increase in the actual level of calorie
needs would
bring about another enterprise
mix.

---

55
TABLE 4.1.7 : OBSERVER AMI CALCULATED LANO SHARES
IN THE CEIURAl. FEAMT 6ASIWZONE
;Crops
Observed
f4odel
Oifferences
I+!i1 let
-12
:Groundnut
50
:o
t10
icowpeas
2
0
t2
Source : observed data from SODEVA, 1982-84
TABLE 4.1.8 : ~~S~E~~
EtlzE
SHARES
Crops
Observed
Mode1
Differences
Mfllet
*
32
33
R;i;e
i:
r :
::
-2
hundnut
33
29
t4
kotton
8
8
0
Source : observed data from SODEVA, 1982-84
. . . “ *
- -
-
_ - - - -
- - .
”
-_-”
. - . . .
-___

---

56
Table 4.2.2 shows the resource range variations
derived from the
UPC model. More than a 13% increase in rainfed land and a 7% increase
in lowland land would affect mode1 results.
The rainfed land in this
zone Shows a greater sensitivity
to land sire than was the case in the
CPB. This plan is sensitive
to a11 herbicide
constraints
and to labor
in period one. An additional
work-hour in that period cari affect the
optimal solution.
The food self-sufficiency
constraints
allow only a
10% increase in calorie
needs before a change in this optimal plan
occurs .
4.2.2 mctive
wction
coefficients
The post-optimality
analysis,
for objective
function
coefficients,
deals with the determination
of the range of variations
of those
coefficients,
within which the actual.enterprise
mix remains
unchanged.
Table 4.2.3 illustrates
the range of optimality
of the objective
function
for the CPB optimal solution.
Home-consumed millet
produced
under 'technology one, Will remain in this optimal plan even if its net
total cost increased by 100%. However, the millet
selling
activities
Will never be part of any optimal solution.
That same information
is shown in Table 4.2.4 for the UPC optimal
solution,
Home consumed millet,
under technology one, Will remain in
the optimal plan even if its net total cost is doubled.
Millet
produced for sale under a11 technologies
is unlikely
to be in any
optimal plan. Home consumed maize produced in packages one and four
show the same behavior. Rainfed rice produced for home consumption is
only aittractive
under technology two.

---

57
TABLE 4.2.1 : RE!jiXWX RASE VARIATIOW
Ill
THE CENTRA1
PEAWT BAW
Con$raints
RHS Unit
Minimum
Maximum
Land
6.5
Hectare
6 :
9
Groundnut seeds
Kg
-2137
313
Fertilirer
millet
Kg
-205
136
Fertiljizer
groundnut
Kg
313
Urea
KS
-23:
105
Insecqicide cotton 1
Treatment
-34
Insecqicide cotton 2
Treatment
-46
0
Fungidide groundnut
Treatment
-17
Starlihg
capital
CFA
3269
3107:
Capitajl group 1
CFA
- 8235113
55148
Capitijl
group 2
CFA
-16731
11071
Food $elf-sufficiency
Calorie
4224
4743
Saurce: LP88 printout
TABLE 4.2.2 : RESWRCE
RANGE
VARIATIONS
IN
THE UPPER
E ZoIyf
Con$raints
RHS
Unit
Minimum Maximum
Rainfqd 1 and
3.9
Hectares
3.7
Lowla ‘d land
0.6
Hectares
ii*"6
Groun nut
a
seeds
0.0
Kg
-15;:
1;4
Cottqq seeds
Fertilizer
millet
0.0
Kg
-1630
-53
:2
FertiIjizer
rice 1
0.0
Kg
-51
74
Fertïliizer
rice 2
0.0
Kg
-41
Fertiliizer
groundnut
0.0
-53
99
Fertilhzer
cotton
0.0
Kg
-1972
54
Urea
0.0
Kg
-2190
101
Herbidide millet
0.0
Treatment
-.
l 2
Herbicide rice
0.0
Treatment
i
Herbidide cotton 1
0.0
Treatment
-20
.!i
Herbidide cotton 2
0.0
Treatment
-17
Labor .in period 1
104.0 Man/Oay
10:
Startirng capital
25000.0 CFA
202::
33820
Capital group 1
0.0
CFA
- 195997
63555
Capitil
group 2
0.0
CFA
-4783
8820
Food s/elf-sufficiency
5165.0
Calorie
4621
5704
Soukjce : LP88 printout.

---

58
4d3 Kncmledge .repmsentation
This section introduces
the concept of expert system (ES) in the
content of this study. In that respect, the various components of an
ES are first
discussed and the remainder of this section is devoted to
the dosign of a knowledge base (KB) from the LP optimal tableau,
As
specified
in Chapter One, only the CPB results
are taken into account
for this purpose.
4.3.1 Çoponents, sf gr! $xrrer& gvstem
Loosely defined, an ES is a set of computer programs capable of
exhfbiting
a certain degree of intelligence
in a given field,
which
simulates to some extent a human expert in that field.
An ES has three
major components illustrated
in Figure 4.3.1.
Tbe knowledge base (KB) is an essential
part of an ES; it stores
a11 the knowledge necessary for it to apply its expertise.
The KB
includes factwal information
on the relations
between entities
and on
the rùles describing
those relations.
The inference engine (IE) is the driving
force of an expert
system. It contains operating rules and principles
designed to use the
KB efficiently
in order to match consistent
conclusions.
The user
interface
(UI) constitwtes
the link between the end user and the
inference engine. Its purpose is to carry out the user's queries and
retwrn: the inferred
knowledge back to him.
The last two components, known as an expert system shell,
interact
with t'he user and the KB to get the system to perform its task.
The Turbo Prolog expert system shell (Borland,
1986) is used to
integrate
the LP optimal solution
into a workable KB. This fifth

---

59
TABLE 4.2.3 : OBJECTIVE
FUNCTION
RANGES
IN
MIE CENTMi. t!YAWT BA$IN
Activity
Net return
Unit
Minimum
Maximum
Millet
consumed 1
-3600.0
CFA/Ha
-11896
159634
Millet
consomed
consumed 3
4
-3600.0
-3298.0 CFA/Ha
-26252
11192
CFA/Ha
- 55057
;:y;;‘*’
Groumdnut 3
-5488.0 CFA/Ha
-8794
Groundnut seeds
-110.0 CFA/Ha
-225
Buy NPK millet
-81.5
CFA/Kg
-315
100
fhy urea
-72.5
CFAfKg
-238
375
Fungicide groundnut
-1000.0 Treatment
- 14863
3586
-0.28 CFA
~w$p~s9roup
By+pital
2
0.0 CFA
999;$*)
0.0 Calorie
Risk dncome
0.0 CFA
Buy rjce
-95.0
CFA/Kg
45
Buy wheat
-278.0 CFA/Kg
77
Sel1 groundnut unshelled
90.0 CFA/Kg
11490
Self groundnut hay
40.0 CFA/Kg
7270
1 ource : LP88 printout.
[*) 99999 or -99999 means no Upper or lower bound
respectively.
FSS is defined as food self-sufficiency.

---

60
TABLE 4.2.4 : WECTIWE FWTION RA#GES IN
THE UPPER
E ZUtdE
Activity
Net return
Unit
Minimum
Maximum
Millet
consumed 1
-20687.0 CFA/Ha
-42645
-11580
Millet
consumed 3
-7149.0
-11694
2770
Millet
consumed 4
-2837.0
CFA/Ha
- 6930
99999(“)
Maize $onsumed 2
-15249.0
-19917
y;;;;(*’
Rainfed rice 2
-30590.0
-41999
rice 3
-14827.0
-25028
-11929
Lowland
Lowladd rice 5
-13542.0
- 16440
-3340
Cottonj 1
-37746.0
-39178
26707
Groundnut 3
-18676.0
CFA/Ha
- 23624
2122
Groundnut seeds
-110.0 CFA/Ha
-384
-71
Cottorj seeds
-110.0 CFA/Ha
-
(“1
99999
30
Buy NRK millet
-90.9
CFA/Kg
-310
Buy N maize
.
-94.0
-140
1;:
Buy N K rice
-115.0
-344
206
Buy T
N K groundnut
-89.0
-130
84
Buy NF% cotton.
-91 .o
-100
Buy urjea
-82.0
CFA/Kg
-107
2:
Buy heirbicide millet
-9750.0
-31709
-644
Buy bdrbicide
cotton 1 -8750.0
- 10182
2450
Buy heirbicide cotton 2 -10000
-11432
2800
Buy iqsecticide
cotton -7350.0
-7827
2058
Fungicjide groundnut
-1000.0 Treatment
-33911
2656
Buy Ca/pital
-0.28 CFA
Capital 2
0.0 CFA
-:;
99&*>
Risk FSS
0.0 Calorie
-14
11
Risk incorne
0.0 CFA
-. 1
0
Buy aifilet
63.0 CFA/Kg
117
Sel1 groundnut unshelled
90.0 CFA/Kg
6;
104
Se11 gkoundnut hay
13.0 CFA/Kg
Sel1 ciotton
95.0 Cfa/Kg
9:
1::
Siource : LP88 printout.
(*) 99999 or -99999 means no Upper or lower bound
respectively.
FSS is defined as food self-sufficiency.

---

61
generation computer language allows knowledge to be represented
in two
different
forms. First,
knowledge cari be expressed with factual
relattonships
between entities
taking an If . ..Else type of
structure.
This representation
leads to the rule based expert system.
Second, knowledge cari take the form of causal relations
between facts
(Predicates),
yielding
the SO- called logic based expert system. This
study makes use of both types of knowledge representation.
Table 4.3.1 (read across columns) shows the status of different
variables
in an LP optimal tableau and the different
characteristics
that those variables
cari have. A "yes" in any ce11 means that a given
variable
has the attribute
of that characteristic.
The set of "yes"
attributes
of a given variable
forms the body of the rule that must be
satisfied
to get a successful matching. For example, a real activity
is in the optimal solution
if:
(1) it is a decision
variable,
(2) has
a lever1 greater than zero and (3) does not have a shadow price (shadow
price equals to zero). A slack is in the optimal
(not binding)
if:
(1)
it is a resource,
(2) has a level equal to zero and (3) has a shadow
price igreater than zero.
TABLE 4.3.1 : VARfABLE
ATTRIBUTES
t
OPTMAL LP TABCEAU
VARIABLE STATUS
-----_--__------_-----------------------------
Real
Real
activity
Activity
Slack
Slack
Characteristics
in
out
in
out
,
Decision variable
w
Yes
Resoukce
no
:2
;ES
Level = 0
çe,
yes
Level, > 0
YES
no
y:,
no
Shadob price = 0
yes
no.
yes
no
Shadop price >= 0
no
Yes
no
yes

---

il
I
j
! 1
g 1 “..e...t<-.--....--
i.__-
.___-.-
-.*
,--.___........-I..._..-.....-.....
..-
“..^..-....<....-....
.
..I
_. “““““‘H Iw,,!L il,&...,H Ke:::
-......-..I,........<...-...-
-
~.........._...
._ ,..i
-..-
..<...__....._.....
62
“”
.
II^
-...
-.
.._~...___.____.
1
1-1
.,...,,..._.,_._,_<..~.~,...,...~
.._.
,......_...
.
..n
8 1 8 /! ! 1

---

63
S&era1 other attributes
should be included in the list
of
characteristics
to allow the expert system (ES) to carry out
sensitivity
analyses on objective
function
coefficients
and resource
rangeci of optimality.
These attributes
are described in the following
list:
(1) The return per unit for the real activities.
(2) The right-hand-side
(RHS) for resources.
(3) The minimum range of optimality
for real activities
and resources.
(4) THe maximum range of optimality
for real activities
and resources.
It may also be convenient,
in the future,
to include labels for
actfviities
and resources in this representation
to make the computer
displays more readable to users.
The expanded list
of characteristics
(see above and Table 4.3.1)
are the cardinality
(number of elements) of the relation
used to store
the LP optimal solution
in a knowledge base. For representational
convenience, the following
modifications
are brought into the data
structure
defined above:
(1) The characteristics
"Decision variable"
and "Resource" are
collapsed into a single variable.
This variable
takes the value "A"
for the former and the value "C" for the latter.
(2) Tbe characteristics
"Level = 0" and "Level > 0" are represented by
an integer variable,
the same precedure is applied to "Shadow price =
0" and "shadow price > 0".
A Illustration
of the knowledge base data definition
used in this
study is shown in Table 4.3.2.

---

64
TA6LE 4.3.2 : KJU'HdLED6E
BASE DATA DEFINITION
'Variable
names
Designation
Type
Var
Activitv
or Constraint
name
Symbol
Type
"A" for"activities
Character
"C" for constraints
Character
Value
Activity
level or Shadow price
Real
Coeff
Return per unit or RWS
Real
Sprice
Shadow price
Real
Unit
Activity
or resource unit
Symbol
Min
Min range of optimality
Real
Max
Max range of optimality
Real

---

CHAPTERV
DIWIOW
W OBJECTIVES
Alto SatfrptY EslrBIIsES
The purpose in this chapter is threefold:
(1) to discuss
the
objectives
formulated in Chapter 1; (2) to discuss the derivation
of
trop supply responses; and (3) to describe the different
functions
avaflable
through the expert system.
5.:1 Discussion of objectives
5;l.l
Stctrtiq
E;Be#tal
This section deals with the impact of varying the starting
capital
levcl on optimal resource allocation
plans. It was hypothesized that
higher starting
capital
levels would shift
the optimal cropping
patterns towards more input intensive
crops. This process is evaluated
by setting
the starting
capital
level at values above the optimality
range of this resource, as determined in the sensitivity
analysis.
For
the purposes of this exercise,
it is assumed that the starting
capital
is doubled in both zones from its initial
levels.
.
Table 5.1.1 shows the new cropping patterns observed for the CPB
zone Jhen the starting
capital
is set at the level of 40000 CFA
francs,.
65

---

66
TAI&E 5.1.1 : CRWPING PATTERN WEN STARTING
CAPITAL - 4WW CFA FRA#CS
IWCEîdTRAL
BASIN
@eturn
-
CFA)
TECMNOLOGIES
TOTAL
-------_--_L------_-----------
CrupS
1
3
4
Ha
%
U-W
!S!e
cons&ption
2.37
U:$'
(;:99,
(:,
{;:;j
(ii)
Groundnut
(0)
(2:60)
(2:60) (40)
S urce : LP88 printout.
9
F:gures in parenthesis
are areas from the base run mode1
Technologies
1 to 4 refer to technical
packages (chapter 3)
The net return increases by 30% and the new cropping patterns cal1
for the following
observations.
Area under millet
decreases by 8% in
favor of groundnut. Within the millet
trop, technology one becomes
more attractive
to farmers while technology three disappears from the
farm plan. The additional
land transferred
to groundnut is used
entirely
in technology one. These facts lend support to the hypothesis
formul:ated before namely, increasing
the starting
capital
level in the
CPB yields
a move towards more input intensive
modules,
Taible 5.1.2 illustrates
the new cropping patterns
observed in the
UPC zone with a starting
capital
level of 50000 CFA francs.
Net return improves by 48% and the new cropping patterns
show the
follo%ing
features.
Millet
land share decreases by 14X, groundnut
share increases by 13% and maize share increases by 1%. Obviously,
there is transfer
of land from millet
in favor of groundnut and maize.

---

67
TABLE 5.1.2 : CROPPING PATTERN MiEN STARTING
CAPITAL
IN TE U
E
(Return - 125836 CFA)
TECHNOLOGIES
TOTAL
---_--------_-_---_--*--------
CrOpS
1
2
3
4
5 Ha % Upland
Mille~ for
consut#ption
.93
.93
24
LW
.08
w34)
y
Maize consumed
(*74)
.22 (O*O) .2g LJ;’
yo)
Lowland rice
(058)
(.02) (.60) *
Groundnut
1.78
46
(1.30)
U;S”
(93)
Cotton
(-36) (9)
Source : LPBB printout.
Fi;gures in parenthesis
are areas from the base run mode1
Technologies
1 to 5 refer to technical
packages (chapter 3)
* Percentages are calculated
only for upland crops.
While millet
and groundnut are grown in the most intensive
technology,
maize and lowland rice show a preference for the less intensive
modules.
5.1.2 Maroinal w
The impact of cultivating
marginal lands on the optimal enterprise
m-ix prevailing
in the zones involved was also investigated.
The amount
of marginal land divailable was set at 0.3Ha and 1.6Ha in the CPB and
UPC zanes, respectively.
This investigation
is performed by introducing
new trop producing
activities
for sale (which use marginal land) in the initial
LP

---

68
tableau. This type of land is more labor intensive
than normal land,
modelled
by doubling the labor input-output
coefficients.
As pointed
out in Chapter One, marginal land is only used for the production
of
millet
or maize in this model.
Table 5.1.3 shows the new cropping patterns observed for the CPB
zone.
TABLE 5.1.3 : CWPPING PATTERNS
I$IE# WWINAL LA#O
IS CULTIVATED
IN CfNTRAL PfAJ4UT
BASIN
(Return - 176653 CFA)
TECHWLOGIES
TOTAL
ClTtPS
1
3
4
Ha
%
(1.48)
‘: l $’ (1) y3 ‘4;’
Groundnut
(2kO)
(2:60) (40)
Millet
for sale
(tnargi/nal land)
0.30
0.30
Sdurce : tP88 printout.
Figures in parenthesis
are areas from the base run mode1
Technologies
1 to 4 refer to technical
packages (chapter 3)
Mode1 results
show that cultivating
marginal land in the CPB has
reduced the normal land share used by millet
by 1% in favor of
groundnut. A new millet-producing
activity
enters the optimal solution
under technology three and net return improves by 3%.
Table 5.1.4 show the corresponding results
for the UPC zone.
From the mode1 results,
it is also observed that in this zone,
cultivating
marginal land has reduced millet
(for home consumption)
normal land share by 4% in favor of groundnut (1%) and rainfed rice
(3%). Al1 marginal land available
is used to grow millet
for sale
under technology three. In comparison with the base run solution,
net
farm return has improved by 24%.

---

69
T@LE 5.1.4 : CROPPING
PMTERWS YHEN BARGIWAL
LMD
IS CULTIUATED fW UPPER CA!YWHCE
(Retum = lQ!i210 CFA)
TECHRUtOGIES
. TOTAL
__-_-------___--_*-_---------
Crops
1
2
3
4
-Ha
49
0.27
0,67
0.38
1.32
34
(0.16)
(0.58) ‘;J:’
y;’
VO)
Maize consumed
0.41
y;'
(0:oo)
t;:;:'
y
Rainfed rice
(0:03)
‘f;;’
348’
Groundhut
1.32
‘p;’
(1.30)
(;:;gJ (;Y
Cotton
(0:36)
(0:36) (9)
Millet, for sale
(marglpal land)
1.60
1.60
So'rce : LP88 printout.
6
Fi: ures in parenthesis
are areas from the base run mode1
Technologies
1 to 4 refer to technical
packages (chapter 3)
5.J.3 &R@& nf pooulation
srowth
Th,is section seeks to evaluate the effect of population
growth on
farm resource allocation
plans. An increase in farm population
size
affects the consumption side of a typical
farm by raising
the family
calorie
requirement to be met by cereals.It
also affects
the
production side of a typical
farm by increasing
the total
family labor
force available
for field work.
The process is carried out by making predictions
on farm
popula'tion levels.
For the purpose of this exercise,
a 5% increase in
farm population
size is assumed in this section.
The next step
involves the recalculation
of new parameters for mode1 coefficients.
These coefficients
concern the minimum food self-sufficiency
needs,

---

70
the farm labor force available
in a11 periods and the nutritional
needs of the farm unit to be met by cereals.
The following
describes
the results
obtained in both zones.
For the CPB, Table 5.1.5 shows mode1 results
with a 5% growth
rate In the population
size. Compared to the base run solution,
land
share for millet
produced for home consumption increases by 3% while
groundnut land share decreases by the same amount. A sustained growth
in farm population
size alters prevailing
cropping patterns.
More food
trop (millet)
is grown to support a higher demand for food.
For the UPC zone, Table 5.1.5 illustrates
mode1 results
with a 5%
growth rate in farm population
size in comparison with the base run
Soluti:on. Here again, millet
for home consumption land share increases
by 3% while groundnut land share drops by 3% and the other crops' land
share remain constant.
The excess demand for food is satisfied
by
growiqg more millet
(for home consumption) rather than groundnut (for
incorner).
5.1.4
Food self-sufficiencv
(m)
rate%
This exercise seeks to determine the impact of varying the FSS
rates during bad rainfall
years on farm resource allocation
plans, assuming that on average, 80% of FSS is desired on good
rainfall
years.
Th;e process is carried out through series of mode1 simulations
by
calcul'ating,
at each step, a new FSS risk coefficient
(see table
3.5.3).
The farm net return is taken as a performance measure of
system behavior in response to different
FSS rates. As pointed out
earlier,
the base run LP problems assumed that 30% FSS rate was
desired during bad rainfall
years (worst states of nature).
Therefore,

---

71
TABLE 5.1.5 CROPPING PATTERNS WHEN POPULATION GRWTH
RATE - 5% IN CENTRAL. PEA#U BASIN
TECHNOLOGIES
TOTAL
------__------------___I______
CropS
1
3
4
Ha
%
b.r!a '1
d (Ha)
Mille iP-- consumed
1.50
1.60
4.10
63
(1.48)
y;'
'; $1
y
Groundnut
(2:60)
(2:60) (40)
TABLE 5.1.6 CRQPPIN6 PATTERNS WHEN POPULATION GROUTH
RATE - !% IN UPPER
E
TECHNOLOGIES
TOTAL
----------------_---______I_
Crop$
1
2
3
4
Ha
49
Are&
Mille
0.23
0.63
0.75
1.61
41
(0.16)
(0.58)
(0.74)
Maize iconsumed
0.74
$J3
Rainfejd rice
(0:03)
Groun+ut
1.16
y5'
(1.30)
Cotton
(0:36)
So/urce : tP88 printcwt.
Fi/gures in parenthesis
are areas from the base run mode1
Technologies 1 to 4 refer to technical
packages (chapter 3)

---

72
mode1 simulation
Will start from that rate and progress by increments
of 1% until
an infeasible
solution
is reached. The rest of this
section discusses the mode1 results
in both zones,
Table 5.1.7 summarizes the results obtained through several
simulation
cycles; the following
comments are made for both zones.
Incredsing the rate of cereal FSS during bad rainfall
years leads, in
both zones, to a deterioration
of net farm return.
The desire to
achieve 7% of FSS during worst years is not an attainable
objective
in either zone (infeasible
solution).
The UPC zone has a better
potential
to protect itself
against food deficit
than the CPB zone. In
practical
terms, a typical
farm in the CPB cari only achieve a maximum
of 40X of FSS during bad rainfall
years while its counterpart
in the
UPC cari sustain 5% of FSS during bad rainfall
years.
5.12 Supply responses
In underdeveloped agriculture,
supply responses are assumed to be
equivalent
to response of acreage under cultivation
to changes in
economic and non-economic factors(
Subrata and Ken, 1984).
In this section,
price assumptions about farm products are
discus:sed first,
followed by the derivation
of normative supply curves
under jalternative
technologies,
5.12.1 Price assumptions
Based on assumptions about the relationship
between price and
quantity
supplied in the regional market, a set of prices
for millet,
maize and rice is determined.
Producer prices of groundnut, and
cotton,
and the rice consumer price are based on government price
setting
policies.
Six price levels are assumed in this study, starting
from the actual

---

73
TABLE 5.1.7
NET FARH RETURN AT DIFFERENT REQUIRED
FOW SELF-SUFFICIEHCY
LEVE@
(Unit - CFA)
WORST YEAR FSS RATES
-1__1_--------------_______3____________------
Zones
30
40
50
60
70
Centr;ll Peanut
Basiin
171656 163194 -370000 -11856557 infeasible
Upper :Casamance 85035
60346
23388
-120247 infeasible
prices used in the base run solutions,
progressing
by increments of
20% up to the level where a11 prices are doubled. Prices are a11
expressed in financial
terms. Tables 5.2.1 and 5.2.2 show the price
levels used for the CPB and UPC zones. Only two crops are considered
in the CPB because rice is not produced in that zone and the consumer
price ,of rice is under goverment control.
5.i2.2 Normative ~u~plv curves
The mechanism used to derive the trop supply responses under
alternative
technologies
is explained at this point.
Cropping
patterns,
for price vector one (base price) are already obtained from
the base run solution.
The next step consists of introducing
price
vector two in the LP tableau for the cereal buying and selling

---

74
activfties
whose prices are determined in the regional
market, The
proce+re
used for that is described below:
(1) The new prices are introduced as coefficients
of the objective
function
for the cereal buying activities
(minus sign) and the cereal
selling
activities
(positive
sign).
(2) Those same prices are also used as coefficients
of the capital
one
constraint
(borrowed capital)
under the cereal buying activities
with
positive
signs.
(3) Price vector two is also entered as coeeficients
of the repayment
of the borrowed capital
constraint
under the cereal selling
activities
with :negative signs.
The mode1 is run with those modifications
and the optimal cropping
patterns are derived from mode1 results.
This process is repeated for
price ivectors three to six.
Tables 5.2.3 and 5.2.4 show the results
obtained.
The following
are the most important points to note for the two zones investigated.
For the CPB, results
show that no land competition
exists between
crops through the different
price levels assumed, Land share between
millet
and groundnut remains constant a11 along. This rigid
situationis
partly due to the fact thàt millet
is the most profitable
food orop grown in that zone, and groundnut is the only cash trop. Any
trade-off
between them cari only occur at high millet
price (150%
increase) as revealed by the sensitivity
analysis.
(A cowpea
production
activity
is included in the model, but it does not enter
the optimal solution,
The analysis of the trop gross margins (Martin,
1987) reveals that, although cowpeas have great potential
in terms of
gross return,
it is the most labor intensive
trop in CPB. This fact

---

75
TJiBLE 5.2.1 CROP P#UCE
VECTOR
IN ME CENTRAL
PEANUT
BASIN
(Unit = CFA/Kg)
L$bels
% increase
Millet
Maire
PFl
base
62.0
79.0
'PF2
20
74.4
94.8
:PF3
40
86.8
110.6
PF4
60
99.2
124.4
'PF5
a0
116.6
142.2
,PF6
100
124.0
158.0
T/'jBLE 5.2.2 CROP PRICE VECTOR -IN THE UPPER CASAMANCE
(Unit - CFA/Kg)
Ldbels % increase
Millet
Maize
Rice paddy
iPF1
base
63.0
fia.0
85.0
'PF2
20
75,6
81.6
102.0
:PF3
40
88.2
9.5.2
119.0
IPF4
60
100.8
108.8
136.0
:PF5
a0
113.4
122.4
153.0
iPF6
100
126.0
136.0
170.0

---

76
makes it too unattractive
to be included in the optimal enterprise
mix.
Tdble 5.2.4 shows for the UPC interesting
aspects of land
çompetition
between crops. After price vector two, groundnut
disappears from the enterprise
mix and maize jumps from 19% of land
share to 60%. This result mainly from the 40% drop in groundnut land
share,: lends support to the hypothesis
that maize cari potentially
be
a leading trop in the UPC, as groundnut did in the CPB, given enough
priceincentive.
Rainfed rice also, after price vector two, leaves the
optimdl plan. Within the lowland rice,
land is transferred
from
techndlogy two to five.
A tentative
explanation
of this would be that
inputç,
currently
used for rainfed trop, should have been transferred
to lowland rice, making it worthwhile to grow more rice for
consumption under technology five.
Millet
land share decreases by 7%
and then remains constant at 30% share in the last two price levels.
Fdr pricing
policy purposes, table 5.2.5 shows the acreage
resporïses derived above, translated
in terms of quantities
produced
for diifferent
crops in the UPC zone. Those quantities
are calculated
by multiplying.the
respective
trop acreages by their corresponding
yieldsi and finally
by the number of producers in that zone.
543 Expert system functioni
This section highlights
the main features involved in the
implementation
of the LP expert system (ES) developed in this study
for the interpretation
of a LP optimal solution.
5.3.1 Design considerations
As defined in chapter four, an ES shell has two major components:
a user interface
(UI) and an inference engine (IE). The following

---

77
ThKE 5.2.3 PRICE LEVELS
AMI CROPPIWi
PATTERNS
IM CENTRAL
PEAWT BASIW
(bits
- Ha)
TECHNICAL PACKAGES
TOTAL
_-----_---_-_------_________c____
Gro#s
1
2
3
4
5
Ha
%
PEa
Mill$t
1.48
1.42
1.0
3.90
:o
Groubdnut
2.60
2.60
LEZ
Mill@
1.48
1.41
1.0
3.89
:1
Groutjdnut
2.61
2.61
.m
Milljt
1.48
1.41
1.0
3.89
59
Grouqdnut
2.61
2.61
41
Mill&
1.48
141
. 1.0
3.09
Grourldnut
2.61
2.61
4:
?I?i
Milbjt
1.48
1.41
1.0
3.89
Groujdnut
2.61
2.61
2
m
Millqt
1.48
1.41
1.0
3.89
59
Growjdnut
2.61
2.61
41
Soiurce : LP88 printout.

---

78
T&E
5.2.4 PRICE LEVELS AMI CROPPING PATTERNS
IN UPPER
E
(bits
- Ha)
TECHNICAL PACKAGES
TOTAL
----"--------------________c_____
% of
Cr0p
1
2
3
4
5
Ha
upland
Mill+
0.16
0.58
0.75
1.49
38.0
Maiz‘
0.74
0.74
19.0
I
Rain ,ed rice
0.03
0.03
0.8
Lowldnd rice
0.58
0.20
0.60
Groujdnut
1.30
1.30
33 0
Cottoll
0.36
0.36
9:2
!?u
Mille$
0.16
0.58
0.75
1.49
38.0
Mai24
0.74
0.74
19.0
Rainffed rice
0.03
0.03
0.8
Lowl nd rice
0.58
0.20
0.60
1
Grou dnut
1.28
IL.30
3; 0
Cottdn
0.36
9:2
pF3
Mill&
0.16
0.29
0.75
1.20
31.0
Mait$
2.34
2.34
60.0
Lowlqnd rice
0.09
0.51
0.60
Cottojn
0.36
0.36
i.0
OFq:
Mill&
0.16
0.29
0.75
1.20
31.0
Mairi
2.34
2.34
60.0
Lowlajnd rice
0.09
0151
0.60
*
Cottdn
0.36
9.2
pF5
Mill&
0.15
0.28
0.75
1.18
30.3
Maim
2.35
2.35
60.3
Lowlahd rice
0.26
0.34
0.60
+
Cott+
0.36
0.36
9.4
Mille$
0.15
028
0.75
1.18
30.0
Maizej
2.36
2.36
61.0
l

Lowlyd rice
0.26
0.34
0.60
Cottop
0.36
0.36
9.0
Wrce
: LP88 printout.
(*!) Percentage'area
are calculated
only for crops cultivated
in rainfed opland.

---

79
discussion
describes some special features inherent to the design of
the E$ shell developed in this study.
The knowledge derived from the LP solution
is represented under
the form of statements of facts between different
LP variables
(predicates).
This type of representation
is more adaptable to the
nature of a LP optimal tableau. The information
is stored in a
computer disk file
as a permanent storage device under the form of a
database. At the beginning of every consultation,
the database is read
into Computer RAM (Random Access Memory) for better performance
(speed). This data management technique is known as a dynamic database
(Borland,
1986). The LJI developed here handles the management of the
database through a user oriented menu. Operations performed are
currently
limited
to the tasks of adding, listing
and deleting
data
from Che knowledge base. The program listing
of the LJI is shown in
Appendix Table 3.
The inference engine designed for this ES is shown in Append ix
Table 4. Its implementation
is based on rules applying to the
interpretation
of a LP optimal tableau.
Rules are executed by using a
"backward chaining"
structure,
meaning that Prolog' s strategy
is to
find a goal by proving that a11 sub-goals in the body of a rule are
satisffied.
5.i3.2 Expert system functions
Tbe expert system designed in this study is "menu driven";
it
perforjns four basic functions
described as follow:
Option one is devoted to the analysis of the LP prima1 solution.
The Es Will display the optimal and nonoptimal real activities
with

---

80
T@LE 5.2.5 SUPPLY RESPWSES AND PRICE EFFECTS
I# UPPER
E
(Quantities
- o100 Kg)
(Price
= CWW
dillet
Maire
Rice
Rainfed
Lowland
Price quantity
price quantity
price
quantity
quantity
63.0
12570
68.0
11641
85.0
289
9470
75.6
12570
81.6
11641
102.0
289
9470
88.2
10662
95.2
36810
119.0
0
5492
100.8. 10662
108.8
36810
136.0
0
5492
113.4 : 10468
122.4
36968
153.0
0
6348
126.0
10468
136.0
37125
i70.0
0
6348
Squrce : LP88 printout.

---

81
the pertinent
information
related to them. A short interpretation
of
the listed
information
is also provided.
Option two performs the analysis of the dual solution
of the
optimhl LP tableau.
Information
related to the shadow prices of the
bindihg constraints
and to the slack resources are displayed
in this
optioh. A brief
interpretation
of results
is also-given.
Options three and four perform sensitivity
analyses on resource
and objective
function
ranges of optimality.
In each case, pertinent
information
is displayed along with a short interpretation
of listed
infor$ation.
T;Lble 5.3.1 shows a sample printout
produced during a
consultation.
A listing
of the complete computer program written
in
TurbojProlog
is given in Appendix Table 4.

---

82
TABLR 5.3.1
RXPRRT SYSTRH SANPLR PRilhTTOUT
o~:~::-
* ~~.~.~*~~~.~~~~~~~~~~~~~~~a~~~;~~~~~~~
. . .._..
.,. . .._ ,. ., .1
1.
PRIMAL SOLUTION ANALYSIS
2.
DUAL VALjUES ANALYSIS
3.
RESOURCEj RANGE ANALYSIS
4.
OBJECTIVjE VALUES ANALYSIS
5.
EXIT TO ;PROLOG SYSTEM-
X*2e*$**tX**jd*#***************
ENTER YQUR CEOIiCE
3
Constr-aint
Name; :
TERRE
Its shadow prick
is
:
84686
CFA
Range of optidllty
:
5.6
8,8(Ha>
57% variation
.in
TERRE
is acceptable
Press space bar . , . when done
Constraint
Name; :
SEMARA
Its shadow pricje
is
:
140.8
CFA
Range of aptima$ity
:
-2137
313cKg/ha>
115% variation;
in
SE&!RA
is acceptable
Press space bar: . . . when done
Constraint
Name
:
NPKl
Its skiadow price
is
:
356
CFA
Range of optimaility
:
-205
136(Kg/ha>
166% variation:
in
NPKl
is acceptable
Psess space bar' , . . when done

---

This study has used a linear programming approach, taking
into
accoutit the risk under which Senegalese farmers operate, to
investigate
cropping patterns and technologies
most profitable
in two
agricultural
zones. This chapter summarizes the major findings
of this
study,~ and highlights
some policy recommendations related to those
findiflgs.
The last section suggests some areas for further
research to
improqe mode1 performance.
6.,1 Sunrary of findings
Several hypotheses have been tested throughout this study. Mode1
results
support the following
points:
(1)
Increasing the starting
capital
level in the two models has two
'major ~effects. First,
there are alterations
in acreage grown, and
intensive
technologies
become more interesting
economically
for millet
and groundnut. Second, land is transferred
from millet
to groundnut in
the CPB and from millet
to groundnut and maire in the UPC.
(2) Xhe introduction
of marginal lands into the two models leads to a
reallocation
of land between crops. In the CPB, millet
grown on
non-mairginal land is reduced in favor of groundnut. This decrease in
83

---

84
millet
share is compensated by the millet
produced in marginal lands.
The same effect
is observed
in the UPC: millet
grown on non-marginal
land $ecreases to the profit
of groundnut and rainfed rice in
partibular.
(3) The study of the impact of farm population
growth on farm
resource allocation
decisions
reveals that a 5% increase in farm
populdtion
Will change cropping patterns.
More food-crops are grown to
support the higher demahd for food while areas used for cash-crops
decreilse. Mode1 results,
in both zones, lend support to this point.
(4) Varying the worst years food self-sufficiency
(FSS) rates reveals
that a desire to secure 70% of food needs during bad rainfall
years is
not an attainable
goal in either zone. However, the UPC zone cari
achieve a maximum of 50% of FSS during bad rainfall
years while the
CPI3 secures a maximum of 40% of FSS rate during bad rainfall
years.
(5) Supply responses are derived under six price assumptions
concerning cereals when cash-trop prices are kept constant.
In the CPB
no land competition
between millet
and groundnut was observed through
the price ranges. This fact leads to the conclusion
that doubling
prices does not affect cropping patterns.
Results from the UPC mode1
show Èhat maize area increases substantially
given a 40% price
increaise. This finding
lends acceptance to the idea that maize cari be
a leading trop in the UPC zone.
6./2 Rtxamnendations
This section suggests some major issues decision makers should
considier in their interventions
in the agricultural
sector. These are
based On mode1 results through the different
simulation
exercises.

---

85
Policies
aiming to increase the starting
capital
available
to
farmers Will favor a shift
towards intensive
technologies.
Planners
shouldj take this fact into account and make agricultural
inputs
availaple
to farmers at the right time.
Exkensive development policies
bringing marginal lands into
cultiv$tion
could reduce millet
produced in normal land and expand
rainfed rice and groundnut production,
in those regions similar
to the
UPC. Tbe magnitude of this reduction Will however, depend on the
amount: of millet
produced on marginal land. This situation
cari
deteribrate
the rural terms of trade through higher millet
price in
the 10~9 run,
Anjr pricing
policy should be introduced with tare in those regions
with cropping patterns similar
to the UPC. High producer price Will
sharply increase maite produced and decrease groundnut production.
If
marketling alternatives
are not anticipated,
the long run price of
maize kil1 suffer.
A high annual increase in the rural population
(5% used) Will lead
to a higher population
pressure on the available
land. This situation
would change farmer strategies
with regard to resource allocation.
Subsis!tence requirements Will dominate the minimum income need and
croppihg patterns Will move towards more food than cash crops.
Puksuing higher rates of national
food self-sufficiency
without
taking' into account regional potentials
and disparities
cari bring
about many side effects.
As seen in this study, farmers want to
achievp certain
FSS goals even in the worst years. The objective
to
secure; 70% of FSS during bad rainfall
years is not sustainable
in the
two zones studied.
A national
goal to caver 80% of food needs may
conflibt
with regional potentials.

---

F;ermers' desire to have a minimum income is another point that
decision makers must keep in mind in national
agricultural
planning
efforts.
Food self-sufficiency
at any cost may not be a desirable
social objective.
A food security
perspective,
linking
both food
self-qufficiency
goals and income needs, may lead to a more optimal
resource allocation.
Mode1 results
show that net returns are higher at
lower FSS levels and that an optimal mix between food-crops
and
cash-trop cari better maximize farmers' objectives.
6*3 Areas for further
research
This section focuses on possible improvements of this model. The
discussion
includes concerns related to mode1 management, mode1
structure
and inclusion
of other sectors in mode1 activities.
IC is known at the present time that agricultural
inputs are not
used %n Senegal as suggested by agronomie research while mode1 results
are based on the assumption that inputs are available.
In order to
accommodate a particular
situation
and use the mode1 as a diagnostic
tool, 'more investigations
should be made about actual technologies
used by farmers.
Under present conditions,
cotton is expected to be a very
promising trop in the UPC, yet mode1 results
do not reflect
that fact.
The cotton trop budget should be reviewed by agronomists and other
knowledgeable people in order to improve mode1 coefficients.
The livestock
subsector is not represented directly
in the mode1
although the trop selling
activities
take into account a possible
integration
between agricultural
and animal feeding. This fact is
reflected
in the groundnut and cotton selling
activities,
It would be
more ippropriate,
in the future,
to investigate
the possibility
of
incfudiing that subsector in the mode1 activities.

---

87
The fishery
subsector also is not taken into account by the modef.
This is not a limiting
factor under present conditions
since it is
assumed that 65% of calorie
needs are satisf ied by cereals.
However,
mode1 'performance could gain credibility
by includ ing activities
and
constraints
related to this subsector in the mode1 structure.
This
could be done through fish purchasing activities
to complement food
needs'associated
with food habits constraints
prevailing
within
regions.
A.representative
farm concept is used to mode1 farmer behavior in
a given zone. This type of representation
does however introduce
aggredation biases into mode1 results
in an upward direction.
It would
have been more realistic
to mode1 at least three categories
of farms
in each zone, ranging from large to small farms. This consideration
Will involve investigating
further
farm structures
prevailing
in the
different
regions used in the mode1 in order to reduce the magnitude
of thc/se biases.
Yi!etd estimates for rainfed crops are based on the amount of rain
observed through a period of twelve years and on rainfall
distribution
across the rainy season for each year. More formai methods of
estimating
yields are available
through the Comprehensive Resource
Inventory and Evaluation
System (CRIES) project which is currently
undeway in the Department of Natural Resources at Michigan State
University
(see Schultink,
1987). The yield mode1 of the CRIES project
is a microcomputer based simulation
mode1 capable of predicting
yields
for a large number of food and export crops based on agro-ecological
zones. The yield mode1 and the other modules available
in the CRIES
project are widely being used world-wide,
in more than twenty nations.

---

APPENDIX TABLE 1
:
CENTRCK PEAPIUT BFtSIN
ItNTTfllL
TFIBLEHU
._
_
^
SIGN
RHS
SIGN
P#IC1
Ii&
PM163
PH IC5
p1IIVl
PMIU2
PWIU3
RETURU
0 RETW
ci
-36c3n
-36cKl
-3298
-3600
-36m
-36cKl
-36clo
TERRE1
<=
6.5
TERRE1
<=
6-S
1
1
1
1
1
1
1
SEHfW?Fi
<=
0 ÇEtiRRfl
<=
a
SEmIE
<=
0 SEtaIIE
<=
0
NPKI
<=
ONFKL
<=
a
1CJO
100
100
100
N+w4
<=
DNPK4
<=
cl
llI!EE
<=
OLtREE
<=
0
50
50
Nt INS1
<=
0 NII%L
<=
1
Nf fi’52
<=
0 NfINç2
<=
ii
FwslARA
<=
D FONRRR
<=
El
MO1
<=
41.2
MO1
<=
41.2
5.5
8
2
5.5
ta2
<=
41.2
Ho2
<=
41.2
7.5
13
7
10
1
7.5
13
?a3
<=
41.2
f%33
<=
41.2
3
2
2
2
3
2
MD4
<=
41.2
MO4
<=
41-2
2
M&!l
<=
137 MUR1
<=
137
3
3
2
2
3
3
TRl
<=
1Q TFtl
<=
10
Cl-5
TR2
<=
10 TF12
<=
10
1
1
1
1.5
0.5
1
1
TR3
<=
10 TA3
<=
10
1
1
1
1
1
1
TR4
<=
10 TR4
<=
10
1
ÇCUP
<=
2mo
çcw
<=
2lxKxl
CftPl
<=
0 CWl
<=
0
GflP2
<=
cl CFP2
<=
0
RCW
<=
D RCRP
<=
0
GIWIIL
<=
0 mJHIL
<=
0
-652
-558
-417
QVNIEP
<=
0 QUNIEP
<=
QYNIEF
<=
0 QUNIEF
<=
:
QYFlRFEG
<=
0 QWRR6
<=
0
QWeF
<=
D WWF
<=
a
OEUF0
>=
0 DEUfll
>=
0
-1103
-977
-727
-1103
-656
581
465
349
AUTE.l
>=
4270 fWTO
>=
4270
1685
1442
1076
1685
656
0EuR1
>=
0 MUR1
3=
0
-18300
- 16652
-12342
OEQR2
3=
0 DEVR2
3=
0
-18800
- 17052
-12642
REUEIW
3=
lOCbJM3 REVENU
3=
lDw31#1
34652
25642
NUTRI
>=
5103 WTRI
3=
5103
1432
1226
915
1432
MIN4
>=
0.05
HIN4
>=
Q,lEi
1
MRX4
<=
<=
1 tmx4
1
FIIPWIL
>=
1026 FIINMIL
3=
lld
455
290
455
177
MftXHIL
<=
1710 MFWIIL
<=
1710
455
zz
290
177
MHWFfI
3=
Cl NINMFtI
>=
a

---

45
5;
ii”
w- -pc-
l
VVVVVVVVVVVVVVVVVVVVVVVVVVVAAAhhAhVhVh II
II
II
Iv
N
Il
II
II
II
II
II
II
II
II
89
II
II
II
Il
II
Il
II
II
II
II
II
II
I 3
II
I aR
II
II
II
II
II
II
II
II
II
II
II
II
II

---

.,
- I “. .
.“_..
I. .
.._
~.
. ..“”
_
SIGN
RHS
FttiJKi
fWREE
FWINl
RNIN2
f&!fjFU
RH01
RH02
RHn3
Rm4
l3MclF!l
RETURN
u
-79.5
-72-5
-6750
-5lcw3
-1iXXl
-SO0
-5m
-5otl
-5fm
-500
TEFSEl
<=
6.5
SEHRRR
<=
0
SENNIE
<=
0
NPKl
<=
NF%3
<=
E
-1
ina%
<=
cl
-1
NI INS1
<=
-1
NIINS2
<=
:
-1
çfJNfwFi
<=
-1
PI01
<=
41.2
-1
Ho2
<=
41-2
-1
tm3
<=
41.2
-1
<=
41.2
-1
HllRl
<=
137
-1
TFIl
<=
LCI
TR2
<=
10
TA3
<=
10
TR4
<=
10
SmP
<=
2clfxm
C@l
<=
0
6750
cFiP2
<=
0
79.5
72.5
RCRP
<=
0
WtlIL
<=
0
CWNIEP
<=
0
QWIEF
<2
a
QURRFIG
<=
a
QURRRF
<=
0
Df3JRl
>=
a
RUTO
>=
4270
DEuRl
.=
cl
cm&!2
>=
cl
REVENU
>=
IOMJDO
MJTR I
>=
5103
t’UN4
>=
0.05
bmx4
<=
1
MINMIL
.=
lQ26
tlfwiIL
i=
1710
MINHRI
>=
D

---

APPTWDTX
TAt3T.E
1 -
_.
SIGN
RHS
RCRP
RISKR
RISKR
CW’T 1
CfWT2
fwIL
-. ‘.
fWRIS
RRIZ
fBLE
‘.
VMIL
RETURH
0
-0-28
-62
-79
-163
-276
62
TERRE1
<=
6-5
SEEIFIRA
<=
0
SEHHIE
<=
0
HPKl
<=
0
NF%4
<=
0
UREE
<=
0
NI INS1
<=
0
NI INS2
<=
0
FtI@tWFH
<=
0
<=
41.2
2:
<=
41.2
m3
<=
41.2
<=
41.2
El
<=
137
TRl
<=
10
m2
<=
10
TR3
<=
10
TA4
<=
10
!XFtP
<=
2mOo
1
1
CFP1
<=
0
-1
-1
62
79
163
278
miP2
<=
0
-1
RCRP
<=
0
1
-62
QWIIL
<=
0
1
RVNIEP
<=
0
WNIEF
<=
0
BVRRRG
<=
0
MIARRF
<=
0
DEVAl
>=
0
1
#U!T0
>=
4270
-0-63
OEURi
>=
0
1
OEUR2
>=
0
1
REUENU
>=
iOOOO0
-1
WTRI
>=
5103
3.15
3.17
2.42
3
HIN4
>=
0.05
mx4
<=
1
ttIt4MIL
>=
1026
1
fiWMIL
<=
1710
1
HINPFII
>=
0
1

---

. .^ .
I..“.
SIGN
RHS
UNIÉP
‘,&ËF
UfV?i&
GiFsF
RETURN
0
79
3Q
90
40
-9999
-9999
TERRE1
<=
6.5
SEHRRR
<=
SEfWIE
<=
ii
WKl
<=
NPK4
<=
“0
UREE
<=
a
HI INS1
<=
cl
HI INS2
<=
0
FfrïNRRR
<=
0
Ht31
<=
41.2
Ht32
<=
41.2
nu3
<=
41.2
‘0
<=
41.2
;3
EIDRl
<=
137
TRI
<=
10
TA2
<=
10
TR3
<=
10
TF14
<=
ta
!xw
<=
2clocm
CRPl
<=
0
cFtP2
<=
0
RCAP
<=
0
-79
-30
-90
-40
QWIL
<=
cl
CK+WIEP
<=
0
1
QWIEF
<=
0
1
QVRRRG
<=
a
1
WRRRF
<=
0
1
llEufl1
>=
0
fWTO
>=
4270
UEVRl
>=
D
nEvR2
>=
cl
REUENU
>=
100000
1
MITR 1
>=
5103
HIN4
>=
0.05
HRx4
<=
1
HINHIL
>=
1026
WIXHIL
<=
1710
HINMFII
>=
0

---

APPENDIX TABLE 2 - Cantinued
RHS
PRCl
-?zEY
.!ibf%42
<=
.u
-33t505
<=
1
TERRE2
O:B
1
1
1
SEPIARA
<=
0
ÇEHCOT
:;
NPKl
<=
ii
100
100
NPK2
<=
200
100
NPKâ
<=
0
150
75
75
Nez4
NPKJ
<=
ii
UREE
<=
<=
0
100
50
50
50
200
100
HIHERB
ZZ
1
1
RIHERB
:=
0
1
1
1
HEC1
HEC2
<=
:
INC
<=
<=
0
FRNf%fl
<=
0
104
19.5
45
11.5
1s
9
3
23.5
LE:
<=
SO15
925
98
11.5
502
49
45
14
17
9
28
14.5
b.0
MoRl
=
&
122.6
61
48
48
35
30
s=-
HoR2
98
TF11
<-
Y
Y
5
4
1.5
6.5
5-S
SCRP
<z
<= 25Oz
CRP 1
CFIP2
:
RCAP
<=
QVHIL
<=
D
-1290
-898
-661
-474
QVHRIS
<-
0
-1908
-1581
QVRPLU
<E
QMP
<=
D
QVARRB
<=
0
QVftRRF
<=
QVCOT
<=
0
DEVAl
>=
-2498
-1668
-1668
-1147
-877
1627
930
697
46s
804
804
RUfO
>=
5165
298
1
2152
2152
1630
1360
DEVRl
>=
-19027
-20388
-14602
-11572
-76163
-54847
DEVR6
=
ii
-20427
-21188
-15202
-11972
-70163
-48847
REVENU
;=
85ooo
82027
56388
41602
29572
128663
107347
NUTRI
>=
64g9
2534
1829
1829
1386
1156
HIN4
>=
0.0s
HRx4
<=
O-75
HINMIL
>=
720
HfWMIL
<=
1440
HfNMf31
>-
HRXHRI
<;
2:
NINRIZ
>-
1047
756
573
478
MFU(RI
2
&
z:
1047
Ez
756
573
478
HINBLE
>=
0
MAXELE
<=
120

---

APPENDIX TABLE 2 - Continued
-!?t!ggL. ??!gg7.
.-. TERRE1
1
1
1
TERRE2
1
1
1
SERRRA
160
160
160
SEF?CCIT
NPK1
NPK2
NPKâ
1fXJ
50
150
75
75
NPK4
120
80
NFKS
100
50
100
50
50
lziERI3
RIHERt3
1
1
1
1
HEC1
HEC2
EklRR
1
1
1
no1
12.5
19.5
:.
MU2
17
11.5
SO!5
5045
49
45
:zi
::
z
6
FIOR
$4
28
61
40
48
35
30
15
21
20
16
;f:
HOU2
TAI
4
4
9
9
5
4
1.5
SCRP
CFP1
cRp2
RCRP
QVNIL
uvrw1'5
QVRPLU
-1162
-967
-1850
-1335
-1335
-1012
-844
iiizzz
-1414
-1097
-993
-782
QVRRRF
-2a32
-1750
-1575
-1227
QVCOT
DEVRl
484
484
484
404
484
AUTO
OEVR
1
-98728
-82229
-131716
-879e4
-87904
-60466
-46240
-03288
-53948
-53511
-41249
DEVRB
-98728
-82229
-1232116
-79484
-79484
-SI?X36
-37740
-72346
-43846
-45089
-34514
REVENU
98728
82229
157216
1134a4
113484
85986
71740
156138
121148
109411
86049
NUTRI
6489
nu44
0.05
MAX4
0.75
FlItelIL
720
HfWtlIL
1440
?lINrlRI
4f.m
RFWMRI
720
PlINRIZ
WXRXZ
ZE
ttIN6LE
0
WSXBLE
120

---

APPENDIX TABLE 2 - Coûtinued
PCOTI
2s
-3?W46-
<=
1
<=
<=
<=
50
50
50
<=
g
<=
<=
150
100
100
<=
50
!3a?
=
<<=
<=
<=
:
:
5
5
5
2
=g
<=
24.5
23.5
46
34
<=
6
.ij
<=
2.5
2.;
2.:
:3
h
<=
23
<=
655
545
2
CRPl
<=
z
<=
<=
ClVPlIL
<=
PrltoRIS
<=
CIWWLU
=
2=
OVRRAG
<=
QV-F
<=
aVC30f
=
-1704
-1430
-86
1
-512
DEVAl
;=
AUTO
=
IEVRl
;=
-46932
-41815
-29561
-27786
REVR6
=
-33632
-24715
-21961
-14486
REVENU
;=
161882
135865
81811
WTRI
WIN4
:z
>-
iYZ&L
-&
misMIL
CIIIWRI
>=
NRXNRI
<=
>=
BINRIZ
YLXRIZ
HIIYBLE
tlfw8LE

---

APPENDIX TABLE 2 - Continued
. . JzEIlmA.
TERRE1
<=
<=
TERRE2
<= 0.6
SEMRRR
<=
-1
SENCOT
<=
ti
-1
NPK 1
<=
-1
NPK2
<=
0
-1
NPK3
<=
-1
NPK4
<=
0
-1
NPK5
<=
0
-1
UREE
<=
0
-1
MCHERf3
-1
RCHERB
>z
:
-1.
. .
HEC1
<=
-1
HEC2
<=
0
INC
<=
FDWtRR
<=
ti
MD1
<= 104
MD2
<=
98
<.o
MDRl
122.6
4
MDR2
:s
9%
Tftl
SCRP
<= 2!xz
CRpl
<=
<=
0
110
110
90.9
81-9
8750
cfe2
<=
0
90.9
93.9
115-Y
88.9
9750
21600
RCRP
<=
QWIL
<=
0
QYWAIS
<=
0
QYRI>LU
ZZZ
Q-
É=
:
QYARRG
<=
BYARRF
<=
0
QYCOT
0
DEVAl
>=
>=
0
AUTD
5165
DEVRl
>=
0
i
OEVR6
>=
REVENU
>=
*SO4
t
NUTRI
>=
6489
MIN4
>=
O-05
j
MRH4
<=
0.75
i
MINMIL
720
MRXMI
L
$Z
1440
MCNMAI
>=
Mfi%MR 1
<=
z
MINRIZ
>=
360
MRXRI
Z
<=
840
MtNE3LE
>=
MIIXBLE
120
/

---

RHS
!3fG!iE
RISKA
RLSKR
CFE'Tl
.
.IixIuaL.
-0
-
R!%I!s&a
Y&=
""
TERRE1
,
<=
=
3-9
TERRE2
*=
0.6
SE-R
<=
SEHCXIT
<=
D
Hf=Kl
<=
HPK2
<=
0
NF-K3
<=
NPK4
=
&
0
FE
<=
a
rlIk#ERe
<=
RINERB
0
HEC1
<=
<=
0
HEC2
<=
<=
0
-1
-1
:E&fl
0
-1
Hi31
<=
<= 104
-1
5,
HQ2
sa
-1
:.o
PlORl
<=
<= 122.6
-1
-L)
Hm!2
-1
TRI
<=
<=
90
SCFlP
<=
-:
Cfwl
=
0
loooo
7350
-1
CAP2
&
0
1OOfB
500
500
500
<=
0
1
E%L
<=
QVMRIS
<=
:
OVRPLU
0
QVRNF?P
:z
QVRRRG
<=
0
OvflRRF
i=
0
CIVCOT
0
DEWl
>=
0
RUT0
>-
>; 5165
-0.3;
OEVRl
ZZ
IEVR6
;=
:
:
REVENU
>= es000
-1
HUTRI
>- 6489
HIN4
<z 0.05
HRx4
>= O-75
HINMIL
=
HfWMI
L
rlINfm1
;=
<=
Fz27:
>=
HFixRIZ
<=
>=
HINW-E
<=
0
H%E#-E
120
1

---

APPENDIX TABLE 2 - Continued
RETURN
TERRE
1
TERREZ
29
SEHRRR
0
SEIYICOT
0
NPKl
NPK2
0
IYPK3
NPK4
0
NfK5
UREE
D
rlIHER8
0
RIHERB
HEC1
0
HEC2
0
INC
FONRRR
0”
104
Ei
12298
“R:
TFIl
M
25000
1
t
ZT
0
284
_.
63
68
169
CFiP2
0
-1
1 .
RCRP
-63
-68
-85
-85
-90
-13
WMIL
:
1
GwlFI1s
0
1
WRPLU
1
WRNFV
:
1
QvRRR8
0
1
QVRRAF
1
QVCOT
iii
DEVRl
fUlTO
516:
DEVRl
DEVR6
:
REVENU
85000
NUTRI
6489
3.15
3.17
2.42
3
MIN4
0.05
wx4
O-75
HINMIL
1
tlRXMIL
1440
1
PlIi4MRI
1
WXMRS
720
1
FIINRIZ
360
1
WXRIZ
840
1
WINBLE
0
rlRX8LE
120
1

---

APPENDIX
TABLE
2
UPPER
Cft5HWMCE
F-ETLCE-Y.
TERRE1
<=
0::
TERRE2
<=
2;
D
5:m
NPKl
<=
0
100
100
NPKZ
<=
200
100
NFK3
<=
D
IOC1
50
NPK4
<=
NPKS
<=
0
UPEE
<=
:
50
50
M[HERB
<=
1
RlliERe
HtC 1
i:
0
HEC2
<=
0
X#c
<=
FDNRRA
<=
D
MO1
<=
104
11.5
3
20
23.5
71
12-5
MB2
<=
98
14
1:
9
25
28
9-s
$4.5
15
MBR 1
<=
<=
122.6
a
6
5
6
4
:4
Til
<=
98
6.5
5.5
1
4
SCRP
<=
5
4
1.S
1
25s
CnPl
<=
CRP2
<=
:
Rf RP
<=
QYNIL
<=
GlYw?15
<=
QYRPLU
<=
QYRNCIP
=
QYflRRG
<(=
QYRRRF
<=
QYCOT
=
REVfll
;=
- 1706
-1389
-1010
-1389
-758
-3581
-2830
-2384
-1872
-1559
-1180
AtJTO
>=
3333
2319
1707
2319
1223
4385
3634
3188
1872
1559
1180
DEVRl
>=
DEVR6
>=
REVENU
>=
NUTR f
>=
2833
1971
1451
1971
1040
3727
3089
W1C.J
1591
1325
1003
MCN4
3s
MRX4
<=
1
:
MCNMIL
>=
899
626
461
626
330
MRXtlIL
<=
899
626
461
626
330
MCNHflX
=
480
1176
975
Mnmx
:=
720
1176
975
855
MLNRIZ
>=
658
548
414
MIXRIZ
ZZZ
:z
658
548
414
’
S=
0
<=
120

---

APPENDIX TABLE 2 - ContSnued
VCOT .95
TERRE
1
<=
<=
TERRE2
<=
ZE?
<=
<=
FEE
<=
<=
WK3
<=
NPK4
<=
!FE
<=
MXHER&
<=
RIHERB
<=
HEC1
<=
tfEc2
<=
<=
:ONflRA
<=
Es
<=
<=
0”
WR1
<=
<=
‘.O
nuR2
-
.
TRI
<=
<=
<=
SE
<=
CRP2
<=
=
-9s
MIL
&
OUMRIS
<=
QVRPLU
<=
<=
iLIzE
<=
=
QVCOT
1
LW3JRl
S=
AUTO
>=
OEuRl
>=
oEuR6
>=
=
REVENU
;=
WTRI
>=
MIN4
<=
mx4
0.75
HXNMIL
>-
<;
720
MWMIL
1440
XINMRI
>=
4Bo
MRXMRI
<=
PIINRIZ
>=
E
FlFWRIZ
<=
840
PllNBLE
=
CQIHBLE
:=
4

---

APPENBIX TABLE 3
/*
,i--,---
------
USER INTERFACE PROGRAM ----------
*f
/*
KNOWLEEDGE
BASE CONSTRUCTION FOR DAT PREDECATE
*/
/*
---c--------------
x/
domains
Var, class,
unit
= symbol
/*Var = name activity(constraint>
*/
type
= symbol
/*type
= A for real activities
*/
val, coeff,
nret
= real
/*
C fur constraints
*/
min, max
= real
/t coeff = ObJ. coefficients
or RHS t/
choice
= integer
file
= data' ile
fname =
1
sym a1
database
ddat(var,type,val,coeff,nret,unit,min,max1
predicates
dat(var,typ+,val,caeff,nret,unit,min,max1
process(int#er>
mybase
nu
saveit<fnam+>
repeat
ciearram
da-file<fnab)
pal
mybase,
clauses
rcpeat
.
rapeat :- rgpeat.
datO'PMIC1" ; " A",1.48,-3600,O,"Ha",-11884,159642>.
.
JX
Module to,clear
the ram
*/
clearram;
:-
retract
(d<pat (,, ,, ,, ,, ,, ,, ,, -)) ,
fail,
clear-ram;:-
!,

---

. .
,)
102
APPEKDIX TABLE 3 - Continued
/*
this
is the main madule
*c/
mybaae
makewindb;(Z,7,7,<'CBaITRAL
PEABUT BASIK ",O,Q,25,80),
I
do-fileCFna+),
/X check if file
exists,
else create
it W
clea.r-ram,
consult(Fna$e>,
menu
saveitCFna*~>.
/f
wrlte
ta file*/
/*
Gheck fiie
*/
dofile(Fna+1
:-
enistfile~F+xame>,
!.
assertz<dda't<Var,Type,Val,Coeff,1Pret,Unit,Min,~x)),
aaveitCFna*),
t
I *
menu
:-
repeat,
clearwind#mj ,
Write ("
**~*********t******************~~~,nl,
write (*'
1:. CREATWEDIT KROWLEDGE
BASE"Z,nl,
write C** 2;. LIST
ALL
DATABASE
"),nl,
wri te ("
ti
QUIT
THE SYSTBM
"),nl,
Write ("
~~*;k*****b**XX**********~**~**~~),nl,
write["BKT$R
YOUR CHOICE
rr>,
readint(Ch#ce>,
nl,
process<Chdice>,
Chaice
= S,
!

---

103
APPIWDIX TABLE 3 - Cantinued
/*
module 1 :: EDIT
*/
process<l>
makewindow(jf;,7,j,CREATE
EDIT HODULE",2,20,18,58>,
shiftwind~c,~(2),
write<j,Bntek
variable
name:
,’ 1’
readlnNar>!,
writePBnt+
type
:
“ ) ,
adln(Typej>,
write(jjEnter
activity
level:
” > 1
readrealWa$),
fficient
: ” ) t
‘1
net return
:
1‘
readrealCB&t),
,’
write(j,Bnte~
unit
:
)t
adln<Uniti),
writeCj'Ent+
minimum
: ” ) ,
rsadrealGli~>,
“
writeC"Ent&
maximum
:
)’
readreal&c),nl,
aesertz~dda~~Var,Type,Va1,Coeff,Bret,Unit,Min,~x~~,
writeIVar,
"bas bsen added ta t,he data base"),
f
&aovswindo&
/*
Xist a11 pf-edicates
module
W
praczess(2)
if
makewindow(4i7,7,"
LIST ~DULE",7,20,8,50),
shlftwindow@,
ddat<Var,Typ~,Val,Cceff,I!Jret,Unit,Min,Max~,
write(Varj,t
*j,Type,j'
“,Val,"
jj,Caeff,'j
j,,ZYret,,,
j,,Unit,j,
'j,Hin,jj
jj,Nax>, nl
write("press;
space bar to continue
"),nl,
readchar(-),'
fail.
process(2)
ik removewindow,
/*
quit menu
t/
1
process(3)
:-
wri te Pi DO YbU WAlfT TO QUIT ? "1,
readlnCAnswe&?,
frontchar~An~wer,'Y',,~,
! +
saveitCFname1
:-
save(Fname>,
t
. .

---

. . . .
104
APPERDIX TABLE 4
/* --e*------------
LIREAR PR6GRAMIaTG IRTERPRETER -------
*/
/*
------------------
t/
domains
Var, unit,
type
= symbol
/*Var = nams activity(constraint>
*/
coeff,
nret
= real
/*
C for constraints
*/
min, max
= real
srrat
= integer
chaice
= lnteger
pcent
= real
database
ddat~var,type,val,coeff,nret,unit,min,max~
predicates
luain
lave11
leve12
shadawl
shadow2
process<fntlger1
El@zLU
ropsat
clear-ram
rrirtsc3urcz
gaal main.
clauses
rapeat
:- rppeat,
Y’*
lodu/le to clear the ram
w
clearramj
:-
retract
Wat (-, ,, ,) ,* -, ,, -, -) 1,
fail,
clearrami
:- !.
/*
this
is thl main module
t/
main
:-
mtli.kewindawCJ,7,7,"LIàTEAR PROGRAPIMIBG ES",0,0,25,80>,
csnsult("pbbsin,
dbf">,
mi(gIlU
ciear-karn
.
/*
end of mains module
w

---

105
APPERDIX TABLE 4 - Cantinued
nIenu
:-
clearwindow
,
WP-ite P
*~******S~*************~*****~*jj~,nl,
Write 0'
lj.
PRIRAL SOLUTION ABTALYSIS jj),nl,
write(jj
2,
DUAL VALUES ARALYSIS
jj),nl,
vmite(jj
3.
RBSOURCE RA?#GE ARALYSIS
"1 *Ill,
write (jj
4,
OBJECTIVE VALUES AIBALYSIS "),nl,
write("
5,
EXIT TO PROLOG SYSTEM
"),nl,
wr ite C"
~**~*******L***X************~*jj),n1,
write(jjEITE/R
YOUR CHOICE
jj),
readint(Chtifee>,
nl,
process(Ghdice1,
Choice
= Fi,
f
/i- * ########a#
IRFEREBCE
E 3 G 1 3 E ######X###X##
W
/x ---a---
Prlnyl
solution
module ---M-M
x/
process<l)
:-
makewindow(~,'7,7,jjPRIXAL
SOLUTION jj,2,10,20,60>,
shiftwindm@2)
,
levell
j
/5
check for optimal
real activities
*/
leve12 ,
/t
check for non optimal
real activities
t/
remvewindoy+
,
sWftwindo~(l),
t
* II
/*
--I------
D*l
solution
module ----------
t,'
pracess<2>
:-
makewindow(@j7,7jjjDUAL
SOLUTION jj,2,10,20,60~,
shlftwïndo$3)
,
shadowl j
/*
check for binding contraints
*/
shadow2 j
/t
check for n5n
binding
constralnts
*/
raavewindo$v >
shlftwindow~<l),
fail.
15 --.------
Resource range module ----------
b/
process(3)
:-
makewfndow<iA,7,7+jjRESOURGE RARGESjj,2,10,20,60~,
sbiftwlndbw$4)
,
rt%siOurce
readchart-;j
remvewindow,
!,

---

106
APPENDIX TABLE 4 - Continu&
/%
--------
ObJectlve
function
module ----------
%ir/
process(43
:-
makewindow@i,7,7,"OBJBCTIVE
VALUES",2,10,20,60>,
sbiftwindo~@)
/t TO BB IRPLB HT& LATTER t/
wrfte<,,CUI&I ? G UP . ...'>,
readcharC_1,
re
vewindow,
t
. ,
pracess(S>
:-
wri te (" Ws YtKJ WART TO QUIT ? "1,
readlnCAnswer1,
frantchar<An~wer,,y',_)l
! b
,'# ########Y# P R 0 D U C T 1 0 R
R U L E S #Y######## t/
/Y --'----
Check for optimal
real actîvity
-------%y
levell
'- '
kewi;dow;(a1,7,7,"OPTIMAL
REAL ACTIVITIES",4,0,18,8O~,
shiftwindo
(21)
writec,,
31111
L ,,,,\\Q:,
"LBVEL,,, , \\Q',f,RETURB,,, '19, ,,,UI?IT', 1, nl,
Write P --f-------------------------------
,,>,nl,
fail
,
levell
:-
:
ddat <Var, Tvpe,Val,Coeff,-,Unit,-,-),
Val > 0
'
Type= " A”
write<Var,,
i91,
Val,,\\9,,Coeff,,\\9,,
Unit>,nl,
fail.
levell
:-
cursor(lO,D>
write(,jThese
Are the real activities
in the enterprise
mix,,, Lnl,
ouraorCll,b),
write(,,Levbls
indicate
how many units
of each activity
,,>,nl,
cursor(l2,p)'
out in this
plan.,,),nl,
space bar . . .,,),

---

107
APPERDIX TABLE 4 - Continued
Jf
--.---
Check for cost of non optimal
real activities
-------t/
leve12
:-
~kewindow~22,7,7,'>ROB
OPTIMAL REAL ACTIVITIES",6,0,16,8O~,
shiftwindow(22)
Write (" ~A~','~Q~,,'RETWRB",'\\Q,,,fCOST",,~Q',"UBIT,'~,
nl,
wite< >',,,i-,,,,-------------------------
>'),nl,
fail
t
level"
*-
ddat(V&,Type,Vnl,Coeff,Bret,Unit,_,_),
Val=0
' ,
Type= "A"
write(Var,
f'\\Q',Coeff,,\\Q',Iet,'\\Q,,
Unit),nl,
fail.
level2
:-
cwsor t10, g, )
write(,+These
aetivities
are not in the enterprise
mix.,'>,nl,
cursorcll,cj1,
write<',Costs
represent
the cost of forcing
one unit
of activity
,,),nl,
csiraor~l2,~~,
wrfteC',in
t/he optimal
plan,,'),nl,
write<',pre46
space bar , ,.,O,
raadcharc-$,
removewindow,
! t
/*
--a:---
Chech for ,bindlng constraints
-------*/
shadowl : -
2
~kewindow¢31,7,7,*,SCARCITY
VALUES ,,, 4,0,16,00>,
aiftwindowC31)
,
write("I!J~t,'\\Q,,',RHS,,,'\\Q,,',SHADOW
P,",'\\Q',',UI?IT,f>,
nl,
mite P --- i------------------r'-------------
,'),nl,
fail
.
shadowl
:-
ddat(Var,Type,Val,Coeff,Bret,Unit,-,-1,
Val = 0
f
Type= "C" ‘,
write~Var,*'\\Qt,Coeff,,\\Q,,Bret,~\\Q\\Q,,,
Unit>,nl,
fail.
stidowl
:-
cursor(lO,Cj>
,'
wr-ite(,,These
represent
the constraints
binding
to this
enterprise
,'),nl,
cursor~ll,
a,,
write ("mix.
Shadow prices
measure the contribution
of an additional,,>,nl,
cussor(l2,cj>,
write<i'unit
of resource
to net farm return.,'>,nl,
write<',pre$s
space bar ,..when
done">,
raradchar( - >',
!

---

108
APPERDIX TABLE 4 - Continued
/*
------
Check for slack resources
-------*/
shadow2 :-
x&akewindow{32,7,7,,,SLACK
RESOURCES,,,6,0,16,80>,
shiftwîndojv~32~
,
writeYalAaEZ$",,\\9,,"RHS',,'\\9,,,,SLACK
,,,,\\9',,'UIYIT',>,
nl,
writeP
:
----------------------------------
"),nl,
fail
.
shadow2
:-
ddat(Var,Tpps,Val,Coeff,-,Unit,-,-),
Val>0
'
,
Type=
” C”
writeWar,
ti9f
,Coeff,,\\Q',Val,,\\9,,
Unit),nl,
fail.
&adow2.:-
cur8or<lo,
$0 (
writePThe$e
constraints
are in the optimal
enterprise
mix,,'I,nl,
cur8or~l1,
b,
mite ("Thelr
levels
represent
unused capacity
or idle far&'>,nl,
CW4tmC12,@),
wsite~"rsth?i'~~es~'~~,nl,
writeC,'pre b space bar . ..when done">,
raadchar(-?,
rsxwvewltidow,
!.
/t
Resaurce ra+ge analyses
X/
resourcs
: .-
:
l

ddatWar,Type,Val,-,Hret,Unit,Min,Kax>
,
Type="C"
,
Val = Q ',
Pcent =c Ma* - Min)/ Win * 100
write~,tCoisfraint
Rame :',,,\\9',Var>
tnl.4
wrlite<"Its
+plicit
price
is
; 't ,t\\9f,Hret,'\\9,,,,CFA,'~
,nL
writeC,tRang+
of optimality
:
,,,,\\9t,Wln,
,\\9,,Max,',
(',,Unit,'t>,t>,nl,
write (PCent,/',X,, , ,, variation
in
"'Var,"
avai lable,'
>,nl,
write("Press
space bar ..+ when done,'),
readcharc-J,inl,nl,
,
fail.
resource
:-
+
I .

---

109
Abt Associates. "Senegalese Agricultural
Policy Analysis."
:Report prepared for the U.S. Agency for International
:Development, Agricultural
Development Office, Dakar,
Senegal by Abt Associates Inc., AAI No. 85-32, 1985.
Bonkia(n, Adama. "EconomIc Analysis of Resource Allocation in
j‘rraditional Agriculture:
Case of the Boromo farm in
Durkina Faso." Masters Thesis, Michigan State
!University, 1985.
Borla
.
Cattinp Michel Benoit and Jacques Faye. axploitattoq
.
iwfcol e Fmiliale en Afrtoue Soudan0 Sahel
a
enne Paris:
Presse Universitaires
de France, ACCT, 1982:
l

Crawfobd, Eric W. and Valery Kelly. "Enquête sur la Distribution
ut 1' Utilisation
de 1'Engrais au Ser@ai: Resumé
Analytique." .Institut Sénégalais de la Recherche Agricole,
Bureau d' Analyse Macroéconomique, Dakar, 1984.
Eicher Carl K. and John M. Staatr. "Food Security Policy in
i5
ub-Saharan Africa."
Invited Paper for the XIXth
lonference of the International Association of
Agricultural
Economists, Malaga, Spain, August 25-
beptember 5, 1984.
Gotsch, Carl H. and Walter P. Falcon. "The Green Revolution and
the Economies of Punjab Agriculture."
food Research
@titute
14 (1) (1975): 3-10s.

---

. .
.
110
HazelT, Peter B.R. and Roger D.Norton. Mathematical Proqramming
.
.
for Econom c Analvsis in Aar iculture
New York: Macmillan
l

Publishing'C
any, 1986.
Happer, W. D.. "Allocation Efficiency in Traditional Indian
,Agriculture." Journal of Farm Economies 47 (3) (1965).
Kennedy, J. 0. S. and E. M. Francisco. "The Formulation of Risk
'Con!traints For Linear Prograa&ng." Journal of
Aarlcultural
Eco~9mic~ 25 (2) (May 1974): 129-144.
ton, A.R.C.. "Decision Taking Under Uncertainty: A Linear
:Progranrning Mode1 of Peasant Farmers Behavior." aournal
pf Aoricultural
Ecom
25 (3) (May 1974): 311-22.
Martin, Frederic. "Analyse des Marges dans le Centre du Bassin
firachidier."
Instjtut S&@alais de la Recherche
Pi?le,
Bureau d Analyse Macrodconomique, Dakar, 1987.
,ra
.
Martin, Frederic. "Budget de Culture au Sénégal .” Institut
SBndgalais de la Recherche Agricole, Bureau d' Analyse
Macro&onomique, Dakar, 1987. Draft.
Martin, Frederic. "Food Security and Comparative Advantage in
Senegal." Michigan State University, Department of
Agricultural
Economies, East Lansing, Michigan, 1986.
Draft.
Martin, Frederic. "Notes on The Mode1 in Zone 1.” Institut
Benégalais de la Recherche Agricole, Bureau d' Analyse d
acro&onomique, Dakar, 1986,
Martin' Frederic and Eric W. Crawford. "Questions A Propos de
\\ 'Autosuffisance Céréalière au Sénegal." Report prepared
for IFPRI conference, Dakar, Senegal, July 15-17, 1987.
Martin. Frederic and Mamadou
Sidibe. "Notes on The Mode1 in
t one 7." Michigan State University, Department of
Agricultural
Economies, East Lansing, Michigan, 1987.
Draft.

---

111
Minist&re du Developpement Rural.
Nouvelle Politiaue
ficrricole. Dakar, S&&gal, 1984.
Ndoye, Ousseynou. "La FiliBre CBréaliére au Sénégal: Cas du
Basin Arachidier."
Institut Sénégalais de la Recherche
Agricole, Bureau d' Analyse Macroéconomique, Dakar, 1985.
Newman, Mark, Papa A. Sou and Ousseynou Ndoye. "Regulatory
Lhtcertainty, Government Objectives and Grain Market
Qrganization and Performance: The Senegalese Case."
Institut
Senegalais de la Recherche Agricole, Bureau
cl' Analyse Macroeconomique, Dakar, 1985.
Niane,'Amadou D.. "Supply and Demand of Millet and Sorghum in
Senegal." Afrf ca-Rural Eçonomv Proqw Workina Paper No,
East Lansing, Michigan: Repartment of Agricultural
Michigan State University,
1980.
Robins'n, Lindon J. and Beverly Fleisher. "Decision Analysis in
A
,griculturaJ
Settings: An Introduction.
" Aaricultural
Michigan State University, Report No.
$84.
Schultink, 6.. "The CRIES Resource Information System:
tomputer Aided Land Resource Evaluation for Devefoptnent
Planning and Policy Analysis." Norwich: 1987.
.
Schulti, T.W.. Jransformim Traditiona 1 Aqriculture.
New Haven:
Yale University Press, 1964.
Shlomo' R.. "Policy Options For Food Security," World Bank,
agriculture and Rural Development Department, December,
1984.
Soned-.EMA. "Modélisation des Prix Agricoles: Rapport Final.”
a ,eport prepared for the Department of Rural Development
Dakar, 1981.
Subrata,
*
G. and Ken Ingersent. Aariculture gnb Econom q
ooment. Baltimore: The John Hopkins Univeriity
Press,
.

---

112
WARDA.
Pice StatistScs Yea book West Africa Rice Development
:Association, Liberfa: Juné 1981.
YacSuh, Gordon. "Post Havest Technology in Senegal: Current
'Practices and Future Needs." International
Development
ilesearch Center, IDRC-l4R 13, Ottawa, 1980.
Yin, tihin H. and David Salomon, bina Turbo Prolog.
tIndianapo1
is: Que Corporation, 1987.
Youtohulos, P.A. and J. B. Nugent, rç9nomics of Develooment.
kew York: Harper & Row, 1976.

---