Front Cover
 Title Page
 Table of Contents
 Partial budget analysis of...
 Capital scarcity and the cost of...
 The use of net benefit curves and...
 Variability in net benefits and...
 More on estimating costs
 More on estimating benefits
 Summary of procedures for deriving...
 Two examples
 Back Cover

Title: From agronomic data to farmer recommendations
Full Citation
Permanent Link: http://ufdc.ufl.edu/UF00102139/00001
 Material Information
Title: From agronomic data to farmer recommendations
Physical Description: Book
Language: English
Creator: Perrin, Richard K
Publisher: Centro Internacional de Mejoramiento de Mai´z y Trigo
Place of Publication: Me´xico
Copyright Date: 1976
 Record Information
Bibliographic ID: UF00102139
Volume ID: VID00001
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: oclc - 2608764
issn - 0304-551X ;

Table of Contents
    Front Cover
        Page i
        Page ii
    Title Page
        Page iii
        Page iv
    Table of Contents
        Page v
        Page vi
        Page 1
        Page 2
        Page 3
        Page 4
        Page 5
    Partial budget analysis of experiments
        Page 6
        Page 7
        Page 8
        Page 9
        Page 10
        Page 11
    Capital scarcity and the cost of investment capital
        Page 12
        Page 13
        Page 14
    The use of net benefit curves and marginal analysis to derive recommendations
        Page 15
        Page 16
        Page 17
        Page 18
        Page 19
    Variability in net benefits and implications for recommendations
        Page 20
        Page 21
        Page 22
        Page 23
        Page 24
        Page 25
        Page 26
    More on estimating costs
        Page 27
        Page 28
        Page 29
        Page 30
        Page 31
        Page 32
        Page 33
        Page 34
    More on estimating benefits
        Page 35
        Page 36
        Page 37
        Page 38
        Page 39
    Summary of procedures for deriving recommendations from experimental data
        Page 40
        Page 41
    Two examples
        Page 42
        Page 43
        Page 44
        Page 45
        Page 46
        Page 47
        Page 48
        Page 49
        Page 50
        Page 51
        Page 52
    Back Cover
        Page 53
        Page 54
Full Text


An Economics Training Manual

Richard K. Perrin
Donald L. Winkelmann
Edgardo R. Moscardi
Jock R. Anderson

Information Bulletin 27

International Maize and Wheat Improvement Center, Apartado Postal 6-641, M~xico 6, D.F. Mexico

Correct citation: Perrin, R.K., D.L. Winkelmann, E.R. Moscardi, and J.R.
Anderson. 1976. From agronomic data to farmer recommendations: An
economics training manual. Centro Internacional de Mejoramiento de Maiz y
Trigo, Mexico City, iv + 51 p.

Spanish edition: Single copies of the Spanish edition of this manual are
available on request.

Reprint rights: Persons wishing to reproduce this manual by any means may do so
without prior approval of CIMMYT. Credit to the authors and CIMM YT would
be appreciated.

CIMMYT: The International Maize and Wheat improvement Center (CIMMYT)
receives financial support from government agencies of Belgium, Canada,
Denmark, Iran, Netherlands, Saudi Arabia, United Kingdom, USA, West Germany
and Zaire; and from Ford Foundation, Inter-American Development Bank,
International Minerals and Chemical Corp., Rockefeller Foundation, United
Nations Development Programme, United Nations Environmental Programme,
and the World Bank. Responsibility for this publication rests solely with CIMMYT.

So dio t~rmino a la impresl6n de este Ilbro el 16 de julio de 1976 en los talleres de
Ediciones Las Americas. Tiro: 3,000 ejemplares. Impreso en Mexico.


Successful farm recommendations 1
Representative experimental conditions 1
Goals of the farmer 2
Relationship between statistical and economic analysis of experiments 4
A ims of the manual 5

Basic concepts 6
Partial budget analysis of fertilizer experiments-an example 9


The net benefit curve 15
Marginal analysis of net benefits 17
Sources of vie/d variability 20
Adjustment of recommendations for yield variability 23
Price variability and sensitivity analysis 25

Identifying and measuring variable inputs 27
Determining the field prices of purchased inputs 29
Determining the field price of equipment 29
Determining the field price of labor 30
Determining the cost of investment capital 32
Summary 33

identifying and assessing benefits 35
Tenancy 37
Summary 38


Maize technology packages 42
Wheat variety trials 45




This manual was prepared by the economics section of CIMMYT for use in its
maize and wheat training programs. We hope that other agronomists will find it
useful. We authorize and encourage the reproduction of any part of the manual.
Comments from users which might improve the manual are solicited.

The idea of a manual was first presented to the CIMMYT Internal Review in
1972 by the economics section. A first version, written by J.R. Anderson,
emerged from discussion between Anderson and Don Winkelmann. This version
was substantially rewritten and expanded by Richard Perrin and Winkelmann. The
second version was reviewed by Edgardo Moscardi while testing it with trainees.
Moscardi and Perrin altered the draft and Winkelmann reviewed it. This version,
the third, was sent out to agronomists and economists for comment. We're espe-
cially grateful to John Dillon, John Lindt, Torrey Lyons, Paul Marko, Matt
McMahon, Robert Osler, Willem Stoop, Alejandro Violic, Pat Wall, and Delane
Welch for helpful suggestions. Moscardi and Perrin incorporated many of these in
this, the fourth version, which Winkelmann again reviewed.

Richard K. Perrin
Donald L. Winkelmann
Edgardo R. Moscardi
Jock R. Anderson


This manual is intended for use by agronomists as they make farm recommenda-
tions from agronomic data. It is not necessarily difficult to make recommenda-
tions which fit farmers' goals and situations, but is certainly easy to make poor
recommendations by ignoring factors which are important to the farmer; some of
these factors may not be very evident.
The philosophy of this manual is that it is better to estimate an effect of a
factor than to ignore it completely, even though it is sometimes difficult to esti-
mate the effect of some factors on farmer choices. This manual provides lists of
these factors and procedures for dealing with them from the farmer's point of

Successful farm recommendations
A good farm recommendation could be defined as a choice which the farmer him-
self would make if he had all the agronomic information available to the agrono-
mist. Such a recommendation will be successful because farmers will adopt it and
continue using it.
For successful farm recommendations, the agronomic data upon which your
recommendations are based must fit the farmer's agronomic conditions. If not,
the farmer will not obtain the results you predict. Also, your evaluation of these
data must be consistent with the farmer's goals and with the factors that in-
fluence his ability to attain those goals. Let's look more carefully at these two
aspects of farmers' circumstances.

Representative experimental conditions
It is impossible to conduct experiments on each farm to make recommendations
tailored to each farm. Instead, you must define a target group of farmers, con-
duct experiments under conditions representative of their farms, and make rec-
ommendations which are applicable to the entire group. We shall call such a
group of farmers a recommendation domain. Generally, a recommendation do-
main will consist of farmers within an agro-climatic zone whose farms are similar
and who use similar practices.
While there are no clear rules for delineating recommendation domains and
agro-climatic zones, you must have these concepts in mind to make successful
farm recommendations. In practice the best rule is to seek out a group of farmers

for whom you can expect a similar choice of variety, fertilizer level, etc. If the
best level of fertilizer for all farmers in a large geographical area is 60 to 80 kg/ha
of N, and if the best variety for virtually all farmers is variety Z, then for pur-
poses of this crop the entire area could be considered a recommendation domain,
even though there may be considerable variability in soils and climate across the
You need several representative experimental sites (not just the accessible ones,
the productive ones, or the flat ones) to provide a sample of the results that far-
mers can expect in a given domain. If you have data from only one year and one
site, they are better than none, but they are not very helpful even for making
recomme nations for the farm on which the experiment was conducted. To make
good recommendations, you need to learn the range of agronomic results obtain-
ed from farm to farm and from year to year in the recommendation domain.
The cultural practices you use in the experiment must be similar to those
which farmers can be expected to use, or the results from the trials may not rep-
resent the results which farmers will obtain when they try the recommendation.
For example, it is not wise to use weed control techniques that farmers can not
adopt or could not profitably adopt. You must take care that the plots are large
enough to avoid border effects which would not occur in farmers' fields. Also, if
most farmers in an area are dependent on rain for water supply, the results ob-
tained from a well-irrigated fertilizer trial may have little relevance to the results
which these farmers can expect to obtain.
These and other problems in the planning of a useful set of agronomic experi-
ments are beyond the scope of this manual. We introduce them here to stress the
point that the agronomic conditions under which the trials were conducted must
be representative of farmers' agronomic conditions if recommendations based on
the trials are to be good ones. But it is not enough that the agronomic data be
representative of the farmer's agronomic conditions. The procedures used to de-
rive recommendations from these data must be consistent with the goals of the
farmer (who will decide whether to accept the recommendation).

Goals of the farmer
To make recommendations that farmers will use, you must be aware of the hu-
man element in farming, as well as the biological element. You must think in
terms of farmers' goals and the constraints on attaining those goals.
In this manual, it is assumed that farmers think in terms of net benefits as they
make farm decisions. For example a weed-conscious farmer will recognize that
he will likely benefit from eliminating weeds from his fields by harvesting more
grain. On the other hand, he recognizes that he must give up some cash to buy
herbicides and then give up some time and effort to apply them, or he must give
up a lot of time and effort for hand weeding. The farmer will weigh the benefits
gained in the form of grain (or other useful products) against the things lost
(costs) in the form of labor and cash given up. The net result of this weighing in
the farmer's mind we refer to as the net benefit from a decision -the value of the
benefits gained minus the value of the things given up.
Two factors complicate our understanding of this decision process. The first is
that we cannot subtract hours of labor from kilos of grain to obtain a useful
estimate of the net benefit which a farmer would perceive. The farmer can per-

From agronomic
data to farmer
recommend tions

haps make such a judgement, but we must find a more systematic method of
evaluating net benefits if we are to avoid the problem of adding and subtracting
hours of labor, kilos of fertilizer, kilos of grain and tons of fodder. The second
factor which complicates our understanding of the decision process is that the
farmer is uncertain of the results which he will obtain from any given decision. In
our weed control example, the farmer knows that in the case of severe drought or
early frost, he may get no grain regardless of the amount of weeds in his fields. If
this happens, there is no benefit at all from killing weeds. Unfortunately, it is
difficult to know just how the farmer sees these risks, and how their existence
affects his decision, but we know that they do affect the decision. In general,
farmers attempt to protect themselves against risks of loss in benefits, and often
avoid choices which subject them to these risks, even though these choices will on
the average yield them positive net benefits.
In order to avoid the problem of subtracting hours of labor from kilos of grain,
we estimate the value to the farmer of a kilo of grain and an hour of labor in
terms of the common denominator, money. This gives us an estimate of net bene-
fit measured also in terms of money. This does not necessarily imply that the
farmer spends money for the labor, nor that he receives money for the grain.
Neither does it imply that we think that farmers are concerned only with money.
It is simply a device which we use to represent the process which we know goes
on in the farmer's mind, the process of weighing the things gained and the things
If our weed-conscious farmer is quite commercialized, that is, if he is contem-
plating hiring the labor, purchasing the herbicide, and selling the extra grain, then
we can attach anticipated market prices to labor, herbicides and grain, and in this
way represent quite accurately the net benefits which the farmer foresees. On the
other hand, if he is a subsistence farmer we have to employ the concept of oppor-
tunity cost to represent the values he places on labor and grain, since there would
be no money prices given up or received. Opportunity cost is the value of any
resource in its best alternative use. Let's consider the opportunity cost of the
farmer's time. If he has a job off the farm which he has to give up temporarily to
weed his field, then we say that the opportunity cost of his time in weeding the
field is the wage which he would have been earning if he had stayed in his job
Suppose, however, that the best alternative use of his time is working on his
tobacco, and that the day's work on tobacco will increase the value of the tobac-
co harvest by $5. (The $ symbol in this manual does not represent any particular
national currency. Also, weights are in metric units). In this case, the opportunity
cost of his time in weeding maize is $5 per day, since that is what he gives up by
weeding the maize instead of tending the tobacco. Suppose the farmer would
merely sit in the shade if he were not to weed his maize? Is the opportunity cost
of his time zero? This is not very likely, since most people place some value on
being able to sit in the shade rather than to work in the sun. But it is difficult to
guess the value which a farmer places on leisure, if that is the highest-valued alter-
native use of his time.
We have suggested here the two main problems in evaluating agronomic alter-
natives from the point of view of net benefits to the farmer. The first is estimat-
ing the relative weights which farmers place on various kinds of goods, and we


introduced the concepts of market prices and opportunity costs as ways to deal
with this problem. The second problem is estimating the effect on farmers' deci-
sions of uncertainty about net benefits. Much of this manual gives procedures
which can be used to estimate prices, opportunity costs, and the effect of risk as
they are viewed by farmers.
One further point is in order with respect to farmers. The conditions under
which farmers live and work are diverse in almost every respect imaginable. They
have different amounts of land and, to an extent, different kinds of land even
within an agro-climatic zone;.they have different degrees of wealth, different atti-
tudes toward change, different attitudes toward risk, different marketing oppor-
tunities, and so on. Many of these differences influence the farmer's response to
recommendations. Unfortunately, it is impractical to attempt to make a separate
recommendation for each farmer. Instead, you must offer recommendations that
will be approximately correct for large groups of farmers in recommendation

The relationship between statistical and economic analysis of experiments
To this point we have not mentioned statistical analysis. Most agronomists are
familiar with the techniques available to determine whether or not the mean
yields from two treatments in an experiment are significantly different from one
another. Some persons say that if treatment means are not significantly different,
then there is no need for an economic analysis. This is not necessarily so, how-
ever. For one thing, most statistical tests are geared to the 0.05 or 0.01 levels of
significance. But farmers may be willing to accept evidence that is much less per-
suasive than this. For instance of variety A yields 3 tons in an experiment, while
variety B yields 4 tons, farmers may be quite happy to choose variety B even
though this difference is statistically significant at, say only the 0. 10 level.
Furthermore, it is quite possible that two treatment means are not significantly
different at any of five trial sites, but the treatment means are different at the
0.01 level of significance when the data are pooled. Because of these considera-
tions, we suggest that both statistical and economic analyses be conducted. If
only one experiment is available, little can be said of the desirability of the treat-
ment for farmers in the area, unless the results are overwhelming. When several
experiments are available (from different sites or years or both), a statistical anal-
ysis of the pooled data should be conducted. The analysis of variance should in-
clude treatments, sites, and site-by-treatment interaction as sources of variation.
If the treatment means are not significantly different, but an economic analysis
shows that one treatment is a better recommendation than others, then a more
careful analysis of the recommendation, using the procedures of Chapter 4 and 5
of this manual, is in order. In all other cases, the agronomist should be guided by
the economic analysis in making his recommendations, for if he has done it well,
his recommendation will indeed be in the best interest of the farmer.
This is not to say that statistical analyses are not useful. They are. However,
their greatest value is not in deriving recommendations, but in determining what
is happening, biologically, in the experiments. For example, only with statistical
analyses can the agronomist determine with confidence whether there is an inter-
action between nitrogen response and phosphate level, or whether the response to
nitrogen varies significantly from location to location. This type of information

From agronomic
data to farmer

may be very valuable in planning further trials, and to some extent in interpreting 1/Introduction
the results of the trials already conducted. But statistical analyses are not neces-
sary in deriving recommendations.

Aims of the manual
The goal of this is to show you how the elements described in the previous sec-
tions interact in the art of making recommendations. By use of the manual you
will be able to:
1. Identify the benefits associated with treatment alternatives, and place val-
ues on them which match farmers' goals.
2. Identify which inputs change from treatment to treatment and place values
on them which match farmers' goals.
3. Identify sources of variability which will make the farmer uncertain about
the net benefits which he will get from each treatment.
4. Derive recommendations from cost, benefit and variability data, which are
consistent with the farmer's desire to increase average income, with the farmer's
desire to avoid risks, and with the scarcity of investment capital which is typical
of mo st farm situations.
Our approach is deliberately non-mathematical and only a few concepts and
special terms from economics are used. This is because we believe that such
knowledge is not necessary for deriving successful farm recommendations.


We have stated that farmers are interested in net benefits and in protecting them-
selves against risk. We have also stated that if you want to make good recommen-
dations, you must keep these goals in mind and evaluate alternative technologies
from the farmer's point of view. Partial budgeting is a method of organizing ex-
perimental data and other information about the costs and benefits of various
treatments. In this chapter we introduce the partial budget concepts. In later
chapters we discuss in more detail some of the problems involved in estimating
costs and benefits. In Chapter 4 we describe procedures for deriving recommenda-
tions from partial budget and risk information.

Basic concepts
The purpose of partial budgeting is to organize information in such a way as to
help make a particular management decision. The types of decisions with which
agronomists will usually be concerned are the choice of fertilizer level, the choice
of variety, the choice of seeding data and rate, and so on, or perhaps the choice
among alternative packages of such practices. Some of these are "yes or no" deci-
sions and others are "how much" decisions, but all of them may be budgeted in
the manner to be described.
To introduce these concepts, let's consider once again the case of the weed-
conscious farmer. He has perhaps seen some experimental results across the fence,
and knows that for the last two seasons, the plots without herbicide yielded an
average of 2 tons per hectare and the herbicide plots averaged 2.5 tons. His own
yields averaged about 2 tons, also, and he thinks he would realize about the same
yield increase from herbicides on his own farm.
We don't know the exact sequence of steps the farmer would use to evaluate
this choice, but in some fashion he weighs the benefits he would receive from
each alternative with the costs which he must give up for each alternative. We can
simulate the same process, and record the results as we go in Table 1. We will first
look at benefits, then costs, and then net benefits.
The first concept used is:

Net yield-the measured yield per hectare in the field, minus harvest
losses and storage losses where appropriate.

Our farmer is satisfied that the yields obtained in the trials are the same as he 2/Partial budget
would obtain, and since he sells his grain immediately after harvest, he need not analysis of experiments
consider storage losses. We can therefore record 2.0 and 2.5 in line one of Table I
as a measure of the yields the farmer expects to receive. The next issue is the
value which the farmer places on the yield, which we designate as:

Field price (of output)-the value to the farmer of an additional unit of
production in the field, prior to harvest. Farmers who sell all or part of
their grain will be concerned with money field price while those who
consume the entire crop will be concerned with opportunity field price.
Money field price is the market price of the product minus harvest,
storage, transportation and marketing costs, and quality discounts.
Opportunity field price is the money price which the farm family
would have to pay to acquire an additional unit of the product for con-

Our farmer always sells his grain to a trucker who comes by, and he expects to
receive $1100 per ton. However, he also knows that it costs him about $100 per
ton to harvest and shell the crop, so that the field price is $1000 per ton. Multi-
plying net yield by field price, we obtain an estimate of the total value or:

Gross field benefit-net yield times field price for all products from the
crop. In general, this may include money benefits or opportunity bene-
fits, or both.

In considering the costs associated with this decision, the farmer need only be
concerned with those costs which are affected by the decision or variable costs.

Example of a per hectare partial budget
Present Use of
practice herbicides
farmer's yield (net yield) 2.0 tons 2.5 tons
farmer's value (field price) $1000 $1000
total. benefit (gross field benefit) $2000 $2500
Variable costs
amount 2 liters
value (money field price) -X$30
total (field cost of herbicide) -$60
labor for application:
amount -2 days
value (opportunity field price) -X$10
total (field cost of application labor) $20
labor for hand weeding:
amount 10 days 3 days
value (opportunity field price) X$10 X$10
total (field cost of weeding labor) $100 $30
total variable costs $100 $110

Net benefit $1900 $2390

Note that the $ symbol in this manual represents no particular
national currency. Weights are in metric units.

Costs which are not affected by the decision (such as plowing and planting costs
in this case) are known as fixed costs. Since these costs will be incurred regardless
of which decision is made, they cannot affect the choice and can be ignored for
the purpose of this decision. The term "partial budgeting" is a reminder that not
all production costs, and perhaps not all benefits are included in the budget-only
those which are affected by the decision being considered.
If the farmer is to make a good decision, he must identify all the inputs which
would change if he decides to apply the herbicide. In his case this includes only
the herbicide and the labor required to apply it, plus the reduction in hand weed-
ing labor (he already has a hand sprayer which can be used). The amount of herbi-
cide required is two liters per hectare, and based on the amount of time it takes
him to apply insecticide, he estimates that application will take two days of his
time per hectare. The value of the herbicide can be simply expressed in terms of
money, because it is money, $30 per liter, which he must give up to acquire it.
This value concept we refer to as:

Field price (of an input)-the total value which must be given up to
bring an extra unit of input into the field. Money field price refers to
money values such as purchase price or other direct expenses. Oppor-
tunity field price refers to the non-money value of inputs which must
be given up. The opportunity price is the value of the input in its best
alternative use. For farm family labor, the opportunity field price may
be the wage which could be earned in off-farm employment, or the
value of the time if spent on another farm enterprise, or the value
which the worker places on leisure.

Field cost (of an input)-is the field price of an input multiplied by the
quantity of that input which varies with the decision. It may be ex-
pressed as money field cost or opportunity field cost, or perhaps both,
depending on the input.

Thus for our farmer, the field cost of the herbicide is $60 per hectare. Regard-
ing his labor, the farmer might perhaps note to himself that he would not do that
kind of work for anyone else for less than $10 per day (otherwise he would rather
sit in the shade). This means that he evaluates the opportunity cost of his time at

Maize yields (tons/ha of 14 percent moisture grain) by fertilizer treatment, 8 trials.

Fertilizer treatment (kg/ha)
N: 0 50 100 150 0 50 100 150 0 50 100 150 *
Trial Pz0s: 0 0 0 0 25 25 25 25 50 50 50 50 Avg.
1 0.40 1.24 3.63 3.76 0.79 2.58 4.23 4.72 1.67 2.51 3.28 3.66 2.71
2 1.53 2.60 5.14 5.32 1.67 3.79 5.10 6.83 1.41 4.13 5.89 6.27 4.14
3 4.15 4.86 4.80 4.87 4.44 5.00 4.97 5.28 5.12 5.66 6.36 6.62 5.18
4 2.42 3.82 5.23 4.48 2.36 4.54 6.26 7.17 1.61 4.41 5.38 6.58 4.52
5 1.64 1.92 2.08 2.19 2.04 3.21 3.12 2.93 1.44 3.44 3.32 3.62 2.68
6 1.61 2.94 4.14 4.34 1.81 3.92 3.61 3.81 1.18 3.89 5.38 4.92 3.46
7 4.74 5.41 4.29 4.92 4.91 5.22 5.38 5.14 5.10 4.88 4.54 5.28 4.98
8 1.21 2.33 1.97 2.23 1.53 2.78 2.49 2.80 1.37 3.51 3.75 4.35 2.53
Avg. 2.21 3.14 3.91 4.01 2.44 3.88 4.40 4.84 2.36 4.05 4.74 5.16 3.76

From agronomic
data to former

$10 per day, and therefore, the field cost of the labor for the herbicide treatment 2/Partia/ budget
is $20 per hectare. He also observed that when herbicides were used, the time analysis of experiments
spent on hand weeding was reduced from 10 days per hectare to just 3. The cost
of hand weeding was thus reduced from $100 to $30. The total of these values for
each treatment is:

Total field cost or Variable cost-the sum of field costs for all inputs
which are affected by the choice. In partial budgeting we refer only to
those inputs which are affected by the decision, so that total field cost
in fact refers to variable costs, i.e. those costs which vary with the
choice. Variable cost can consist of either money costs or opportunity
costs or both.

The total variable cost of the herbicide alternative is $110 per hectare. The
total variable cost of the present practice is $100 per hectare. Subtracting these
from the benefits received gives:

Net benefits-total gross field benefit minus total variable costs.

In the net benefit figure we want to represent the value which the farmer places
on additional production minus the value he places on those things which he must
give up to attain the extra production. In the case of the weed-conscious farmer,
the net benefits from the herbicide alternative are $2390 per hectare, versus
$1900 for his current practice. Remember that net benefits are not the same thing
as profit, because we have left many costs out of the budget because they are
irrelevant to this particular decision.
While it may appear that this farmer will choose to use herbicides, this is not
clear since there is uncertainty surrounding his yields, and since money may be
quite scarce. In later chapters we will deal with these matters. We now proceed to
apply the concepts just described to make partial budget analyses of some ferti-
lizer experiments.

Partial budget analysis of fertilizer experiments-an example
Table 2 presents the results of 8 maize fertilizer trials conducted in a rainfed rec-
ommendation domain. The purpose of these trials was to derive recommended
fertilizer levels for farmers of the domain. Here we have presented the average
yields obtained from three replications of the treatments. (We have averaged theI
replicates because these averages are the best estimate of the yield which would be
obtained on the entire field in which the experiment was located).
Although it is obvious that there is considerable variability in yields and yield
response from trial to trial, we shall postpone a discussion of the implications of
the variability for farmers' decisions. For now, we'll consider only the average
yields obtained for each treatment over the eight trials, and we will treat the data
just as we would a single experiment. The yield curves in Figure 1 provide a graph-
ic picture of the resulting average yield response.
Table 3 provides a convenient format for organizing the partial budget informa-
tion. We show the alternative choices of fertilizer level as column headings, then
the average yield for each, followed by net yield after adjusting down-ward 10 'o/

FIG. 1. Average yield response to nitrogen.

Partial budget of averaged data from fertilizer trials (per hectare basis)

50 kg/ha P205
25 kg/ha P205

0 kg/ha P205

0 25 50 75 100 125 150
Nitrogen applied, kg/ha

Fertilizer treatment (kg/ha)
N: 0 bO 100 150 0 bU 100 1bo 1 bu 100 ibu
P20s: 0 0 0 0 25 25 25 25 50 50 50 50
2.21 3.14 3.91 4.01 2.44 3.88 4.40 4.84 2.36 4.05 4.74 5.16
1.99 2.83 3.52 3.61 2.20 3.49 3.96 4.36 2.12 3.64 4.27 4.64
1990 2830 3520 3610 2200 3490 3960 4360 2120 3640 4270 4640

0 400 800 1200 0 400 800 1200 0 400 800 1200
0 0 0 0 250 250 250 250 500 500 500 500
T '40([ 8 TYOUW Ti!RF fib~56TF450r Wb W~ f3DW6 T70

0 1 221 12 2 112 2
50 59_ so so so so so so so so so so
O 5 100 10 50 50 100 10 0 50 0 0

O 450 900 1300 300 700 1150 1550 550 950 1400 1800

1990 2380 2620 2310 1900 2790 2810 2810 1570 2690 2870 2840

(1) Average yield (ton/ha)
(2) Net yield (ton/ha)
(3) Gross field benefit ($/ha at $1000/ton)

Variable money costs:
(4) Nitrogen ($8/kg N)
(5) Phosphate ($10/kg Pz0s)
(6) Variable money costs ($/ha)

Variable opportunity costs:
(7) Number of applications
(8) Cost per application (2 days at $25.)
(9) Opportunity, cost ($/ha)

(10) Total variable costs ($/ha)

(11) Net benefit ($/ha)

for assumed harvest and storage losses. The market price for maize in this area is
$1200 per ton, but after making corrections for harvest costs, transportation
costs, and shrinkage, (see Chapter 7), we determine that the field price of addi-
tional yield is $1000 per ton. Resulting gross field benefit is shown in line 3. Of
course, the largest gross field benefit is obtained from the treatment with the
highest yields, which in this case is also the highest level of fertilizer.
In considering the costs associated with each choice, we must be familiar with
the cultural practices used by farmers if we are to determine which inputs are to
be affected by the choice of fertilizer level. In this particular area, horse and plow
technology is the dominant tillage method, and fertilizer is applied by hand.
Therefore, the only inputs affected by this decision are the amounts of fertilizer
and the labor required for application (the value of harvest labor has been deduc-
ted from field price-see Chapter 7). The price of nitrogen at the store is $5 per kg
of N and the price of phosphorus is $7 per kg of P205, but after making adjust-
ments for transportation (see Chapter 6), we determined the field price of N and
P205 to be $8 and $10 per kilo, respectively.
In these experiments, nitrogen levels in excess of 50 kg were applied in two
doses, and we estimate that two man days are required per hectare for each appli-
cation. After visiting with farmers in the area we calculated that $25 per man-day
is a reasonable estimate of the average value of farmers' time, although we recog-
nize that for some farmers in the area the amount should be closer to zero, while
for others it should be more (see Chapter 6). In lines 7, 8 and 9 of Table 3, we
have calculated the cost of labor for each treatment, and in line 10 we show the
total of all variable costs associated with each treatment.
We have now completed the task of assessing the field benefits and variable
costs associated with each of the alternative choices of fertilizer level. But the task
of making a choice among them, from the farmers' point of view, is far from
complete. Next we calculate net benefit, gross benefit minus variable costs, and
record these amounts in line I1.
The listing of net benefit for each treatment, as shown in line 11 of Table 3,
completes the partial budget analysis of the average yields from these experi-
ments. One might be tempted at this point to choose treatment 100-50 as the
fertilizer recommendation for this area. But this would be a poor choice, because
we have so far ignored some crucial aspects of farmer conditions, namely capital
scarcity, yield uncertainty and risk aversion. In the following three chapters, we
consider these complicating factors and their effects on our recommendations.

2/Partial budget
analysis of experiments


In the previous chapter we were careful to include the costs of all inputs which
change with a given production decision. These costs included the cash costs of
purchased inputs but we did not include the cost of using investment capital. By
investment capital we mean the value of inputs (purchased or owned) which are
allocated to an enterprise with the expectation of a return at a later point in time.
By the cost of investment capital we mean the benefits given up by the farmer
due to having the investment capital tied up in the enterprise for a period of time.
The cost of using investment capital (or, more simply, the cost ofcapital) may be
a direct cost, as in the case of a person who borrows money to buy fertilizer and
must pay an interest charge. It may also be an opportunity cost, the earnings
which are given up by not using money, or an input already owned, in its best
alternative use.
We suggested in the last chapter that the cost of capital may be very important
to farmers' decisions. This is because the cost of investment capital for agricul-
tural use is generally quite high, particularly in less developed countries. Interest
charges by local money lenders often are in the vicinity of 100 % per year. This
can effectively double the price of inputs purchased with such loans. Even in the
case of a low cost government agricultural loan program, service charges and insur-
ance fees can result in interest rates which are much higher than the interest rate
announced by the loan agency. Furthermore, most small farmers have very little
capital of their own, and they want to invest it in only those inputs which yield
high returns. This means that the opportunity cost of capital, as well as the direct
cost, is quite high for these farmers.
One way of incorporating the cost of investment capital into the budgeting pro-
cedure is to increase the cost of each input by an appropriate amount. Due to the
critical importance of capital availability, however, we have rejected this approach
in favor of another alternative. We charge no cost to capital in the budgeting pro-
cess, but instead attribute net benefits as a return to investment capital. We can
then compare this rate of return to capital with the rate which farm investment
capital can realize in alternative activities. If the calculated rate of return for a
production alternative is above the opportunity rate of return, i.e. its return in
other alternatives, then we can judge the first to be desirable from the point of
view of the farmer. This assumes that all alternatives are equally risky. This issue
is considered further below.
This brings us to the difficult question of the minimum rate of return which

3/ Capital scarcity

will be acceptable to farmers. Let us consider two separate farmers to see why this
is a difficult question and what we can do about it.
First let's consider Farmer A who can borrow money for production through
his local credit cooperative. If he borrows money for a new production alter-
native, the cost of investment capital will be a direct cost, for he must pay interest
at the rate of 12 %b per year on the loan. Since he will be borrowing for only six
months, the cost of the loan is 6: %/ of the amount of the loan. But he also must
pay a service charge which amounts to 5% of the amount of the loan. Thus the
cost to him for a six-month loan is I1%O/ of the amount of the loan.
Now, if a production alternative promises an average return of just 1 1%, then
Farmer A will not want to adopt the alternative, because after paying the direct
cost of capital he will have exactly zero gain in net benefits. For example, suppose
this farmer can spend $100 on fertilizer and he expects an average increase in net
benefits of $11. If he borrows the $100 from his co-op, he will have to pay $11 in
interest and service charges in addition to his other costs, and his increase in net
benefits will be reduced to nothing.
So we can safely conclude that Farmer A will not choose a production alterna-
tive unless the rate of return on capital is more than 11% the direct cost of his
investment capital. But how much more? This will depend in part on the risk of
the investment, the other important factor which we have not yet included in our
discussion. Farmer A would be well aware that the net benefits as calculated in
our partial budget analysis are based on average yield results. In some years, the
net benefits from the investment may be very low. We will postpone to later a full
discussion of how to evaluate this type of risk, but it should be clear that farmers,
poor farmers particularly, do not want to place themselves in the position of
losing what little capital they have.
Because of this aversion to risk, Farmer A may not want to accept a new pro-
duction alternative unless the average returns (over time) to his scarce capital are
considerably in excess of the direct cost of his capital. As a rule of thumb, we
believe that most farmers of the less developed world will not invest in alterna-
tives unless the average rate of return is at least 20 percent points per production
cycle above the direct cost of capital. We do not claim any great accuracy for this
estimate, but we are convinced it is better to make an estimate of this risk pre-
mium than to ignore it completely. For investment alternatives which are not very
risky, we know that farmers would be willing to accept a smaller risk premium.
For very risky alternatives, we are sure that the required risk premium can be
much higher. Therefore, unless we had more information about Farmer A or
about the riskiness of the alternatives he is considering, we would estimate that he
would not adopt an alternative unless the rate of return for the average yield with
that alternative is at least 31%/ :

Farmer A, Cost of Capital

amount borrowed for fertilizer $ 100
interest for 6 months (12% /yr) $ 6
service charge
total amount of loan $ 111
direct cost of capital (11/100) 1 1 %
risk premium 20'
Farmer A cost of capital 31%~

From agronomic
data to farmer

Now let's consider Farmer B who will not be borrowing, but instead will be
using his own funds to invest in alternative technologies. The opportunity cost of
using his investment capital in a particular alternative is the rate of return which
he would receive from his capital in its best alternative use plus the risk premium
appropriate for that alternative use. We think that in general, a rough estimate of
this opportunity cost is about 40%q per production cycle. Again, we claim no
great accuracy for this rule of thumb, but it is consistent with behavior that we
have observed among farmers of both the developed and less developed agricul-
tural areas, and again it is better to make an estimate, than to ignore the matter
completely. Some people place the figure at 50%O/ or even at 100%O/ and these
levels will be appropriate in some cases, particularly for subsistence farmers in
areas with high yield variability.
To summarize, we have argued that the cost of using investment capital is very
high for most farmers of the world. While the cost of capital will vary from farm
to farm, as a general rule we think that a technology should not be recommended
unless the rate of return to the additional investment is at least 40%6 for the
cropping season. When you have specific information regarding either the direct
cost of capital, the opportunity cost of capital, or the riskiness of the alternatives,
you may wish to use a different rate as a criterion. In Chapter 5 we will discuss
further the measurement and implications of the riskiness of alternatives. In Chap-
ter 6 we present a more thorough discussion of how to estimate the cost of


In Chapter 2, we explained how to evaluate alternatives from the point of view of
average net benefits to the farmer. We suggested that farmers will not necessarily
choose the alternative with the highest average net benefits because of the scarcity
of capital in agriculture and because of the risks that may be associated with the
average net benefits from a given production alternative. In this chapter we bring
these concepts together and show how to derive recommendations which are con-
sistent with both capital scarcity and risks.

The net benefit curve
A very revealing device for summarizing the results of a partial budget is the net
benefit curve. This curve shows the relationship between the variable costs of the
alternatives and the average net benefits from the alternatives. We can best des-
cribe this by plotting the net benefit curve from the fertilizer experiments des-
cribed earlier.
In Figure 2 we have plotted each of the fertilizer treatments from Table 3
according to the net benefit from the treatment and the variable costs of the
treatment. Beside each of the 12 points plotted, we show in parentheses the nitro-
gen level and phosphate level. It is apparent from the points plotted that some of
the fertilizer alternatives would not be chosen by any thoughtful farmer. For
example, the phosphate-only treatments (0-25 and 0-50) have net benefits lower
than the check treatment (0-0), yet require variable costs of $300 and $500 per
hectare. No farmer is likely to choose these alternatives when he could receive a
higher net benefit with zero variable cost. The same is true of treatments 100-0
and 50-50. The average returns from these two treatments are lower than the
return from 50-25, and 50-25 has a lower variable cost. Fertilizer levels such as
0-25, 0-50, 100-0, and 50-50, we refer to as dominated alternatives, because for
each of these there is another alternative with a higher net benefit and lower
variable cost. In normal circumstances, we would never expect a farmer to choose
one of these dominated alternatives.
The choices which are not dominated we have connected together with a solid
line. This solid line is the net benefit curve. Two aspects of this net benefit curve
are noteworthy. The first is that the curve rises steeply at first, then rises more
slowly to a peak and begins to fall. The curve shows diminishing returns to ferti-
lizer expenditures. This is important because it demonstrates clearly that we can

From agronomic
data to farmer

reduce costs considerably from the point of maximum net benefits with little
reduction in those benefits. Said another way, this demonstrates that the returns
from expenditures on initial amounts of fertilizer are much greater than the re-
turns to additional expenditures for larger amounts of fertilizer. Experience shows
that this is often the case for fertilizer.
The second interesting aspect of the net benefit curve is its shape between the
0-0 point and the 50-25 point. The two solid line segments drop below the broken
line connecting these two points, whereas we would normally expect a fertilizer
response curve or net benefit curve to fall above the dotted line. In other words,
we normally expect these curves to begin steeply, with the slope gradually falling
as expenditure on inputs increases. The irregularity of the curve we observe here
may be due to an interaction between nitrogen and phosphate at low fertilizer
levels, or it may be due to chance (even though these are the combined results of
many trials).
Whatever the cause of this unusual shape, the implications for further experi-
mentation are clear. There is surely no reason to conduct any further trials with
fertilizer costs in excess of $650, since it seems clear that net benefits increase
little if any above that point. On the other hand, intuition suggests that there may
be some fertilizer treatments which would result in points above the broken line
between 0-0 and 50-25. Since it appears there might be an important interaction


.25) *(150-50)






200 400 600 800 1000 1200 1400
Variable cost, $/ha

1600 1800

FIG. 2. Net benefit curve for the fertilizer trials. Numbers in parentheses
represent kg/ha of N and P205 respectively.

between N and P205, it would seem wise to experiment further with treatments 4/ Net benefit curves
costing between $300 and $500, such as 40-15, 30-15, 25-25, etc. These treat- andmarginal analysis
ments may result in the discovery of points above the broken line. If so, these are
treatments which further reduce farmer costs without appreciably reducing net

Marginal analysis of net benefits
We have observed that the net benefit curve for the fertilizer data rises quite
sharply at first and then more slowly to a maximum. We have found this to be
true of most net benefit curves. It implies that the rate of return to the invest-
ment in the first units of fertilizer is much higher than the return to the additional
units required to achieve the maximum net benefit. Look at figure 2. You may be
tempted to conclude that not many farmers would want to invest more than $700
per hectare for fertilizer (for 50 kg of N and 25 kg of P205). This is because the
first $700 provides an increase in net benefit of about $800, while the second
$700 provides an increase in net benefit of only $80. To explore this observation
in more detail, we need to introduce the concept of marginal analysis.
The purpose of marginal analysis is to reveal just how the net benefits from an
investment increase as the amount invested increases. Marginal net benefit is the
increase in net benefit which can be obtained from a given increment of invest-
ment. In the fertilizer example, the marginal net benefit from $450 invested in 50
kg of N (the smallest non-dominated investment included) is $390. The next
possible increment of expenditure is to spend an additional $250 for 25 kg of
P205 (taking us to the 50-25 treatment). The marginal net benefit from this
increment in expenditure is $410. The marginal rate of return to a given incre-
ment in expenditure is the marginal net benefit divided by the marginal cost
(increment in expenditure). The marginal rates of return of the first two incre-
ments in fertilizer investment capital are determined as:

2380 1990 = marginal net benefit --0.87 = 87%
4$0 0 = 450 = marginal cost
The marginal rate of return of the second increment is:

2790 2380 = 410 = marginal net benefit 1.64 = 164%
700 450 = 250 = marginal cost

It is clear from the shape of the curve that the marginal rate of return on expendi-
tures above $700 per hectare is quite small. We verify this with later calculations.
It is possible to make a marginal analysis of the fertilizer data without refer-
ence to the net benefit curve itself. The first step is to list all the alternatives from
the highest to the lowest net benefit. We have taken the information from Table
3 to make such a listing as shown in Table 4. Next, proceed from top to bottom
down the list to identify and eliminate the dominated alternatives. For instance,
the second-highest net benefit is obtained from treatment 150-50. But the var-
iable cost for this treatment is higher than the variable cost for the treatment
above it. Thus it is dominated, and can be eliminated (as indicated by italics in
Table 4). Moving down the list, we eliminate any treatment which has a variable
cost equal to or higher than the treatment above it. We are left with five non-
dominated alternatives, which are of course the same as those represented by the
solid net benefit curve of figure 2.

From agronomic TABLE 4.
data to farmer Dominance analysis of fertilizer response data.
Net benefit treatment (kg/ha) Variable cost
($/ha) N P205 ($/ha)
2870 100 50 1400
2840 150 50 1800
2810 100 25 1150
28 10 150 25 1550
2790 50 25 700
2690 50 50 950
2620 100 0 900
2380 50 0 450
23 10 150 0 1300
1990 0 0 0

To proceed with the marginal analysis, we take these five alternatives from
Table 4 and place them in Table 5. Here we calculate and present the marginal
cost, the marginal net benefit and the marginal rate of return for each increment
of expenditure. Beginning at the bottom, the marginal cost of the first increment
is $450, the marginal net benefit is $2380 $1990 = $390, and the marginal
rate of return is thus 390/450 = 87%~0. The marginal cost of the second incre-
ment is $700 $450 = 250, the marginal benefit is $2790 $2380 = $410,
and the marginal rate of return is 410/250 = 164b. The next increment in ex-
penditure, on an additional 50 kg of N for $450, returns only 4%q, but the follow-
ing increment of another 25 kg of P205 returns 24 %o.
The question remains what level of expenditure would the average farmer
choose if he had all this information? We have previously stated that as a general
rule, farmers will not want to make an investment unless the average rate of re-
turn is at least 40 % per crop season. Thus in general, farmers would be willing to
invest both the first $450 for 50-0 and the additional $250 for 25 kg of P205,
for both increments have rates of return well over 40%/. But farmers in general
would not want to invest more. Clearly 4%6 is not a very attractive rate of return,
although 24%b might be for some farmers. But if a farmer were to go from 50-25
to 100-50 (two increments at once), the rate of return would be
80/700 = 1 1.4% This is not a very good rate of return, and it is doubtful that
very many farmers would be willing to make such an investment. Thus using this
marginal analysis approach, we can feel quite comfortable in recommending a fer-
tilizer rate of 50 kg N/ha and 25 kg P205/ha.
But there are other questions which should be asked before an agronomist can
afford to be satisfied with this recommendation. The first question is whether
40%b is the correct figure for the cost of capital. Suppose, for example, that far-
mers in the recommendation domain have access to government credit programs
with an interest charge of 8 % for the crop season. Recall that the cost of capital
in this case can be approximated by adding a 20 %o risk premium to the direct
cost of the capital. In this fertilizer example, this change would not result in an
increase in the fertilizer recommendation, because the next increment in capital
does not return more than 28% But it is quite possible that a reduction in the
cost of capital (interest charges plus risk premium or capital's opportunity cost)
will increase the recommended level of fertilizer.

The second is the question of how risky this alternative is relative to no ferti-
lizer at all or to, say, treatment 100-50. If these investment alternatives are not
very risky, it is possible that farmers would be willing to accept a rate of return
lower than 40%Oo. Procedures for addressing this question will be considered in the
following chapter.
The partial budget analysis and then the marginal analysis of these fertilizer
data have involved considerable effort (though not much compared to the effort
required to carry out the experiments). It is useful to review what we have gained
by it. Had we based recommendations simply on maximum yields, we would have
recommended 150-50, which would have subjected the farmer to very large costs
($1700) compared with the $700 expenditure which marginal analysis shows to
be best. Had we based recommendations on maximum net benefit per hectare, we
would have recommended 100-50 with an expenditure of $1400. But marginal
analysis has shown us that the returns to the last $700 of this amount are much
too low for most farmers. By reducing expenditures from $1400 to $700 per
hectare, net benefit is reduced only $80. Even though yields for the recommend-
ed level are more than I ton per hectare less than the maximum attainable, this
analysis has shown that it will not be in the interests of most farmers to approach
this ma ximum.
Before leaving this topic, we wish to point out a mistake which we would prob-
ably have committed had we not used marginal analysis. We determined that the
rate of return to the investment capital required to go from 50-25 to 100-50 was
11% 0. But what is the average rate of return to the entire $1400 (700 + 700)
required for the 100-50 treatment? The net benefit is $880 higher than for the
no fertilizer treatment, so the average rate of return is $880/$1400 = 63%qbBy
our criterion of 40 %l this seems to be enough to warrant recommending it. But
what we discovered in marginal analysis was that, while the farmer would earn
63 %b on this investment, he would in fact be earning 114 %b on the first $700 and
only 11 % on the second $700.
Clearly, we would have been misleading both ourselves and the farmer if we
had recommended that he spend $1400 on 100-50 on the basis that the (average)
rate of return, 63%q, is very good. We are far more correct to recognize these
marginal changes in rates of return and to make recommendations accordingly.

Marginal analysis of the undominated fertilizer response data (per ha).

Examples of calculations: the amount in column 4, line a (4a) is the amount in
column 2 line a (2a) minus the amount in column 2 line b (2b). Also, 3a =
la 2b, and Sa = 3a/4a.

4/Net benefit curves
and marginal analysis

Change from next highest benefit
Marginal Marginal Marginal
increase in increase in rate of
net benefit variable cost return
(3) (4) (5)
$ 60 $250 24%
20 450 4
410 250 164
390 450 87

N P205
100 50
100 25
50 25
50 0
0 0


Net benefit
(a) $2870
(b) 2810
(c) 2790
(d) 2380
(e) 1990


We have previously stated that farmers want to avoid the possibility of occasional
high losses as they seek higher average net benefits. This is especially true of far-
mers near the subsistence level. For them an occasional net loss can have very
serious consequences.
This view of the farmer has important implications for recommendations. Be-
cause risk aversion is important to the farmer, variability in yields and net benefits
must be important to the agronomist. You cannot be content with recommenda-
tions which promise to increase average net benefits. You must recognize that the
best choice will change from year to year and from field to field, and you must
somehow estimate the risks which this variability causes. On the other hand, this
variability means that you need not try to be very precise in deriving the recom-
mendation from any one experiment.
We're not saying that care and attention to those trials which are made is unim-
portant. We're not saying that care in identifying and estimating costs is unimpor-
tant. We are saying that because of the role of risk aversion and because of varia-
bility there is a limit to how precisely recommendations can be made. In this
sense, excess precision is pretense and a waste of time and funds.
What kind of variability should you look for? The variation that occurs in net
benefits even when you administer the same treatment. This kind of variation
emerges from several sources which can be grouped under two headings: yield
variability and price variability. The purpose of this chapter is to discuss those
sources of variability and what they mean for recommendations.

Sources of yield variability
Yields that farmers get from a particular treatment will not be the same as the
yield that you get. There are several factors which cause this. There will be differ-
ences between the soils, the weather and the pest infestation at your site (or sites)
and the soils, weather, and pest infestation at the farmers' sites. Because of this,
you would obtain different yields at each of these sites even if you conducted
identical trials at the same time. We call this site-to-site variability in yields.
Another type of yield variability is that which you will get from year-to-year at
a given site even with the same treatment. This year-to-year variability may mean
that the treatment which gives the highest net benefit in one year may give a
disastrous loss the next year in the same experiment in the same location.


5/ Variability in net

Fertilizer treatment (kg/ha)
N: 0 50 100 150
P205: 0 0 0 0
360 670 2370 2080
1380 1890 3730 3490
3740 3920 3420 3080
2180 2990 3810 2730
1480 1280 970 670
1450 2200 2830 2610
4270 4420 2960 3130
1090 1650 870 710
1990 2380 2620 2310









For examples of these two kinds of variability look at the data presented in
Table 6. These are the data of the fertilizer trials discussed earlier. Though we did
not mention it, the data from the first four trials are from one year and those for
the second four trials are from a second year. Now compare trial 1 with trial 2.
We've held prices constant in computing the net benefits in the table, so the
difference in net benefits at the two sites is due to the site-to-site yield varia-
bility. You can see that no single treatment gave the same yield in one site as it
did in another.
Now, compare trial 1 with trial 5, two trials conducted at the same site but in
different years. The treatment which gave the biggest net benefit the first year
(150-25) gave one of the smallest net benefits the second year. (You can also
compare site 2 with 6, 3 with 7, or 4 with 8). These comparisons show that no
single treatment gave the same yield at the same site in the two years. That is
year-to-year variability.
These two kinds of variability are facts of life. They make it impossible to pre-
dict accurately what a particular treatment will yield in one place based on data
from a different place or to predict accurately what will happen on a given site
next year based on data from last year. You know that such variability exists.
Farmers also know it. It is good to be skeptical about your ability to predict the
results that a given farmer will obtain in a given year.
Let us for the moment suppose that the eight trials are representative of the
kind of variability that a given farmer or the farmers of an area might expect from
the treatment applied. That is, if a farmer were to apply (50-25), the net benefits
he might expect on his farm in any given year are represented by the column of
benefits under (50-25) in Table 6. Notice that the highest benefit is $4000 and
the lowest is $1620. This is a wide range of variation, with an average net benefit
of $2790. More importantly, notice that no single treatment consistently gives the
highest net benefits across the trials.
There is still another source of yield variability which you should consider.
That is the kind of variability that arises from farmers using different practices
than you use in your experiments.
It is well known that the agronomist typically maintains more control over the
environment of the crop than does the average farmer. This happens even when
you are working on farmers' fields. You will probably take more care in the tim-
ing and thoroughness of planting, in weed control (herbicides are more often used

Net benefits to fertilizer treatments by site ($/ha).

From agronomic
data to farmer
recommend tions

on trials) and in insect and disease control.
The reasons for these differences in management intensity are many. In some
cases it is because farmers are not aware of the techniques which you use. More
often it is because farmers cannot wisely allocate as much time, thought, and
money to their fields as you can and should. Regardless of the reasons, because of
these differences the yields obtained by a farmer using a given treatment on a
given field and in a given year might be different from the yields you would attain
on the same field in the same year. Furthermore, because management intensity
differs from farmer to farmer, different farmers will also get different yields from
the same treatment, even if everything else is the same.
Unfortunately there seems to be no simple rule of thumb to correct for these
manageme nt differences. We can say only that they can dramatically affect yields.
Border effects, for example, can have a substantial effect on the absolute level of
yields. CIMMYT trials comparing small and large plots suggest that the yields for
small plots should be reduced by 20%/, to compensate for border effects.
All in all, you will tend to get higher yields from any given treatment than will
farmers. Some suggest that yields should be reduced by 20 to 30 %l to account for
the more intense management given the experiments.
But this isn't the entire problem. Management practices can cause changes in
the ranking of treatments. For example, data from a trial undertaken in CIM-
MYT's wheat program shows that if the wild oat population is controlled (by
spraying, for example), the 100 centimeter-high Jupateco variety outyields the
taller durum variety Anhinga. But if the wild oat population is quite dense, the
reverse is true. Yields for both varieties decline because of the wild oats, but that
of the durum declines by far less. Another example is planting density and ferti-
lizers in maize. With high plant densities heavy fertilizer applications can give
higher net benefits than light applications. At lower plant densities the reverse can
be true.
You are probably aware of other examples. The examples we have presented
stress the importance of still another source of variability. They also warn that
you should be quite familiar with the standard practices of farmers before you
organize your experiments. You should also try to understand why these practices
are used so that you will know whether farmers will find it difficult to change
There is still another source of variability that we could discuss but it really
isn't necessary. This is the variability that occurs among replications, often called
experimental error but better called "within-site variability." This variability just
signals that the fields are not homogeneous. Farmers know this and they tend to
think in terms of the whole field. When differences are really notable they tend to
make two fields or more where they had one. In any case, you need not regard
within-site variation as another source of variability.
To summarize, there are three sources of yield variability which you must rec-
ognize when you attempt to predict what will be farmers' yields based on data
from trials. They are:
1. Site-to-site variability under the same management conditions;
2. Year-to-year variability under the same management conditions;
3. Management level variability on a given site in a given year.

5/ Variability in net

Adjustment of recommendations for yield variability (minimum returns analysis).
In the analysis of the net benefits in the previous chapter, we considered only the
average yields for each of the treatments. In this chapter we have pointed out the
sources of variability in yields, and examined the variability in net benefits which
resulted from the variability of yields in the fertilizer data. We have already sug-
gested one procedure for incorporating risk aversion into the process of deriving
recommendations. This was to add a 20 %o "risk premium" onto the direct cost of
capital. This is because farmers (and others, too) want a margin of protection. In
general, an alternative which offers an average 20%O/ risk premium will be less like-
ly to lead to ruin from a bad year than will an alternative which offers only a
10 %o risk premium.
But the idea of a 20%"orisk premium is a general rule of thumb. There may be
some new technology options which are basically no more risky than traditional
technology, or perhaps even less risky. But even in this case, farmers would be
likely to insist on a small risk premium because the new option entails the risks
of the unknown. On the other hand, a new option may be much more risky than
the traditional options. This will be true if the new option calls for a large invest-
ment and crop failure is likely to occur.
To examine the relative risks of "disaster" among the alternatives, we use mini-
mum returns analysis. Of all the experimental trials available, we look at the worst
25 %/ or so of the outcomes of each treatment. A comparison of these worst re-
sults will give us some idea of the relative riskiness of the various treatments. If
the recommended practice (from marginal analysis) appears to be very little more
risky than the current farmer practice, you can be even more confident that this
recommendation is a good one for the farmer. If, on the other hand, the recom-
mended practice has "worst" results which are worse than the poorest from
current farmer practice, then you need to reconsider the recommendation. One
way to reconsider is to use an opportunity cost of capital higher than 40%~b. The
exact level depends upon the riskiness which is observed, but risk premiums of
50 %/ or even 100%/, (added on to the direct cost of capital) might realistically
represent farmer circumstances.
A minimum returns analysis such as this will be meaningless unless you have at
least five or six experiments. It will also be misleading if you don't include all of
the experimental (or demonstration) sites in the analysis. It is common practice to
abandon agronomic trials if weather or other factors damage the site to the extent
that the agronomist is satisfied that he will observe no significant yield differences
between treatme nts. Thus if 20 sites are planted, it might be that 5 are abandoned
because of drought, flood, severe insect or disease infestation or other factors. It
is common practice to analize only the results from the 15 "successful" trials. But
this is a mistake, because the farmer must accept unsuccessful as well as successful
results. It is just as important for you to know what results the farmer will obtain
in unfavorable circumstances as it is to know the results for the successful circum-
Therefore, it is very important for you to consider carefully the reasons why a
particular trial has been abandoned. If the cause was an obvious error on your
part (you applied the wrong chemical, broke the plants with a machine, etc.),
then the site could properly be deleted as being unrepresentative of farmers' re-

From agronomic TABLE 7.
data to farmer Minimum net benefits from eight fertilizer trials ($/ha).
N e rlere f!sh _Fertilizer treatment (k/a
Net benefit P205: 0 0 0 0 25 25 25 25 50 50 50 50
Worst 360 670 870 670 410 1620 1090 970 510 1310 1550 1460
Second worst 1090 1280 970 710 1080 1800 1660 1090 680 2150 1590 1490
Average of
worst two 725 975 920 690 745 1710 1375 1030 595 1730 1570 1475

sults. If a part of the plot was damaged by livestock, this could be dismissed as
not representative of results that a farmer would get on his entire parcel. Other-
wise, the data should be included as being representative of farmers' conditions
and therefore, very relevant. In some cases you will collect no yield data from
such a site, even though you want to include the site in the analysis. This is an
unfortunate situation, but if it should occur, you should estimate the harvestable
yield from the entire plot, and assume that this yield occurred on all treatments.
This way the calculated net benefits will reflect the loss in variable costs. This is a
relevant measure of the worst results which farmers could expect from a given
treatme nt (the loss of the variable costs).
We have suggested that you consider the worst net return, but this is not en-
tirely satisfactory. Due to random chance, this level of return may be a lot lower
than the remaining outcomes. Furthermore, the farmer may be able to survive just
one bad outcome, if the others are relatively more favorable. So in addition to the
worst possible outcome, we suggest that you look at the average of the worst
25Yo or so of the outcomes for each treatment.
In table 7 we show the worst net return from the eight trials for each treatment
(taken from Table 6). For this set of experiments we are lucky, for the treatment
which we chose using marginal analysis (50-25) is also the treatment which has
the highest net return in the worst of the eight situations ($1620). Therefore, a
farmer who is concerned about occasional low returns could not do better than to
choose 50-25.
The last line in Table 7 shows the average net returns for the worst two
outcomes of each treatment. Again, the previously selected treatment, 50-25, pro-
vides nearly the highest average (50-50 provides an average which is $20 higher,
but this cannot be a very important difference to the farmer).
This analysis of minimum net benefits has provided a check on the riskiness of
the treatment chosen by marginal analysis as compared with other alternatives. In
this case, the previously chosen treatment has less down-side risk than do the
others, so it seems to be a good choice for risk averters. Often, however, the alter-
native selected by marginal returns analysis will prove to be inferior to others in
minimum net return. In such a case, you will need to assess the importance of risk
aversion to the farmers in the recommendation domain before you can decide
whether or not to alter the recommendation because of the result of minimum
returns analysis.
A further comment is in order with respect to the minimum net returns real-
ized from a particular treatment. Ocassionally, something goes wrong in an experi-


ment and one or more of the applications of a treatment may have very low yields 5/ Variability in not
relative to other replications or other treatments. If this is the case, that yield bnft
figure mnay result in a "worst" return for that treatment that is misleading. Thus
in examidning the array of outcomes as in Table 7, you should view with suspicion
any net returns which are far below both other net returns for that treatment and
net returns for other treatments. You should go back to the field book to deter-
mine whether or not some extraneous factor was lowering the yields for just one
of the treatment plots.

Price variability and sensitivity analysis
In making a partial budget, you will be unable to accurately estimate prices or
costs. This is especially true of the prices estimated for the product and for labor.
Variability from year-to-year and farmer-to-farmer in prices paid or received is a
fact of life which you must somehow consider.
With product prices it is sometimes tempting to use guaranteed prices. We all
know, however, that the prices which farmers actually receive often differ from
guaranteed prices. Sometimes they are higher. Often they are lower. This is why it
is essential for you to find out what prices farmers are actually receiving. Even so,
your estimates of product prices are apt to be in error because of season-to-season
or year-to-year variability which you cannot anticipate.
Your estimates of labor prices are apt to be in error because some farmers will
have a higher or lower opportunity cost for their time than do other farmers.
The implications of these difficulties in estimating prices may or may not be
serious. Fortunately, it is usually easy for you to determine whether this is the
case. This can be done by using a technique called sensitivity analysis. The object
of this procedure is to change the product (or labor) price within reasonable
bounds of the original estimate to determine if the ranking of alternatives is af-
To demonstrate this technique we will first apply it to the question of whether
errors in estimating labor price could have an important effect on our fertilizer
recommendation example. Look again at Table 5, the undominated treatments.
Note that of the five treatments listed, the first two require 4 extra days of labor,
the second two require 2 extra days of labor, and the last, the check plot, requires
no extra labor. What effect would a change in labor price have on the ranking of
the treatments?
Using the previously established field price of labor, $25 per day, treatment
100-50 returns a net benefit $80 higher than treatment 50-25. Note however, that
if we increased the field price of labor to $65 per day, both would return about
the same net benefit.

50-25 100-50
gross field benefit $3490 $4270

variable money costs 650 1300
variable labor costs (at $65/day) 130 260
total variable costs 780 1560

net benefit

$2710 $2710

We have already noted that the alternative 100-50 does not quite offer enough
extra net benefit to warrant the extra expenditure of fertilizer over 50-25. For
farmers whose opportunity cost for labor approaches $65, it would offer no in-
crease in net benefits whatsoever. This is one more reason for being reluctant to
recommend 100-50, even though it has the highest estimated net benefit. The
marginal return from the additional investment falls rapidly with higher labor
Comparing 50-25 with 0-0, however, we can determine that for any labor field
price up to $212 per day, the former alternative would still offer a higher net
benefit. Since this figure is far above our estimate, we can be confident that
errors in estimating the field price of labor will not affect our recommendation of
Suppose now we were interested in whether maize price changes of up to 20%
would effect the fertilizer recommendation. One could complete the entire bud-
get analysis again using field prices of $800 and $1200 per ton, but this is not
really necessary. We know that if the maize price rises the return to all levels of
fertilizer will increase, and the main question of interest is whether the returns to
100-50 increase enough to warrant its recommendation for poorer farmers. Given
a field price of $1200 per ton, the net benefit for 100-50 would increase from
$2870 to $3724:

From agronomic
data to farmer

at a field price of at a field price of
$1200 $800
50-25 100-50 0 -0 50-25
$4188 $5124 $1592 $2792
- 700 -1400 0 700
$3488 $3724 $1592 $2092

gross field benefit
variable costs
net field benefit

marginal net benefit

marginal rate of return





The net benefit from 50-25 would be $3488, and the rate of return of the extra
fertilizer would be 236/700 = 0.34 = 34% compared with 24%6 at the old
price. This is nearly a high enough return to warren its recommendation to far-
mers. If there were a good chance that a field price of more than $1200 would
prevail, we would want to reconsider the recommendation.
At a maize field price of $800, on the other hand, the question is whether
50-25 remains profitable enough to be recommended. At this field price, the in-
crease in net benefit over the check plot is about $500, compared with $800 at
the original price, and the rate of return falls from 114%0 (800/700) to 71%0
(500/700). This is still adequate to warrant recommendations of 50-25.
So the result of this maize price sensitivity analysis is that the recommendation
for most farmers does not change for maize prices within 20%O/ of our best esti-
mate price, $1000 per ton, though it might for prices in excess of $1200. Sensi-
tivity analysis with respect to maize price and labor price has given us further
confidence that the 50-25 recommendation will indeed be in the farmers'
interest, even if prices should differ from what we expect.


In Chapter 2 we discussed the general procedures for reporting gross benefits and
variable costs, but we said very little about the problems and procedures involved
in estimating costs and benefits. In this and the following chapter we discuss in
more detail how these estimates are made, and provide check lists which can help
you to insure that you do not overlook significant costs or benefits.
The first task in estimating costs is to identify which input items are changed
in any way by changing from one treatment to another. These inputs are called
variable inputs. They include changes in chemicals, seed, amount or type of labor,
and amount or type of machinery. The second task is to determine field price of
that input the money cost or opportunity cost per unit of the input.

Identifying and measuring variable inputs
To identify which inputs are affected by the alternatives included in an experi-
ment, you must be familiar with local practices as well as the practices used on
the experiment. You must then ask yourself which operations change from treat-
ment to treatment, and which operations are different from those used by the
farmers in the recommendation domain.
Following is a check list of operations which should be considered:

Land preparation
is it the same for all treatments and on farms?

is the same seed used for all treatments?
is the same amount of seed used?
is the planting technique the same?

is there reason to think the amount of time required
for this chore will differ from treatment to treatment?
is the technique the same for all treatments and on farms?

is it required for all treatments?
is the amount of time required the same?
do farmers do it?


From agronomic Application of pesticides and fertilizer
data to farmer are these practices identical for all treatments?
If the practices for the above operations are not identical for all treatments and
for the farms in the recommendation domain, one must then consider which of
the following types of inputs might be affected by the differences, and by how

C~hemicals-(fertilizer, insecticide, herbicide)
do they differ in either type or amount?

does it differ in type or amount?

are the kinds of equipment needed the same?
is the amount of equipment time required the same?

how much does labor differ due to differences in operations
-weeding practices, thinning practices, irrigation practices,
planting density, land preparation, etc.
does the amount of labor required vary significantly
with type or amount of seed applied or fertilizer applied?
is the type of labor required different between treatments?

For inputs such as labor and equipment time, it is usually difficult to make
estimates of the differences in the amount required for each treatment. Informa-
tion about labor use from the experimental plots, even if they are conducted on
farmer fields, is not very useful because of the small size of the plots. The best
way to get this information is to visit with several different farmers. Each will
have his own opinion as to the time required for various operations, but a number
close to the average of these opinions will provide a good estimate. For activities
with which farmers are completely unfamiliar, a guess will have to suffice. Re-
member, not all farmers take the same amount of time for a given job, so your

Assessing variable field costs for a particular treatment
(per hectare).

Field cost ($/ha)
Money Opportunity
Number price price Total
Operation Input of units per unit cost per unit cost cost
Planting seed 15 kg 1 15 - 15
labor 2 days -- 25 50 50

Fertilizer N 50 kg 8 400 -- 400
P205 25 kg 10 250 -- 250
labor 2 days - 25 50 50
Total variable costs 665 100 765

estimate cannot be precise. The danger is that you will overlook an important
change in labor requirements.
Once variable inputs for each operation have been identified and their amounts
estimated, it is sometimes useful to record them in an orderly manner such as the
first three columns of Table 8. We say sometimes, because in relatively simple
experiments such as fertilizer trials, only fertilizer and application labor are varia-
ble inputs, and they may be recorded directly in a partial budget table such as
Table 3. But for experiments with a larger number of variable inputs, such as a
trial demonstrating technological packages, a table such as Table 8 will be very
helpful to you in organizing field cost information. Differences in field cost from
treatment to treatment can then be quickly determined by comparing the total
costs from the table for each treatment.
To this point we have discussed only the identification and measurement of
variable inputs, the first three columns of Table 8. We now turn to some consider-
ations related to assessing the cost of each of these.

Determi ning the field prices of purchased inputs
How can you estimate the field price of inputs which are purchased and used up
during the season? (This would include such items as seed, pesticides, fertilizer,
and irrigation water.) It's very easy. Go to the local retail outlets or wherever the
farmers must buy the inputs, and check the retail price for the appropriate size of
Then find out how the farmers get the input to the farm. In the case of non-
bulky inputs such as insecticides and herbicides, the item can be carried by the
person and transportation costs are insignificant. But for fertilizer, and perhaps
seed, this is not the case. Usually the farmer must use a truck or perhaps a pack
animal to get the input home. If this is so, a transportation charge must be added
to the retail price. If the farmer pays others to transport the item for him, it is not
difficult to learn what the normal charges are. If he transports it himself, one may
want to include the opportunity cost of his own time and for his own truck.
When budgeting for farmers in general, one will have to be guided by the practice
which would be followed by the majority of the farmers in the recommendation
In some situations, the farmer will be selecting seed from his previous crop,
rather than buying the seed. This seed is not costless either, as he has other alter-
natives for it. In general, the opportunity cost of this seed should be the local
market price, less transportation and marketing costs, plus cost of storage and
seed treatment (if any).

Determining the field price of equipment
Some treatments or alternatives may require the use of small hand equipment not
required by other treatments. Then for the types of budgets we are describing,
you must derive a field price per hectare for the use of the equipment.
The retail price of the equipment is the appropriate starting point in deter-
mining field price per hectare of use. To obtain a pro-rated cost per hectare of
use, one can first divide the retail price by the approximate life of the piece of
equipment (in years). This provides a pro-rated annual cost, which must then be

6/More on estimating

From agronomic divided by the average number of hectares per year grown by farmers in the area
data to farmer to obtain the pro-rated per hectare field price of the piece of equipment.
recommendations For example, suppose you are considering recommending a herbicide which is
applied with a knapsack sprayer which costs $500. You might estimate that most
farmers could use this for 5 years, and that the average farmer has 5 hectares of
the crop. You can calculate the per hectare cost as

$500/5 years = $100 per year
$1~00/5 hectares ='$20 per hectare per year

In rare cases you may be considering alternatives which differ in their use of
tractor-drawn implements or perhaps small self-powered implements. The above
procedure can be used for this type of equipment also, but there are other factors
involved in the cost, such as repair costs, fuel costs and the possibility that the
equipment would have other uses on the farms. Thus for these larger pieces of
equipment, it is best to seek the advice of an agricultural engineer or an agricul-
tural economist who is familiar with the equipment and costing techniques.
The above approach to estimating equipment field price may seem a bit crude,
and it is true that more sophisticated costing techniques could be used. But one
cannot hope for much precision in estimating these costs per hectare, as they may
vary widely from farm to farm. Once again, precision can be a waste of time and
money. And it is far better to use a crude method of estimating the costs then to
ignore them altogether.
Thoughtful readers might have noticed that we have included no interest
charges in this procedure for determining the field price of equipment. This is
because we are using the rate of return to investment capital as a decision crite-
rion (Chapter 4). If the rate of return is not at least as high as the interest rate, we
will not recommend the treatment which requires the equipment.

Determining the field price of labor
For farmers who hire labor for the operations in question, the field price of labor
is the going wage rate for day laborers in the area, plus the value of non-monetary
payments normally offered, such as lunch. (The value of such non-monetary pay-
ments may not be trivial. In parts of Pakistan for example, the value of lunch
represents a fourth of the wage.) There are two problems with using this price.
First, it may be that most farmers for whom the recommendations are intended
do not hire outside labor and will do the work themselves with family labor.
Second, it may be that the operation in question is of such a critical nature that
the farmer would not want to entrust the task to anyone other than himself.
Where the farmers or members of his family will in general be doing the work,
we must use the concept of opportunity cost to determine labor's field price.
Opportunity cost represents value which is given up in order to do the work and
thus represents a real cost. For example, if the farmers must take a day off from
his job in town to do the extra work, he will give up at least a day's wages, and
this opportunity cost is just as real as if he had paid someone else to do the work
for him.
As we mentioned earlier, it may be that the extra work is required at a time
which is critical in the care of some other crop, such as tobacco. If taking a day

from the more important crop results in a reduction in earnings from that crop,
then that loss is the opportunity cost of the labor. Again it is a very real cost, even
though it does not directly involve money.
It is all very well for us to give you the opportunity cost principle as the
approach to estimating the field price of labor, but how can you discover what
the opportunity costs are for the average person for whom the recommendation is
to be made? The point of departure is the going agricultural wage rate for the
season in the area, which can be discovered by talking to several farmers. Remem-
ber that it is not unusual to find the going rate higher during some periods of the
year than during others.
Then you must call upon your familiarity with farming practices to determine
whether the extra labor will be required at a time when the family labor will be
fully occupied, or if it will occur at a time when there is likely to be slack labor
available. If the extra labor is to occur during a relatively slack period, we suggest
an opportunity price of about 50 to 75 %b of the going wage rate. We suggest this
lower price because the farmer will have the opportunity, if he wishes, to work
off his farm, in which case he could have earned the going rate for the season. But
since it is probably some trouble for him to obtain outside work, and since he
probably prefers to work for himself, we think most farmers are willing to work
at home for somewhat less.
We have suggested the figure of 50 to 75 %o but this is, of course, a rough esti-
mate of values which probably vary from farm to farm. We want to caution you
not to be swayed by the possible fact that the farmer would be sitting around
doing nothing if he were not doing the extra work. For if jobs are available, and
he does not choose to take them, this is evidence that he values his leisure time
more than the amount which he could obtain by working. Of course, if no off-
farm work is available for most farmers, it may well be that the opportunity cost
of most farmers' time is very much closer to zero. In this case, the opportunity
cost of labor can be set even lower than 50 %l of the going wage, but in no case
should it be set at zero.
On the other hand, if the extra labor is to occur during a very busy time, when
the farmer is likely to be able to earn more in another enterprise, then we suggest
using an opportunity cost of about 125 % of the going wage rate for that season.
While the opportunity cost of the farmer's time may be more than this, he usually
has the opportunity of hiring in workers to assist him. Since it is some bother to
him to do this, the true cost of hiring in the labor would be more than the going
wage rate, and this is why we suggest the figure of I25%~b. (While the busy farmer
may not actually hire the labor, the fact that he does not do so indicates that he
does not feel that the value of labor in the alternative uses is worth more than
125%O/ of the wage rate.)
Summarizing what we have said about the field price of labor, we point out
that the going agricultural wage rate (including lunches, etc.) in the area for the
season of the year in question is the starting point for estimating the opportunity
price of labor. If the farmers for whom the recommendations are being made will
be very busy at this time of year, then we suggest a figure of 125 %o of the wage
rate (for that season) as the opportunity cost. If the farmers can be expected to
be less than fully employed at the time in question, we suggest a figure of 50 to
75 % of the wage rate for that season. In Chapter 5 we described a way to see

6/More on estimating

From agronomic how important the estimated field price of labor is in choosing a treatment to
data to farmer recommend.

Determining the cost of investment capital
The opportunity rate of return is the concept which we use to estimate the cost
of using capital, and though we do not use it in calculating field costs, we do use
it to derive recommendations as described in Chapter 4. We now want to discuss
in more detail how the opportunity rate of return may be estimated.
Suppose a partial budget analysis of a $100/hectare investment in fertilizer
shows an average net benefit of $25 per hectare. This is a rate of return of 25 %
per 6 months. We now need to estimate the opportunity rate of return to invest-
ment capital if we are to decide whether or not this 25 %/ is satisfactory.
If the farmers will be borrowing money to finance the investment, the rate of
interest which they must pay on their loan is a first approximation to the oppor-
tunity rate. But you must not forget service charges and insurance premiums
associated with the loans but not included in the interest rate. These charges will
often cost more than interest charges, thus perhaps doubling the effective interest
rate which the farmer must pay. Also, one must consider that the loan interest
rate is expressed in percent per year, while the period of investment in fertilizer
may be only six months. Perhaps we can best show how to account for these
factors with an example.
Suppose the farmers can borrow from the agricultural bank to buy this $100
worth of fertilizer. The annual interest rate is 12%~, there is a $5 service fee, and a
charge for loan insurance. The bank makes a loan for $121, determined as

$ 100 cost of fertilizer
x0.12 interest rate per year
$ 12 annual interest charge
x 0.5 fraction of year
$ 6 interest charge

$ 100 cost of fertilizer
6 interest charge
5 service fee
10 insurance prennum
$ 121

But this bank, like most others, actually gives to the farmer only the value of the
fertilizer, $100, and asks the farmer to pay back $121 at the end of 6 months.
The cost of investment capital is the total charges for the loan divided by the
amount received, or

21/100 = 0.21 = 21%/, the cost of capital.

The effective loan rate which this farmer pays is 21%O/ per 6 months (42%Olper
year). The fertilizer investment returns 25 %r per 6 months (a total of $125), a
little more than enough to repay the loan. But most farmers would require a risk
premium of 20 %l or more above the effective loan rate to provide an income
safety margin, given the risks of production. This would increase the cost of capi-
tal from 21 %o to 41%O/ The 25%q rate of return on this fertilizer investment would

probably not be sufficient for most farmers. The greater the yield and price un-
certainty, the larger average return over loan costs will be required to convince the
farmer to invest.
Now let's consider those farmers who will invest their own money in fertilizer.
The opportunity rate of return to their own capital is (1) the rate at which they
could loan their capital to others (with comparable risks as for investment in ferti-
lizer) or (2) the rate which they could earn by investing their capital in alternative
enterprises of similar riskiness. Unfortunately it is much easier to conceive of
these two rates than it is to measure them. However, our experience with poorer
farmers around the world suggests that local (private) interest rates are generally
quite high, up to 100% oas we mentioned earlier, and that available investment
opportunities on farms generally promise rates of return of 30%q and more. Thus
we have suggested a figure of 40 %b (per crop season) as a minimum opportunity
rate of return. Where the variability of net benefits is high, the figure should be
50 %b or more. In areas where the private money-lending trade is active, the rate of
interest for these loans can be used as the opportunity rate of return.
Let's summarize what we have said about capital charges and rates of return.
We have not made any charge for the use of capital for inputs in our partial bud-
get approach. Rather, we have calculated marginal net returns as a percentage of
marginal variable costs, and have compared this rate of return with the oppor-
tunity rate of return to determine if the rate of return is sufficiently greater to
warrant undertaking the risk involved. Where loans are widely available for fi-
nancing the investment, the rate of return should be around 20 %~ above the effec-
tive loan rate for investments with average risk. Where farmers will generally be
financing the investment themselves, we have suggested an opportunity capital use
cost of 40 % per crop season for investments with average risk.

We have presented many details which you should consider in estimating variable
costs. These details may appear tedious, but they will not seem so once you incor-
porate them into your way of thinking about the value of your research to far-
mers. The details are important. Failure to recognize all of the important costs
associated with a treatment can destroy the credibility of your recommendations.
To assist you in identifying these important costs, we offer the following check

Check list for estimating field costs
1. Identify all operations which will be performed differently from treat-
ment to treatment, including:
land preparation
planting (density, technique, seed)
application of pesticides and fertilizers
2. For each of these operations, note which inputs are different, and estimate
the quantities required, including:
chemical inputs (fertilizer, insecticide, herbicide of the correct kind)
seed (kind and amount required)

6/More on estimating

From agronomic
data to farmer

equipment (kind and amount required)
3. Determine the field price of each of the above inputs
purchased inputs
retail price (for appropriate size of package)
transportation costs
retail price
average years of life
average crop hectares for farmers of area
going agricultural wage during relevant season
full employment or slack employment period
effective loan rate if loans are generally available
information about private credit rates


In Chapter 2 we presented an overview of how to assess the benefits and costs of
alternative recommendations using partial budgeting. Here we will look more
closely at some of the problems that can arise in assessing benefits. We will discuss
the identification of sources of benefits and the assignment of values to benefits,
and conclude with a check list for easy reference.

Identifying and assessing benefits
Recall our earlier statements about the need for completeness. What that means
for benefits is that you must identify all items which:

1. have a positive value to the farmer
2. which change from one treatment to another

Let's return to the example of maize production presented in Chapter 2. Maize
grain has value to the farmer and the data presented in Table I show maize yields
changing as the application of fertilizer changes. But the farmer will not salvage
the entire harvest, as some grain will be lost in the harvest, shelling and storage
process. For this case, we estimated that the farmer will lose 10%of the grain in
these operations. The net yields (90%Ooof the measured yields) represent the quan-
tities which the farmer will benefit from.
Should we distinguish between maize sold and maize consumed on the farm?
One could argue that sales provide income while maize used on the farm does not,
hence that only the maize sold is a source of the benefits. Clearly that is a short
sighted view as the maize used on the farm has value in satisfying nutritional re-
quirements. Indeed it also has potential value in the market it could be sold for
cash and the cash used to buy something else which could satisfy the require-
ments of food and feed.
Let's agree, then, that we want to value the total production of maize whether
sold or consumed on the farm. Notice here that maize can be used on the farm in
three ways: as seed, as animal feed, and as human food.
We are now left with the problem of valuing the maize produced.
A first approximation of the value is the market price. From this you must
deduct certain costs associated with getting the maize from the field -the point at
which the farmer makes agronomic decisions- to the market.

Which costs must be deducted? The rule we suggest is to deduct from the mar-
ket price all of those costs which will vary directly with the quantity of maize
produced. To see what these costs are, suppose yields were zero. Then there are
no harvest costs, no costs for storage, no costs for bagging the grain, no costs for
transporting grain to market. Alternatively, for a harvest of 2 tons, the cost of
harvesting, shelling, storing, bagging, and transporting 2 tons is almost exactly
twice the cost of performing the same activities for a harvest of 1 ton. These
costs, harvesting, shelling, storing, bagging, and transporting, are the costs which
vary in direct proportion to production and should be deducted from market
price. The remaining value is the field price of maize.
Before continuing, we need to describe what we mean by market prices. These
are not the retail prices in central cities. They are the prices which farmers can
expect to receive in local markets. Because prices vary over the course of the year,
it is a good idea to get an average price. If you want to be conservative, you can
use the price expected at harvest time. Again the price to use is not necessarily the
official price. We all know of cases where prices to farmers have been higher or
lower than official prices. We want the price our decision-making farmer antici-
pates, whatever it is.
It might be asked why the cost of, say, weeding isn't also deducted from mar-
ket price. This is because the cost of weeding is not a constant proportion of
yields. It is this distinction which differentiates those costs which can be sub-
tracted from the market price and those which can better be treated separately.
Returning now to the costs, you'll recall that in the previous chapter we distin-
guished between money costs and opportunity costs. We must do the same thing
here. Consider Table 9.
If payments are to be made in cash -for buying bags, for hiring custom shelling
or threshing, for transportation- then one need only write in that cost per ton. If
it is contemplated that an activity or several activities will be done with family
labor, then the concept of opportunity cost must again be used. In the example
we've assumed that the farm family will harvest the maize. Again, all such costs
must be assigned in per-ton terms. If for example, a day's labor is valued at $25
and if in one day a worker can harvest 0.3 tons, then the cost per ton is
$25/0.3 = $83.30.
Storage needs special consideration. The items which figure in costs are fumi-
gants and insecticides along with a charge for the cost of constructing the space
which the grain occupies. It is likely that the cost of storing grain will be small on
a per ton basis. We include it for completeness and because, while small, it is still
greater than zero.
Storage losses create another complication in calculating proportional costs.
Suppose you pay $135 ( 1 83 + 17 + 35) to harvest and store a ton of grain.
Suppose further that 20 % of the grain is lost in storage. The cost per ton of grain
remaining after storage is then $135/0.8 tons or $ 169/ton, rather than $135/ton.
The storage loss of 20%~ has increased these proportional costs by 25%0. In situa-
tions where storage losses are large, you should make this correction in deter-
mining proportional costs.
Subtracting proportional market costs from the market price of maize gives the
fieldprice of maize, $1200 200 = 1000. This is the value to the farmer of a ton

From agronomic
data to-farmer

of maize standing in the field. Notice how much lower it is than the market price.
These costs will normally be at least 10%o of the market price, and you should not
ignore them.
It is now time to reconsider the problem of identifying sources of value which
will vary among alternative treatments. Maize for grain has been discussed above.
Does anything else vary? Clearly yes, the production of fodder varies. If fodder
has a value, as it so often has in poor countries, then you should estimate its gross
field benefit also.
The procedure for estimating the gross field benefit for fodder is exactly the
same as that for estimating the value of grain. First you must estimate production
and deduct anticipated losses to get net production. Then follow each step out-
lined above for assessing costs. Of course "harvesting" becomes "collecting",
"shelling"' becomes "baling" and it seems likely that the costs of storage and of
bagging will be virtually zero. The important thing to remember however, is to
consider each potential activity will "chopping" be included? and follow the
procedure laid out for maize to estimate the proportional cost per ton of process-
ing the fodder from field to market.
Once the per ton proportional cost is estimated, all that remains to do is to
subtract that amount from the market price to get field price and then multiply
field price by adjusted production. The result is the gross field benefit from
Adding the two benefits together that from grain and that from fodder -
gives the estimated gross field benefit from the treatment.. It then remains to con-
tinue with the cost categories as described earlier.
Now, it is unlikely that calculations for maize or wheat will show potential
benefits from more than two sources, grain and fodder. For other crops or for
crop mixtures more than two sources of benefits might well emerge. Again the
procedure for treating each potential source of benefits is the same as the proce-
dure described above for handling maize.

So far our discussion has assumed that the farmer is an owner-operator of his
farm. However, in many farm communities considerable numbers of farmers are
tenant farmers. The form of tenancy varies widely from country to country and
from region to region. It is common for the landlord and tenant to share the crop
according to some formula with the tenant supplying all the cash inputs such as

Assigning costs (per ton) to activities proportionally related
to production.
Money Opportunity Total
Harvest -$83 $83
Shelling/threshing $17 -17
Storage 35 -35
Bagging 25 -25
Transport 40 -40
Total proportional costs per ton 117 83 200


7/More on estimating

From agronomic fertilizer, seeds, etc. If we assume that the (tenant) farmer wants to reap as much
data to farmer net benefit as he can, then such crop-sharing arrangements may exert a very signi-
recomendaionsficant influence on the choice of practices.
To demonstrate the importance of this, let's assume that the maize fertilizer
recommendation discussed previously is to be made for tenant farmers who pay
all the costs of inputs, but receive only half of the production. We need to calcu-
late the net benefits again under this assumption. For the 0-0 treatment, for
example, the net benefit would be 0.995 tons x $1000 = $995, exactly half of
the previously calculated figure. For the 50-25 treatment, the net benefit is:
1.745 tons
x$ 1000 field price per ton
$ 1745 gross field benefit
700 total variable costs
$ 1045 net benefit

instead of $2790 as calculated for an owner-operator. If you proceed with these
calculations for all treatments, you will find that none of the other fertilizer treat-
ments give a net benefit as high as the $995 from the 0-0 check plot. Then treat-
ments 0-0 and 50-25 are the only undominated alternatives for the tenant farmer.
The marginal rate of return on the $700 investment for 50-25 is:

$ 1095 $ 995 = $100 marginal net benefit
$ 100 / $700 = 0.14 = 14olomarginal rate of return

This is not an adequate rate of return to warrant recommendation of any fertilizer
to tenant farmers who must pay all the fertilizer costs and receive only half the
crop. This is a drastic change from our previous conclusion. It demonstrates that
you cannot afford to overlook the effect of tenancy arrangements when you cal-
culate net benefits.
And what about tenants who pay a fixed rent for their land? A little thinking
about our earlier discussion where we said that things which don't change with
treatments can be eliminated argues that we needn't worry about cash rents.

Some persons might regard the foregoing discussion of costs and returns as being
unduly concerned with detailed aspects of accounting. While there may be an ele-
ment of truth in such a point of view, we contend that disregard for some of these
"little details" has been an important factor in explaining the non-adoption of
technology which was thought to be "profitable." People who make recommen-
dations and who desire not to be surprised by low rates of adoption should recog-
nize these little-but-important details. They should also bear in mind the likeli-
hood and impact of high opportunity costs of farm labor and scarce financial

Check list for benefits
.1. Identify all sources of potential benefits which can be expected to vary
from one treatment to another -for cereals these are likely to include
only grain and fodder.
2. For each potential source of benefits, estimate harvest and storage losses
and calculate net yield. Make adjustments for tenancy if appropriate.

3. For each potential source of benefits estimate a market price or oppor-
tunity price, with proper attention to quality discounts.
4. Identify all activities whose costs vary proportionately with production
per hectare. These are usually the processing activities from harvest to
market, including harvest, shelling/threshing, bagging, storage, and trans-
5. Estimate the unit cost, for example per ton, of each of the activities iden-
tified in (4). Adjust for storage losses if appropriate.
6. Add the adjusted costs per unit of the activities identified for each poten-
tial source of benefits (e.g. for grain and for fodder) and subtract each
total from the relevant market price. The resulting values are the field
prices of grain, of fodder, etc.
7. For each potential source of benefits, multiply the field price times the
net yield and sum over all potential sources of benefits. This is the gross
field benefit of the treatment.

7/More on estimating


I. Calculate average net benefits for each treatment.
A. Estimate benefits for each treatment (see check list, Chapter 7)
1. Calculate average yields for each treatment including grain and fodder if
appropriate. Adjust yields first for differences between experimental man-
agement levels and farmer management levels (0-50%/). Then adjust for nor-
mal harvest and storage losses (at least 10%/).
2. Estimate the field price of grain and fodder. For sellers, this will be the
local farmer market price less cost of harvest, shelling/threshing, storage,
transportation and marketing. These costs will generally total at least 10%/
of the market price, sometimes much more. For subsistence farmers, local
market price plus transportation and marketing costs may be more appro-
3. Multiply field price times adjusted average yield for each product and
sum to obtain gross field benefit for each treatment. Correct for tenent's
share if appropriate.
B. Estimate variable costs for each treatment (see check list, Chapter 6).
1. Identify the variable inputs, those items which are affected by the
choice of treatment. Include chemicals, seed, labor and equipment. Esti-
mate the quantity of each of these inputs used for each treatment. To esti-
mate the quantity of labor and equipment required under farmer condi-
tions, familiarity with farmers' practices is required.
2. Estimate the field price of each input. Normally this will be retail price
plus transportation costs for purchased inputs. Field price of labor will nor-
mally be an opportunity cost. Start with the farm labor wage rate and ad-
just if the labor is needed at a very busy season or a very slack season.
3. Multiply the field price of each input by the quantity and sum over in-
puts to obtain the variable cost for each treatment. This will include a
money cost component and an opportunity cost component.
C. Subtract variable costs from gross field benefit to obtain the net benefit for
each treatment.
II. Choose a recommended treatment using marginal analysis.
A. Array treatments from high to low net returns. Eliminate dominated treat-
ments. Calculate the rate of return to each increment in capital. Graph the net
returns curve if several treatments are involved.

B. Select as the recommendation the treatment which offers the highest net
benefit and a marginal rate of return of at least 40 % on the last increment of
III. Check the suitability of the recommendation from the point of view of yield
and price variability.
A. Use minimum return analysis to compare the minimum returns from the
selected treatment to those from all other treatments. If it compares unfavor-
ably, a different recommendation may be more consistent with local farmers'
B. Use sensitivity analysis to determine whether the choice of recommenda-
tion is sensitive to product or input prices which are particularly subject to
estimation error. If the recommendation is sensitive to these changes, consider
changing the recommendation or obtaining more information about the prices
in question.

s/summary of
procedures for deriving


Our purpose here is to describe two more examples in which we use proce-
dures of the manual to derive farmer recommendations from agronomic data.
These two examples are different from the fertilizer example we used in the text
in that they are more nearly "yes-no" problems rather than "how-much" ques-
tions. The first involves a choice between two treatments, the second involves a
choice among six. As we have mentioned, the procedures of the manual are useful
in both "yes-no" and "how much" situations.
In the first example we examine a series of on-farm maize trials with two treat-
ments: The "current" technology package and an "intensive" technology pack-
age. The question is whether or not to recommend the intensive package. In the
second example we examine a series of wheat trials in which there were six treat-
ments: three varieties, each with and without fertilizer. The question here is
which of the treatments should be recommended to farmers.

Maize technology packages
A series of demonstration/research trials were conducted in three high tropical
valleys. The intensive technology package included fertilizer at the rate of 100
kg/ha of N and 40 kg/ha of P205, one soil insecticide application, two foliar in-
secticide applications, and herbicide. The current technology package, which was
designed to represent current farmer practice, included half the amount of ferti-
lizer, no soil insecticide, and hand weeding rather than herbicides. In all other
respects the two packages were identical. The purposes of the trials were to dem-
onstrate to the farmers the results they could obtain with the two packages, and
to evaluate the desirability of the intensive package over a number of locations
(26) in the recommendation domain. (Other trials were being conducted simul-
taneously to examine components of the package fertilizer response, insecticide
response, variety comparison, etc.)
In Table 10 we show the calculations of variable costs for the two treatments
(following the format of Table 8). Agricultural labor in the area is commonly
hired byr the day for $3 per day, and the rental of hand sprayers is common at the
rate of $4 per day. Labor estimates for applying fertilizer, insecticides and herbi-
cides and for hand weeding were based on discussions with farmers. Fertilizer and
chemicals are available from both private and government-sponsored stores at the
prices indicated (including a delivery charge for fertilizer).

Field cost ($/ha)
Money Opportunity
Operation Input Amount Price Cost Price Cost Total
Casrrent technology package
fertilization 46-0-0 65 kg 0.54 35.10 35.10
20-20-0 100 kg 0.54 54.00 54.00
labor 6 days -_ 3.00 18.00 18.00
89.10 18.00 107.10

weed control weeding 10 days 3.00 30.00 30,00

insect control insecticide '24 kg 1.60 38.00 38.00
(2 applications) sprayer 2 days 4.00 8.00 8.00
labor 2 days 3.00 6.00 6.00
46.00 6.00 52.00

Intensive technology package
fertilization 46-)-0 130 k~g 0.54 70.20 70.20
20-20-0 200 kg 0.54 108.00 -- 108.00
labor 9 days - 3.00 27.00 27.00
178.20 27.00 205.20

weed control herbicide 2 kg 17.00 34.00 34.00
sprayer 3 days 4.00 12.00 12.00
labor 3 days 3.00 9.00 9.00
46.00 9.00 55.00

insect control insecticide 36 kg 1.60 58.00 58.00
(3 applications) sprayer 3 days 4.00 12.00 12.00
labor 3 days - 3.00 9.00 9.00
70,00 9.00 79.00

9/ Two examples

Maize is marketed by most farmers through independent truckers who regu-
larly pass through the villages offering to buy grain. It is not usually stored before
being sold. Prices the last two seasons have been about $250 per ton. The official
government price is higher, but since the government buys only a limited amount,
and since farmers must accept quality discounts and pay delivery costs, the price
from truckers is taken to be relevant to most farmers. The field price per ton was
determined as:

Price offered by truckers

field price



The cost of picking was determined by dividing the number of man-days per hec-
tare by the average yield in the area, multiplied by the going wage. The costs of
husking and shelling were estimated from contract rates reported by various far-
mers in the area.
The partial budget of these trials can now be completed as shown in Table 11
(which follows the format of Table 3). Notice that the average yields as measured
have been reduced by 10%6 to account for losses from shelling and other factors
not reflected in the experimental harvest and measurement procedures. The mar-
ginal cost of the intensive technological package is $150 per hectare, and the rate
of return to this investment is 77%6 (last line, Table I1 ).

Variable cost calculations
(per hectare).

for the current and intensive technologies


From agronomic TABLE 11
data to farmer Partial budget of the maize technology trials.
Current Intensive
Item technology technology
Average yield, (ton/ha) 2.78 4.04
Adjustment for harvest loss (10%) X 9 X .9
Net yield (ton/ha) 2.50 3.64
Gross field benefit ($/ha at $232.50/Ton) 581 846
Variable costs (from Table 10):
fertilization ($/ha) 107 205
weed control ($/ha) 30 55
insect control ($/ha) 52 79
Total variable cost ($/ha) 189 339
Net benefit ($/ha) 392 507
Rate of return = (507 392)/(339 189) = 115/150 = 0.77 = 77%

A rate of return of 77%0 is adequate to warrant its recommendation to farmers
unless the risks are unusually high. (Similar trials elsewhere in the same country
showed rates of return too small to be recommended). The risk analysis procedure
suggested in Chapter 5 is to list the net benefits from each treatment at each loca-
tion. In this case we suggest a modification of this procedure which is quite useful
where only two treatments are included.
In Table 12 we present the yields for each treatment at each of the 26 sites, as
well as the yield gain offered by the intensive technology as compared with cur-
rent technology. Notice that four of the 26 trials were lost due to drought. Since
drought is a hazard which farmers must consider, we must include these results in
the analysis.
What size yield gain is required in order to pay for the cost of the extra in-
puts? This is determined as follows:

marginal cost in money $150
marginal cost in grain ($150/$230.50) 0.65 ton
marginal yield required to give 40 %/
rate of return (0.65 x 1.40) 0.90 ton

We now know that if a farmer is to receive a 40%orate of return to his investment,
the intensive technology package must yield 0.9 ton/ha more than the current
We see in Table 12 that this occurs in 14 of the 26 trials. Furthermore, we see
that in three more cases, the farmer would have got his investment back, but at a
lower rate of return than 40%b. This leaves nine trials of the 26 in which the gain
in yield was not enough to pay for the additional inputs.
The above approach offers a convenient way to present the range of results
from a two-treatment experiment. But it does not directly help us to assess the
down-side risks of the two alternatives. To do this, we again turn to a minimum
returns analysis as described in Chapter 5. First, look down the column of yields
for the current technology. The worst eight of these (the worst 25%/, including
the four drought trials and the four yields in italics) average 0.36 tons. To get net

Minimum net benefits (ton/ha) from
the 26 trials.

Yield data from a set of wheat variety
trials (ton/ha).

Vo ale Variety V1 Variety V2
Trial 04) 60-20 0-0 60-20 0-0 60-20

"Yield gain sufficient for rate of return of
40% or more. *Yield gain sufficient for rate
of return between 0% and 40%.

9/ Two examples

benefits, expressed as grain, we can subtract from these figures 0.81 tons (variable
cost of $189 divided by $232.50 per ton of grain), giving the results in Table 13.
Similarly, subtract 1.46 tons from the worst yields in the complete technology
column ($339 divided by $232.50).
It is clear from Table 13 that the complete technology is more risky only when
there is a complete crop loss. In these cases (15%~ of the trials in this year) the
farmer would lose the equivalent of 0.65 tons of grain more with the intensive
technology than with the current technology. This is a serious risk, and will dis
courage many subsistence farmers from applying the extra inputs, even though
the average return on the investment is 77%/. It is probably not wise to recom-
mend this investment to farmers close to the subsistence level, because of the
probability that a crop failure in the first or second year would put the family in a
very bad debt situation. In the recommendation domain of this study, however,
the number of subsistence farmers is small, and the intensive technology is an
appropriate recommendation despite the risks.

Wheat variety trials
The set of wheat variety trials which will be analyzed were conducted at six dif-
ferent sites in a rainfed area. Varieties were tested under conditions of no ferti-

lizer and 60 kg of N plus 20 kg of P205 per hectare. The results (replication
means, again) are presented in Table 14. Figure 3 also helps show the relationship

Yields (ton/ha) from 26 trials with
two levels of maize technology.

Intensive Current Gain
6.98 5.17 1.81 "
6.24 6.34 -0.10
5.49 3.25 2.24 "
5.84 4.97 0.87 *
5.26 4.04 1.22 "
3.00 3.01 -0. 10
6.07 2.51 3.56 "
7.81 7.11 0.70 *
5.25 3.14 2.11 "
6.10 1.15 -0.05
3.04 0.21 2.83 "'
4.86 1.36 3.50 "
3.33 0.39 2.94 ",
2.06 1.01 1.05 "+
4.63 1.47 3.16 "
3.43 3.81 -0.38
3.71 2.99 0.72 *
3.41 1.24 2.17 "
5.43 3.76 1.67 "
3.67 2.64 1.03 "

02 405 0.2
0 0 0
0 0 0
0 0 0
4.04 2.78 1.26 "

en logy



Tete oogy



23 (drought)
24 (drought)
25 (drought)
26 (drought)

Net benefit
Second worst
Third worst
Fourth worst
Fifth worst
Sixth worst
Seventh worst
Eighth worst
Average of
eight trials

1 0.84 1.67
2 0.72 1.50
3 1.23 1.38
4 1.22 1.51
5 1.36 1.30
6 1.58 1.99
Avg. 1.16 1.56

1.08 2.25 1.46 2.58
0.98 2.00 0.76 1.94
1.68 2.33 0.95 2.27
1.34 2.31 1.67 2.58
1.10 2.24 1.40 2.68
1.53 2.01 1.74 2.97
t.28 2.19 1.33 2.50

FIG. 3. Wheat variety yields and fertilizer levels.

Partial budget of wheat variety trials (per hectare).


local Variety V1 Variety V2
N: 0 60 0 60 0 60
Item P205: 0 20 0 20 0 20
(1) Average yield (ton/ha) 1.16 1.56 1.28 2.19 1.33 2.50
(2) Net yield (ton/ha) 1.02 1.37 1.13 1.92 1,17 2.20
(3) Field price ($/ton) 1000 1000 900 900 1000 1000
(4) Gross field value ($/ha) 1020 1370 1017 1728 1170 2200
Variable money costs
(5) Seed (75 kg at $1/kg) 75 75 75 75
(6) Fertilizer ($ 5/unit) 400 400 400
(7) Variable money costs ($/ha) 0 400 75 475 75 475
Variable opportunity costs:
(8) Days per application 2 2 2
(9) Cost of application ($50/day)- 100 100 100

(10) Total variable costs ($/ha) 0 500 75 575 75 575
(11) Net benefit ($/ha) 1020 870 942 1153 1095 1625

2.5 -



D 1.5



p local

1.15 2.0
Average yield of all varieties, tons/ha

TABLE 16 9/ Two examples
Marginal analysis of the wheat data (per hectare).

change from next highest
treatment variable increase in increase rate of
Net benefit variety N P205 cost variable cost net benefit return
$1625 V2 60 20 $575 $500 $530 106%
1153 V1 60 20 575 - -
1095 V2 0 0 75 75 75 100%
1020 local 0 0 -

between variety yields and fertilizer levels. For each fertilizer level the average
yield of each variety is plotted against the average yield of all varieties. Such a
diagram is useful in visualizing variety-environment interactions. In this case there
are only two environments represented, the fertilized environment and the non-
fertilized environment.
It is clear that V2 is the variety most responsive to fertilizer, VI is about aver-
age, and the local variety is least responsive.
Only an economic analysis such as the one which follows can indicate the im-
plications of these data for farmer recommendations.
The first step is to adjust average yields for harvest and storage losses, which
we estimate at 12%b. After appropriate inquiries among farmers and merchants in
the area, we determine that the field price of the local variety is $1000 per ton.
Variety VI is a new variety which had been previously introduced, and merchants
state that local people will not buy it because of its color, though it can be ship-
ped out of the area and sold. Because of this, the price of VI has been 8 to 10
lower than the local variety, which means a field price of $900 per ton. The other
new variety, V2, has not previously been released, but the grain is virtually indis-
tinguishable from the local variety, so the field price is assumed to be the same,
$ 1000 per ton. The gross field values, based on the average yields for each variety,
are shown in line 4 of Table 15.
Most farmers would have to purchase seed of the two new varieties at a field
price of $2 kg, so at a seeding rate of 75 kg/ha the new varieties would require a
cash outlay of $150/ha. Seed for the local variety costs only $1.00 kg, so the
increase in seed costs above the local variety is $1.00/kg. The field price of both N
and P205 were determined to be $5/kg of nutrient, and it was estimated that 2
man-days were required to apply fertilizer to I hectare, at a field price of $50 per
man-day. The resulting estimates of total variable cost by treatment are shown in
lines 7 and 10 of Table 15.
Finally, in line 1 1 of Table 15, we present the resulting net benefit for each of
the alternatives. Variety V2, when fertilized, offers clearly the highest average net
benefit, but again due to considerations of capital scarcity and risk aversion, we
will need to examine these results using the procedures described earlier to be sure
which alternatives to recommend.
The first task is a marginal analysis of the partial budget results. To this end,
we rank the alternatives by net benefit gs shown in Table 16, omitting those treat-
ments which gave net benefits smaller than the check plot (the local variety with-
out fertilizer).
The treatment VI with fertilizer is dominated by V2 with fertilizer, since both
have the same variable cost, but the latter has a higher net benefit. Only V2 with-

out fertilizer and V2 with fertilizer remain as reasonable alternatives by this cri-
terion. The smallest investment alternative available to the farmer is to spend
$75/ha on the seed for the new variety. In exchange for this he can expect to
receive a net return of $75/ha (the first year), for a rate of return of 100%o. This is
an adequate rate to warrant farmers' investment, and morevoer, the farmer can
expect to receive additional benefits in the future without the necessity of again
investing in seed. The actual rate of return is then underestimated by this figure.
However, the absolute amount of increase in net benefits ($75/ha) is quite small,
being only 7 %l or so higher than net returns attainable from the local variety.
Thus farmers might not be too enthusiastic about making this change, even
though the rate of return on the investment in seed is quite high.
What about the alternative of investing an additional $500/ha to apply ferti-
lizer to variety V2? The expected increase in net benefits is $575/ha, for a rate of
return of 106%/. This rate is acceptable (if the risks are not unusually great), and
the size of the increase in net returns is quite significant, being about 50%q of the
net benefits from the local variety without fertilizer. Thus we can assume that the
recommendation of variety V2 with fertilizer would be consistent with farmers'
As a check on the riskiness of these alternatives compared with the others, we
need to examine the returns in the worst of the six outcomes and the worst two
of the six. These are presented in Table 17 (along with the average of all six for
reference). Here we find, as we did in the case of the fertilizer trials, that the
treatment chosen by marginal analysis of the average yields is also the treatment
with the highest minimum returns. The minimum return analysis again supports
the recommendation.
However, the minimum returns analysis does reveal something which was not
evident from the marginal analysis of the average yield data. Suppose that there
are some farmers in the area who are unable or unwilling to invest in fertilizer.
The marginal analysis suggests that for these farmers, V2 would be a good recom-
mendation, or perhaps they could as well stay with their old variety. However,
Table 17 shows that these two alternatives have minimum returns much lower
than the alternative VI. Then very risk-averse farmers who cannot or will not
apply fertilizer might prefer VI, even though on the average they would receive
lower returns than from either of the other varieties when grown without ferti-
lizer. The variety VI appears to be more stable across environments than the
other two. Furthermore, it outyields the local variety on the average under unfer-
tilized conditions, and yields almost as much as V2 without fertilizer. The reason
that it shows up as a relatively unattractive alternative in the marginal analysis is
because of the price discount. It is therefore appropriate to use sensitivity analysis
to examine the implications of possible changes in this discount.
We can ask ourselves at what price discount VI without fertilizer would pro-
vide higher net benefits than the local variety without fertilizer. The answer is a
discount of about 3%/. With such a discount (a price of $970), the net benefits
would be $1020/ha, the same as for the local variety. In order to provide a rate of
return equal on the average to that from investing in V2 seed, variety VI would
have to have an even higher price than the local variety.
What then are we to conclude? First, the recommendation of V2 with ferti-

From agronomic
data to farmer

TABLE 17 9/ Two examples
Minimum returns analysis of the wheat data
(per hectare).

average of average of
Treatment worst net worst two all six net
variety N P205 return net returns returns
local 0 0 $ 633 $ 686 $1 020
local 60 20 640 677 870
V1 0 0 701 740 942
V1 60 20 1008 1012 1153
V2 0 0 594 678 1095
V2 60 20 1131 1276 1625

lizer is a good one, as it is verified by both marginal analysis and minimum returns
analysis. If there are some farmers who will not be applying fertilizer regardless of
the recommendation, one might want to recommend that they plant VI because
of its high minimum net returns. The judgement must be made on the agrono-
mist's judgement of the size of this group of farmers and the seriousness with
which they might view the differences in minimum net returns between the lo-
cal variety and V1.
Since six observations are not many, the agronomist might be wise to wait
another season before making any recommendation with respect to VI versus the
old variety under unfertilized conditions. Additional observations may show that
the difference in minimum returns is not so great as is estimated here on the basis
of just six outcomes.


Cost of capital-benefits given up by the farmer due to having investment capital
tied up in an enterprise for a period of time.
Dominance-one alternative is said to dominate another when the first has higher
net benefits and equal or lower variable cost than the second.
Field cost(of an input)-the field price of an input multiplied by the quantity of
the input which varies with the decision.
Field price (of an input)-the total value which must be given up to bring an addi-
tional unit of input into the field.
Field price (of output)-the value to the farmer of an additional unit of produc-
tion in the field, prior to harvest.
Gross field benefit-net yield times field price for all products from the crop.
Investment capital value of inputs (purchased or owned) which are allocated to
an enterprise with the expectation of a return at a later point in time.
Marginal cost-the increase in variable cost which occurs in changing from one pro-
duction alternative to another.
Marginal net benefit-the increase in net benefit which can be obtained by chang-
ing from one production alternative to another.
Marginal rate of return- the marginal net benefit divided by the marginal cost.
(Calculated for non-dominated alternatives only).
Money field price (of an input)-refers to purchase price plus other direct expenses
such as transportation costs.
Money field price (of output)-the market price of a unit of product minus har-
vest, storage, transportation and marketing costs, and quality discounts.
Minimu m returns analysis-a process on each production alternative which features
examining net returns to the individual treatments and selecting that alter-
native whose lowest return or whose lowest average return is highest among
the alternatives being considered.
Net benefits-the value of the benefits less the value of the value of the things
given up in achieving the benefits, total gross field benefit minus total variable
Net yield-the measured yield per hectare in the field, minus harvest losses and
storage losses where appropriate.
Opportunity cost-the value of any resource in its best alternative use.
Opportunity field price (of an input)-refers to the non-money value of the input
in its best alternative use.
Opportunity field price (of output)-the money price which the farm family
would have to pay to acquire an additional unit of the product for consumption.
Proportional cost costs which vary directly and proportionally with yield.
Recommendation domain-a group of farmers within an agro-climatic zone whose
farms are sufficiently similar and who follow sufficiently similar practices that
a given recommendation is applicable to the entire group.
Risk premium-an amount, given as a percentage, which a farmer requires before
exposing himself to a variable income.

Sensitivity analysis-a process which features changing a price or a cost within rea- Glossary
sonable bounds of the original estimate to determine if the original ranking of
alternatives is affected.
Total field cost/variable cost--the sum of field costs for all inputs which are affect-
ed by the choice.

University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - Version 2.9.7 - mvs