Cover letter
 Front Cover
 Title Page
 Table of Contents
 Part one: The partial budget
 Part two: Marginal analysis
 Part three: Variability
 Part four: Summary

Title: Second draft of : From agronomic data to farmer recommendations
Full Citation
Permanent Link: http://ufdc.ufl.edu/UF00080820/00001
 Material Information
Title: Second draft of : From agronomic data to farmer recommendations
Physical Description: Book
Language: English
Creator: CIMMYT Economics Program
Tripp, Robert
Publisher: CIMMYT Economics Program
Place of Publication: Gainesville, Fla.
Publication Date: 1986
 Record Information
Bibliographic ID: UF00080820
Volume ID: VID00001
Source Institution: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: oclc - 187981166

Table of Contents
    Cover letter
        Cover letter
    Front Cover
        Front Cover
    Title Page
        Title Page
    Table of Contents
        Table of Contents 1
        Table of Contents 2
        Page 1
        Page 2
        Page 3
        Page 4
        Page 5
        Page 6
        Page 7
        Page 8
        Page 9
        Page 10
        Page 11
        Page 12
        Page 13
        Page 14
        Page 15
        Page 16
        Page 17
        Page 18
    Part one: The partial budget
        Page 19
        Page 20
        Page 21
        Page 22
        Page 23
        Page 24
        Page 25
        Page 26
        Page 27
        Page 28
        Page 29
        Page 30
        Page 31
        Page 32
        Page 33
        Page 34
        Page 35
        Page 36
        Page 37
        Page 38
        Page 39
        Page 40
        Page 41
        Page 42
        Page 43
    Part two: Marginal analysis
        Page 44
        Page 45
        Page 46
        Page 47
        Page 48
        Page 49
        Page 50
        Page 51
        Page 52
        Page 53
        Page 54
        Page 55
        Page 56
        Page 57
        Page 58
        Page 59
        Page 60
        Page 61
        Page 62
        Page 63
        Page 64
        Page 65
        Page 66
        Page 67
        Page 68
        Page 69
        Page 70
        Page 71
        Page 72
        Page 73
        Page 74
        Page 75
        Page 76
        Page 77
        Page 78
        Page 79
        Page 80
        Page 81
        Page 82
        Page 83
    Part three: Variability
        Page 84
        Page 85
        Page 86
        Page 87
        Page 88
        Page 89
        Page 90
        Page 91
        Page 92
        Page 93
        Page 94
        Page 95
        Page 96
        Page 97
        Page 98
        Page 99
        Page 100
        Page 101
        Page 102
        Page 103
        Page 104
        Page 105
        Page 106
        Page 107
        Page 108
        Page 109
        Page 110
        Page 111
        Page 112
        Page 113
        Page 114
        Page 115
        Page 116
        Page 117
        Page 118
        Page 119
        Page 120
        Page 121
    Part four: Summary
        Page 122
        Page 123
        Page 124
        Page 125
        Page 126
        Page 127
Full Text

(". C Centro Inttmacional de
SMejoramiento d Maiz y Trig
International Maize and
W 'hat Improvement Center
Li a27 Apdo. Postal 6-641, Cdl Jwm,
Deg Cuaubter 06&oo Mixico, DE, Mexico
elex: 1 772023-CIMTME Cable: CENCIMMYT Tel: Texcoco (595) 421-00/420-11

October 29, 1986

Dr. Peter Hildebrand
Food and Resource Economics Dept.
University of Florida
Gainesville, Florida 32611

Dear Peter:

I enclose a draft of a revised version of the CIMMYT
manual on the economic analysis of on-farm experiments.

It maintains the basic concpets of the previous version
(Perrin et.al.), but tries to take advantage of ten years'
experience in teaching these ideas. Thus the ordering, the
presentation, and some of the vocabulary are a bit
different. The idea, as previously, is to present what any
beginner needs to know about the economic analysis of
on-farm experiments. We have resisted the temptation to go
beyond the basics in this publication. We look upon this as
the bare minimum, and will look to other publication for
exploring more advanced topics. The new manual will be
accompanied by a new version of our workbook of exercises.

I would very much value your comments on this new
version. Any ideas, suggestions or criticisms will receive
full attention, and any comments received before the end of
January 1987 can be accommodated in the final version. I
look forward to hearing from you.

Sincerely yours,

Robert Tripp
Economics Program





CIMMYT Economics Program

October 1986




CIMMYT Economics Program

October 1986

For review purposes only. Not for citation or distribution.


On-Farm Research .. . . . 1
Goals of the Farmer . . . . 5
On-Farm Experiments . .. . .... 7
Experimental Locations and Recommendation Domains . 10
Introduction to Basic Concepts .. . . 11
The Partial Budget .. . . 12
Marginal Analysis ... . . . . 16
Variability .. . . . .. 17


Identifying Variable Inputs . . . 21
Purchased Inputs ........... . 22
Equipment and Machinery . . . . 24
Labor . . . . .. . .. 25
Total Costs That Vary .. . ... 27

PARTIAL BUDGET . . . . . . 31
Pooling the Results From the Same Recommendation Domain 32
Statistical Analysis . ... . . 33
Average Yield . .. .. . .. 34
Adjusted Yield . . . . 34
Field Price of the Crop . . . . 38
Net Benefits .... . . . . 40
Including All Gross Benefits in the Partial Budget . 40


RETURN .. .. .. . . .. .. 45
Dominance Analysis . . .. . . 45
Net Benefit Curves . . . 47
Marginal Rate of Return . ... . .. .. 47

A First Approximation of the Minimum Rate of Return 52
The Informal Capital Market . . . . 53
The Formal Capital Market . . . 54
Summary . . . . 55

An Introductory Example ..... .. 56
A Fertilizer Experiment . . . . 57
Analysis Using Residuals . . . 64
A Second Example: An Insect Control Experiment . 67
Some Questions About Marginal Analysis . . 74


Reviewing the Purpose of the Experiment . .. 85
Tentative Recommendation Domains . ... 87
Reviewing Experimental Results . ..... . 89
Recommendation Domains Defined by Socioeconomic Criteria 92
Statistical Analysis . . . . 93

Dealing with Risk in On-Farm Research ... .. 98
Risk and On-Farm Data. . .... . .... 100
Prerequisites for a Minimum Returns.Analysis . 100
The Farmers' Point of View . .. .. . 102
Minimum Returns Analysis . .. .. 106

Which Costs and Prices Should be Used in the Partial
Budget? . . . . . . 113
Sensitivity Analysis . ........... 115




This manual presents a set of procedures for carrying out'the economic
analysis of on-farm experiments. It is intended for use by agricultural
scientists as they develop recommendations for farmers from agronomic data.
Development of recommendations which fit farmers' goals and situations is
not necessarily difficult, but it 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.

A recommendation is information that farmers can use to improve the
productivity of their resources. A good recommendation can be defined as a
choice which farmers would make, given their current resources, if they had

all the information available to the agronomist. It may be very specific

information, as in the case of a particular variety. Or it may be more
general information, as in the case of a fertilizer level or storage
technique, that farmers will probably adjust to their own conditions and
needs. The agronomic data upon which the recommendations are based must be
relevant to the farmers' own agro-ecological conditions. The quality and
quantity of these data must be sufficient to inspire confidence that the
results will be repeated, and the evaluation of these data must be

consistent with the farmers' goals and socioeconomic circumstances.

On-Farm Research

The steps of an on-farm research program are shown in Figure 1.1. The

first step is diagnosis. If recommendations.are to be oriented to farmers,
research should begin with an understanding of farmers' conditions. This

National goals,
input supply,
credit, markets,

Choice of target
farmers and
of policy issues

Figure 1.1
Stages of On-Farm Research

On-Farm Research'

1. Diagnosis
Review of secondary data,
informal and formal

2. Planning
Selection of priorities
for research and design
of on-farm exDerirents.

3. Experimentation
Conduct experiments in
farmers' fields to
formulate improved --
technologies under
farmers' conditions.

-Agronomic evaluation"
-Statistical analysis
-Farmer assessment
-Economic analysis

New components
for on-farm

of problems
for station

Experiment Station
Develops and screens
new technological

5. Recomrmendation
Demonstrate improved
technologies to farmers.

iu^ M-mi mj~BMa CiM.Mj B

requires some diagnostic work-in the field, including observations of

farmers' fields and farm surveys. This diagnosis is used to help identify

major factors that limit farm productivity and to help specify possible


The information from the diagnosis is used in planning an experimental

research program which includes experiments in farmers' fields. After the

first year, the experimental results form an important part of the

information used for planning subsequent cycles. In addition, other

diagnostic work continues during the management of the experimental program,

as researchers continue to seek information about farmers' conditions and

problems which will be useful in the planning of future experiments.

The on-farm experiments are planted on the fields of representative

farmers. Some of these experiments may explore the problems identified

during diagnosis, while other experiments will test possible solutions to

those problems.

The results of the on-farm experiments must then be assessed. There

are several elements in such an assessment. First, the experimental results

must be subjected to an agronomic interpretation in order to decide if the

responses of the various treatments can be understood agronomically.

Second,.a statistical analysis of the results helps researchers decide what

degree of confidence they can place in the observed responses. In addition,

researchers must discuss the results with farmers, in order to get their

opinions of the treatments they have seen in their fields. This farmer

assessment is very important. Finally, an economic analysis of the results

is essential. The economic analysis helps researchers decide which

treatments merit further investigation, and which recommendations can be
made to farmers. The procedures for carrying out such an economic analysis

are the subject of this manual.

The results of an assessment of on-farm experiments can be used for

several purposes. First, they may be used for helping plan further

research. Some experiments will have as their goal the clarification of

production problems (e.g. Is production limited by the availability of

potassium? Will improved weed control give a significant increase in

yields?) The answers to these questions provide researchers with

information for further work. As Figure 1.1 shows, this information can be

fed back to the planning of subsequent experiments. It is also an example

of the type of on-farm research information that may help orient work on

the experimental station.

Second, the results may be used for making recommendations to farmers.

Some experiments will compare possible improvements to current farmer

practice (e.g. Which level of potassium should be applied? Which weed

control method gives the best results?) The answers to these questions

provide information for farmers.

Finally, the results of on-farm experiments may be used to provide

information to policy makers regarding current policy towards such matters

as input supply or credit regulations. Experimental results can be used to

help analyse policy implementation (e.g. Given a significant response to

potassium, is the appropriate fertilizer available? Do local credit

programs allow farmers to take advantage of new weed control methods?)

Although the emphasis in this manual will be on the economic analysis of

on-farm experiments for guiding further research and making recommendations

to farmers, mention will be made at several points of this link between

on-farm research and policy implementation.

Goals of the Farmer

In order to make recommendations that farmers will use, you must be

aware of the human element in farming, as well as the biological element.

You must think in terms of farmers' goals and the constraints on

achieving those goals.

In the first place, farmers are concerned with assuring an adequate

food supply for their families. They may do this by producing much of what

their family consumes, or by marketing a certain proportion of their output

and using the cash to obtain their food supply. Farm enterprises also

provide other necessities for the farm family, either directly or through
cash earning. In addition, the farm family is usually part of a wider

community, towards which the family may have certain obligations. In order

to meet all of these requirements, farmers usually manage a very complex

system of enterprises that may include various crops, animals, and off-farm

work. Although the procedures of this manual concentrate on the evaluation

of improvements in particular crop enterprises, it is essential that these

new practices be compatible with the larger farming system.

Second, no matter whether farmers market little or most of their

produce, they are interested in economic return. Farmers will consider the

costs of changing from one practice to another and the economic value of

the results of that change. Farmers will recognize that if they eliminate

weeds from their fields they will likely benefit by harvesting more grain.

On the other hand, they will realize that they must give up some cash to

buy herbicides and then give up some time and effort to apply them. Or,

alternatively, they must give up alot of time and effort for hand weeding.

Farmers 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. What farmers are doing in this case is assessing the

difference in net benefits between practices the value of the benefits

gained minus the value of the things given up.

SAs farmers attempt to evaluate the net benefits of different
treatments, they know there is some risk involved. In the weed control

example, the farmers know that in the case of drought or early frost, they

may get no grain, regardless of the type of weed control. 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 may on

the average yield them positive benefits. The fact that farmers will often

prefer yield stability to high average yields is referred to as risk


Another factor, related to risk aversion, is the fact that farmers

tend to change their practices in a gradual, stepwise manner. Farmers

compare their practices with alternatives, and seek ways of cautiously

testing out new technologies. It is thus more likely that farmers will

adopt individual elements, or small combinations of elements, rather than a

complete technological package. This is not to say that farmers will not

eventually come to use all the elements of a package of practices, but

simply that in offering recommendations to farmers it is best to think of a

strategy which allows them to make changes a step at a time.

On-Farm Experiments

What are the characteristics of agronomic .experiments that will allow

an appraisal of alternative technologies in a way that parallels farmer

decision making? The following are five characteristics of on-farm

experiments which must be fulfilled if the procedures described in this

manual are to be useful.

(1) In the first place, the experiments must address problems that are

important for the farmers. It may be that farmers themselves are not

initially aware of a particular problem (e.g. a nutrient deficiency

or a disease), but if research does not lead to possibilities for

significantly improving farm productivity, it will not attract the

interest of farmers, nor merit assessment. Thus the experiments demand

a good understanding of farmers' agronomic and socioeconomic


(2) The experiments should examine a relatively few factors at a time. An

on-farm experiment with more than, say, four variables will be

difficult to manage and of little relevance to farmers' stepwise

adoption behavior.

(3) If researchers are to compare the farmers' practice with various

alternatives in order to make a recommendation, then the farmers'

practice should be included as one of the treatments in the

experiment. The farmers will want to see this comparison in any case.

(4) The non-experimental variables should be those of representative

farmer practice. It is sometimes tempting to use higher levels of

management for the non-experimental variables of an on-farm

experiment, in order to increase the chances of observable responses

to the experimental variables. For example, in a variety experiment

researchers may want to use fertilizer levels higher than those used

by farmers, in order to ensure vigorous plant growth and the

observation of clear differences among the varieties. This type of

*experiment may certainly be justified in order to get more information
about the performance of the varieties, but the results cannot be used

to recommend a particular variety for use under current farmer

fertilization practices.

Another example may be useful. Assume that researchers wish to plant a

fertilizer experiment in an area where insects cause crop losses but

farmers do not control insects. There are four possibilities:

a) Plant the fertilizer experiment with good insect control. The
experiment will give interesting information on fertilizer

response, but will not provide a fertilizer recommendation for

farmers who do not use insect control. An analysis of this

experiment with the procedures in this manual will give a

misleading result.

b) Plant the fertilizer experiment with the farmers' level of insect

control. The results can be analysed in order to decide what

level of fertilizer is appropriate, given farmers' current insect

control practices.

c) If insects are more serious problem, it may be better to

experiment first with insect control methods, before

experimenting with fertilizer. The diagnosis and planning steps

of on-farm research are meant to help set these priorities. The

methods of this manual could be used to help identify an

appropriate insect control method for recommendation to

d) If insects and fertility are both serious problems, then an

experiment can be designed which takes both insect control and

fertilizer as experimental variables. As long as one treatment

represents the farmers' practice with respect to insect control
and fertility, the experiment can be analyzed with the procedures

of this manual.

(5) Finally, not only must the management of non-experimental variables be

representative of farmers' practice, but the experiments must be

planted at locations that are representative of farmers' conditions.

_/ Once this has been done, and there is evidence that farmers will adopt
the new insect control method, it could be used as a non-experimental
variable in the fertilizer experiments, as long as it is understood
that the fertilizer recommendation to be developed from such trials
depends on the farmers first adopting the insect control method.

If most of the farms are on steep slopes, the results of experiments

planted on an alluvial plain will probably not be relevant. Similarly,

if most farmers plant one crop in rotation with another, experiments

from fields that have been under fallow may provide little useful

information. More will be said in the following section about the

selection of locations.

Experimental Locations and Recommendation Domains

The development of recommendations for farmers must be as efficient as
possible. The conditions under which farmers live and work are diverse in

.almost every respect imaginable. They have different amounts and kinds of

land, different degrees of wealth, different attitudes toward risk,

different access to labor, different marketing opportunities, and so on.

Many of these differences can influence the farmers' response to

recommendations. But it is impossible to make a separate recommendation for

each farmer.

As a practical matter, one must compromise by identifying groups of
farmers that have similar circumstances, and for whom it is likely that the

same recommendation will be suitable. Such a group of farmers is called a

recommendation domain. Recommendation domains may be defined both by

agroclimatic and by socioeconomic circumstances. A recommendation domain is

specific to a particular recommendation. For example, a new variety may be

appropriate for all farmers in a given geographical area, while a
particular fertilizer recommendation may be appropriate only for farmers

with a certain type of soil or rotation pattern. Thus the recommendation

domain for variety would be different from the recommendation domain for


Recommendation domains are identified, defined and redefined
throughout the process of on-farm research. They may be tentatively

described during the first diagnosis. Experimentation adds precision to the

definition of domains. The final definition may not be ready until the

recommendation is ready to be passed to farmers.

In interpreting agronomic data to make recommendations you must have a

fairly clear idea of the group of farmers which will be able to use this

information. This must include not only the agroclimatic range over which

the results will be relevant, but also whether such factors as different
management practices or access to resources will be important in causing

some farmers to interpret the results differently from others.

For the purposes of this manual, it is important that the on-farm

experiments must be planted at locations that are representative of the

recommendation domain. The economic analysis is done on the pooled data

from a group of locations of the same domain. The economic analysis of the

results of a single location is not very useful because, first, you are not

making recommendations for individual farmers, and second, one location

will rarely provide sufficient agronomic data to be extrapolated to a group

of farmers. Thus all of the examples in this manual will represent data

from groups of locations from one recommendation domain.

Introduction to Basic Concepts

In order to make good recommendations for farmers, researchers must be

able to evaluate alternative technologies from the farmer's point of view.

The premises of this manual are: (1) farmers are concerned with the net

benefits of particular technologies; (2) they usually adopt innovations in

a stepwise fashion; and (3) they are concerned with the risks involved in

adopting new practices. These are the concerns that will be treated in

subsequent chapters of the manual.-Part One of the manual describes the

construction of a partial budget, which is used to calculate net benefits.

Part Two of the manual presents the techniques of marginal analysis. This

is a way of evaluating the changes from one technology to another, by

comparing the changes in costs and net benefits associated with each
treatment. Part Three of the manual describes ways of dealing with the

variability that is characteristic of farmers' environments. Variability in

results from location to location and from year to year, and in the costs

of the inputs and prices of crops, are all issues of concern to farmers as
they decide what choices to make. Part Four provides a summary of these


The following sections provide a brief overview of these topics.

The Partial Budget

Partial budgeting is a method of organizing experimental data and
information about the costs and benefits of various treatments. As an

example, consider the farmers who are trying to decide between their
current practice of handweeding and the alternative of herbicide. Suppose

that some experiments have been planted on farmers' fields, and the

results show that the current farmer practice of handweeding results in

average yields of 2,000 kg/ha, while the use of herbicides gives an average

yield of 2,400 kg/ha.

Table 1.1 shows a partial budget for this weed control experiment.
There are two columns, representing the two treatments of hand weeding and

herbicide. The first line of the budget presents the average yield from all

locations in the recommendation domain for each of the two treatments. The,
Second line is the adjusted yield. Although the experiments were planted on

representative farmers' fields, researchers have judged that their yields

exceed by about 10% those that could be achieved by farmers using the same

technologies. They have therefore adjusted the yields downwards by 10%.

(Yield adjustment will be discussed in Chapter Three).

Average Yield (kg/ha)

Adjusted Yield (kg/ha)

Gross Field Benefit ($/ha)

Table 1.1
ample of a Partial Budget

Hand Weeding





Cost of herbicide ($/ha)
Cost of labor to apply herbicide ($/ha)

Cost of labor for hand weeding ($/ha)

Total Costs that Vary ($/ha)

Net Benefits ($/ha)








The next line is the gross field benefit, which values the adjusted

yield for each treatment. To calculate the gross field benefit you need to

know the field price of the crop. The field price is the value of.a kilo of

the crop to the farmer, net of harvest costs. In this example the field

price is $2/kg2/ (e.g. 1,800 kg/ha x $2/kg = $3,600/ha).

The farmers can now compare the gross benefits of each treatment, but

they will want to take account of the different costs as well. In

considering the costs associated with each treatment, the farmers need only

be concerned by those costs that differ across the treatments, or the costs

that vary. Costs which do not differ across treatments (such as plowing and

planting costs) will be incurred regardless of which treatment is used.

They cannot affect the choice and can be ignored for the purpose of this

decision. The term "partial budget" is a reminder that not all production
costs are included in the budget only those which are affected by the

decision being considered.

In this case, the costs that vary are those associated with weed

control. Table 1.2 shows how to calculate these costs. Note that they are
all calculated on a per hectare basis. The total costs that vary for each

treatment is the sum of the individual costs that vary. In this example,

the total costs that vary for the present practice of hand weeding is $400

per hectare, while the total costs that vary for the herbicide alternative

is $600 per hectare.

SThe use of the $ sign in this manual is not intended-to represent any
particular currency.

The final line of the partial budget shows the net benefits. This is

calculated by subtracting the total costs that vary from the gross field

Table 1.2

Calculation of the Costs that Vary

Price of herbicide

Amount used

Cost of herbicide

Price of labor

Labor to apply herbicide

Cost of labor to apply


Price of labor
Labor for hand weeding

Cost of labor for hand


= $250/liter

S 2 liters/ha

= $500/ha

= $50/day

= 2 days/ha

= $100/ha

= $50/day

= 8 days/ha

= $400/ha

benefits. In the weed control example, the net benefits from the use of

herbicide are $3,720 per hectare, while those for the Current practice are

$3,200 per hectare. Remember that net benefits are not the same thing as

profit, because the partial budget does not include the other costs of

production which are not relevant to this particular decision. The

computation of total costs of production is sometimes useful for other

purposes, but will not be covered in this manual.

A partial budget, then, is a way of calculating the total costs that

vary and the net benefits of each treatment in an on-farm experiment. The

partial budget includes the average yields for each treatment, the adjusted

yields and the gross field benefit (based on the field price of the crop).

It also includes all the costs that vary for each treatment. The final two

lines are the total costs that vary and the net benefits.

Marginal Analysis

In the weed control example, the net benefits from herbicide use are

higher than those for hand weeding. It may appear that farmers would choose

to adopt herbicides, but the choice is not obvious, because farmers will

also want to consider the increase in costs. Although the calculation of

net-benefits accounts for the costs that vary, it is necessary to compare

the extra (or marginal) costs with the extra (or marginal) net benefits.

Higher net benefits may .not be attractive if they require very much higher

If the farmers were to adopt herbicide, it would require an extra

investment of $200 per hectare, which is the difference between the costs

associated with the use of herbicide ($600) and the costs of their current

practice ($400). This difference can then be compared to the gain in net

benefits, which is $520 per hectare ($3,720-$3,200).

So in changing from their current weed control practice to an

herbicide the farmers must make an extra investment of $200 per hectare,

and will obtain extra benefits of $520 per hectare. One way of making this

comparison is to divide the change in net benefits by the change in costs

($520/$200 = 2.6). For every $1 per hectare invested in herbicide, farmers

recover their $1, plus an extra $2.6 per hectare in net benefits. This

ratio is usually expressed as a percentage (i.e. 260%) and is called the

marginal rate of return.

The process of calculating the marginal rate of return, and deciding
if it is acceptable to farmers, is called marginal analysis.


In addition to being concerned about the net benefits of alternative

technologies and the marginal rates of return in changing from one to

another, farmers also take into account the possible variability in

Results. This variability can come from several sources, and researchers

need to consider these in making recommendations.

Experimental results will always vary somewhat from location to

location and year to year. An agronomic assessment of the trial results

will help researchers decide whether the locations are really

representative of a single recommendation domain, and can therefore be
analyzed together, or whether the experimental locations represent

different domains. This type of agronomic assessment helps refine domain

definitions and leads to more precisely targeted recommendations.

Another source of variability in experimental results derives from
factors that are impossible to predict or control, such as droughts,

floods, or frosts. These are risks that the farmers have to confront, and

if the experimental data reflect these risks, they will have to be included

in the analysis.

Finally, farmers are also aware that their economic environment is not

perfectly stable. Crop prices change from year to year, labor availability

and costs may change, and input prices are also subject to variation.

Although such changes are difficult to predict with any accuracy, there are

techniques that allow researchers to consider their recommendations against

possible changes in farmers' economic circumstances.




The first step in doing an economic analysis of on-farm experiments is

to calculate the total costs that vary for each treatment. Farmers will

want to evaluate all the changes that are involved in adopting a new

practice. It is therefore important to take into consideration all inputs

that are affected in any way by changing from one treatment to another.

These are the items associated with the experimental variables. They may

include purchased inputs such as chemicals or seed, the amount or type of

labor, and the amount or type of machinery. These calculations can be done

before the experiment is planted, as part of the planning process, in order

to get an idea of the costs of the various treatments'that are being

considered for the experimental program.

In developing a partial budget, we need a common measure. It is of

course not possible to add hours of labor to liters of herbicide and

compare these with kilos of grain. The solution is to use as a common

denominator the value of these factors, measured in currency units. This

provides an estimate of the costs of investment measured in a uniform

manner. This does not necessarily imply that the farmer spends money for

labor or receives money for the grain. Neither does it imply that farmers

are concerned only with money. It is simply a device to represent the

process the farmers go through of comparing the value of the things gained

and the value of the things given up.

An important concept in these calculations is that of opportunity

cost. Not all costs in a partial budget necessarily represent the exchange

of cash. In the case of labor, for instance, farmers may do the work

themselves, rather than hiring others to do it. The opportunity cost can be

defined as the value of any resource in its best alternative use. Thus if
farmers could be earning money as laborers, rather than working on their

farms, the opportunity cost of their weeding is the wage they would have

earned had they chosen not to stay on the farm and weed. The concept of

opportunity cost will be discussed at several points in the following

The field price of a variable input is the price expressed in units of
sale (e.g. $ per kilo of seed, liter of herbicide, man-day of labor, or
hour of tractor time). The field cost is the field price multiplied by the

quantity of the input needed for one hectare. Thus field costs are always
expressed in $ per hectare. If the field price of herbicide is $10 per

liter, and if 3 liters per hectare are required, then the field cost of the

herbicide is $30 per hectare. In both cases, the emphasis is on the field,
i.e. what the farmers pay to obtain the input and to get it to their farms.
Such field prices may be quite different from official prices.

Identifying Variable Inputs

To identify which inputs are affected by the alternatives included in
an experiment, you must be familiar with farmers' practices as well as the
practices used in the experiment. You must then ask yourself which

operations change from treatment to treatment, and make a list of these.

For example, consider an experiment in which three different
fungicides (A, B and C) are being tested, along with the farmers' practice

of no fungicide application. There are thus four treatments in the

experiment. The list of variable inputs is as follows:

Fungicide A

Fungicide B

Fungicide C

Labor to apply fungicide

Labor to haul water for mixing with the fungicide

Rental ,of sprayer to apply fungicide

This list includes purchased inputs (the fungicides), two types of

labor, and equipment (a sprayer). The following sections explain how to

calculate the costs for all of these.

Purchased Inputs

Purchased inputs include such items as seed, pesticides, fertilizer,

and irrigation water. For purchased inputs, you need to know the field

price of the input, the value which must be given up to bring an extra

unit of input into the field. The best way to estimate the field price of

a purchased input is to go to the place where the majority of the farmers

in the recommendation domain buy their inputs, and check the retail price

for the appropriate size of package. For instance, if farmers normally

purchase their insecticide in 1 kg packets in a rural market, that is the

price that should be used, rather than the price of insecticide in 25 kg

sacks in the capital city.

In some situations, the farmers will be selecting seed from their

previous crop, rather than buying seed. This seed is not costless either.

The best way to estimate the opportunity field price of the farmers' own

seed is to use the price that farmers pay when they buy local seed, either

from each other or at the market.

The next step is to 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 farmers and transportation costs are

probably not important. But for fertilizer, and perhaps-seed, this is not

the case. Usually the farmers must use a truck or an animal to get the

input home. If this is so, a transportation charge must be added to the

retail price. As most farmers pay others to transport such items for them,

it is not difficult to learn what the normal charges are. In general, you
will have to be guided by the practice that is followed by the majority of

farmers in the recommendation domain.

For example, if a 50 kg sack of urea costs $375 in the market, and it
costs $25 to transport the 50 kg to the farm, then the field price of urea

is calculated as follows:

$375 cost of 50 kg urea in market
25 cost of transporting 50 kg to farm
$400 field price of 50 kg urea

or $-0 $8/kg, field price of urea

Very often fertilizer experiments, especially those in the early
stages of investigation, utilize single nutrient fertilizers. The

treatments are usually expressed in terms of amounts of the nutrient (e.g.
50 kg N or 40 kg P205). In these cases, it is useful to go one step

further and calculate the field price of the nutrient. This can be done by

simply dividing the field price of the fertilizer by the proportion of

nutrient in the fertilizer. In the case of urea, which is 46% nitrogen,

$8/kg fertilizer = $17.4/kg N, field price of N
.46 kg N/kg fertilizer

The field cost of 50 kg nitrogen in a particular treatment would be 50

x $17.4, or $870 per hectare.

This should only be done when working with single nutrient

fertilizers, and it presumes that the field price of nitrogen (for

instance) is roughly the same in any nitrogen fertilizer available. If it

is not, researchers should of course be aware of this and take account of

these differences when considering which fertilizer to experiment with and


A final point is in order about purchased inputs. This discussion has
assumed that the inputs in the experiments are available in local markets,

or can be made available. If this is not the case, then the economic

analysis of experiments involving such inputs may be of little immediate

use to farmers. The results may be used, however, in order to communicate

to policy makers the possible benefits of making available a particular


Equipment and Machinery

Some treatments or alternatives may require the use of equipment not

required by other treatments. It is then necessary to estimate a field cost

per hectare for the use of the equipment.

The best way to estimate the per hectare field cost of equipment is to
use the average rental rate in the area. If farmers will rent their

sprayers for $20 per day and if the sprayer can cover 2 hectares in a

day, then the field cost may be taken as $10 per hectare. When estimating

the field cost.of tractor-drawn or animal-drawn implements, or small

self-powered implements, the average rental rate in the area can also be

used. This is especially appropriate if most farmers are renting these

implements anyway, but even for farmers who own their equipment, the rental

rate is a good estimate of the opportunity field cost. In certain cases a

pro-rated cost per hectare can be calculated, using the retail price of the

equipment and its useful lifetime, but this calculation involves factors
such as repair costs, fuel costs, and the possibility that the equipment

will have other uses on the farm. If a pro-rated field cost is to be
calculated, it is best to consult an agricultural economist who is familiar

with the equipment and costing techniques.


It is very important to take into consideration all of the changes in
labor implied by the different treatments in an experiment. Estimates of

labor time should come from conversations and perhaps direct observation

with farmers in their fields. Information about labor use derived from the

experimental plots is not very useful, however, if small plots are used in

the experiments. The best way to get this information is to visit with
several different farmers. Each will have an opinion as to the time

required for a given operation, but a number close to the average of these

opinions will provide a good estimate. Not all farmers take the same amount

of time for a given task, so the estimates will only be approximate. For

new activities, with which farmers are completely unfamiliar, some educated

guesses will have to be made until more reliable estimates can be

If farmers hire labor for the operations in question, the field price
of labor is the local wage rate for day laborers in the recommendation

domain, plus the value of non-monetary payments normally offered, such as
meals or drinks. This wage rate can be estimated by talking to several
.farmers. The field cost of labor for a particular treatment is then the
field price of labor multiplied by the number of days per hectare required.

When members of the farm family will do the work, it is necessary to
estimate the opportunity cost of family labor. This is the value which is
given up in order to do the work and thus represents a real cost. For
example, if farmers must take a day off from working in town to do extra
weeding, they will give up a day's wages, and this opportunity cost is just
as real as paying a laborer to do the work. And even if the farmer has
nothing to do but sit in the shade, the opportunity cost is not zero, as
most people put some value on being able to sit in the shade rather than

work in the sun.

The best place to start in estimating an opportunity field price for
family labor is the local wage rate (plus non-monetary payments). It is not

unusual to find the rate higher during some periods of the year than

others, and this must be taken into account.

It is sometimes difficult to estimate an opportunity cost of family

labor, especially if local labor markets-are poorly developed. Labor

availability may vary seasonally, or across different types of farm

households. Labor availability and labor bottlenecks are one of the most

important types of diagnostic information to aid in the selection of

appropriate treatments for experiments and in the definition of

recommendation domains. If labor is scarce at a particular time, extreme

caution must be used in experimenting with technologies that further

increase the labor demand at that time. In cases such as this, it is

reasonable to set the opportunity cost of labor above the going wage rate.

On the other hand, if extra labor is to be used during a relatively slack

period, an opportunity cost below the going wage rate might be appropriate.

But in no case should the opportunity cost of labor be set at zero.

In situations where most farm labor is done by the family, and where

new technologies are being considered that change the balance between cash

expenditures (i.e. for inputs) and labor, special care must be taken in

estimating labor costs. If a particular treatment, for instance, involves a

large change in labor input, relatively small differences in the

opportunity cost of labor will have significant effects on the estimation

of the.cost of the treatment. It must be remembered that assigning a

monetary value to labor is simply a way of helping to understand the

trade-offs involved for the farmers and to quantify net changes in the

value of all resources implied in moving from one practice.to another.

Total Costs That Vary

Once the variable inputs have been identified, their field prices have

been determined and the field costs have been calculated, the total costs

that vary for each treatment can be calculated. The following is an


Table 2.1

Weed Control by Seeding Rate Experiment (Wheat)






Weed Control

no weed control

herbicide (2 It/ha)'

no weed control

herbicide (2 It/ha)

Seeding Rate

120 kg/ha

120 kg/ha

160 kg/ha

160 kg/ha

price of seed

price of herbicide

price of labor

price of sprayer

to apply herbicide

to haul water

$ 20/kg


$250/day (local wage rate)

$ 75/day (rental rate)

2 days/ha

One laborer can haul 800 liters per day

(400 liters of water per hectare are

required for the herbicide)








Table 2.2

Calculation of Costs that Vary

Cost of

Treatments 1

Treatments 3

Cost of

Treatments 2

Cost of

Treatments 2

Cost of

Treatments 2

Cost of

Treatments 2


and 2: 120 kg/ha x $20/kg = $2,400/ha

and 4: 160 kg/ha x $20/kg = $3,200/ha


and 3: 2 It/ha x $350/ha = $700/ha

labor to apply herbicide:

and 3: 2 days/ha x $250/day = $500/ha

labor to haul water:

and 3: 400 liters required x $250/day = $125/ha
800 liters per day


and 3: 2 days/ha x $75/day = $150/ha

These costs that vary are then listed by treatment and the total costs

that vary are calculated.



Table 2.3

Total Costs that Vary for Weed Control by Seeding Rate Experiment

Treatment Treatment Treatment Treatment

1 2 3 4

($/ha) $2,400 $2,400 $3,200 $3,200
icide ($/ha) 700 0 700

Labor to apply herbicide

Labor to haul water ($/ha)

Sprayer (,$/ha)

Total Costs That Vary
















The perceptive reader will have noticed that all of the costs that

vary have not been treated in this chapter. There are two important

exceptions. Costs that are associated with harvest and marketing are

discussed in the next chapter, where they are included in the field price

of the crop. Costs associated with obtaining working capital, such as

interest rates, are discussed in Chapter 5.



There are several steps involved in calculating the benefits of the

treatments in an on-farm experiment:

1. The first step is to identify the locations that belong to one

recommendation domain for the experiment. The economic analysis is

done on the pooled results of an experiment that has been planted in

several locations for one recommendation domain.

2. Next, the average yields across sites for each treatment should be

calculated. If the results of these experiments are agronomically

consistent and.understandable, a statistical analysis of the pooled

results can then be carried out. If there is no reasonable evidence of

differences among treatment yields, researchers need only consider the

differences in costs among the treatments. But'if there are

significant yield differences, then researchers should continue with

the partial budget.

3. The next step is to adjust the average yields downwards, if it is

believed that there are differences between the experimental results

and the yield the farmers might expect using the same treatment.

4. A field price for the crop is then calculated, and multiplied by the

adjusted yields to give the gross field benefits for each treatment.

5. Finally, the total costs that vary are subtracted from the gross field
benefits to give the net benefits. With this calculation the partial

budget is complete..

Pooling the Results From the Same Recommendation Domain

The first line of a partial budget is the average yield for each

treatment for a particular experiment for all locations for a

recommendation domain. Recall that a recommendation domain is a group of

farmers whose circumstances are similar enough so that they will all be

eligible for the same recommendation. Tentative identification of

recommendation domains begins during the diagnostic and planning stages of

on-farm research. This tentative identification is used for selecting

locations for planting experiments. If, for example, soil fertility seems

to be a serious problem for farmers in a given area, the recommendation
domain for a fertilizer experiment might be defined in terms of farmers who

plant the target crop, have certain types of soil, and follow a particular
crop rotation. Locations for the experiments are chosen so as to represent

farmers with these particular circumstances. On analyzing the results it

may be that a factor not previously considered, such as slope of field, is
responsible for distinct results. In such a case, the experiments from the

tentative domain would not all be combined for economic analysis. Instead,

they would be divided into two domains (further defined by slope, in this
case), and two separate analyses would be carried out. More detail on how
and when to pool experimental results is presented in Chapter 7.

It should be noted here that although locations can'be eliminated from

analysis if it can be demonstrated that they do not belong to the

recommendation domain in question, this does not hold for locations that

were lost .due to drought, flooding, or other environmental factors that are

not predictable. Locations lost or damaged by such factors must be included

in the economic analysis. Further discussion of risk analysis is to be

found in Chapter 8.

Statistical Analysis

Before doing an economic analysis of the pooled results of an

experiment for a particular recommendation domain, researchers must assess

the experimental data to verify that the observed responses make sense from

an agronomic standpoint. The statistical analysis must then be reviewed in

order to judge the probability that the observed treatment differences

could be repeated. Performing an economic analysis on experimental data

that researchers do not understand, or do not have confidence in, is a

misuse of the techniques of this manual.

If the statistical analysis of the results of an experiment indicate

little likelihood that the observed responses represent real differences

between two treatments, then the lower cost treatment would be preferred.

When the statistical analysis indicates that treatment yields are probably

the same, the gross benefits for these treatments will also be the same,

and the lowest cost method of achieving those benefits should be chosen. If

two methods of weed control give equivalent results, for instance, the

methods with the lower costs that vary should be chosen (for further

experimentation or for recommendation) and no further economic analysis is


More details on the relation of statistical analysis to economic

analysis are given in Chapter 7.

Average Yield

When the recommendation domain for a particular experiment has been

established and agronomic and statistical assessments have indicated that

it is worthwhile to proceed with a partial budget, the average yields of

each treatment are entered on the first line of the partial budget.

Table 3.1 shows the results from 5 locations in one recommendation

domain of the weed control by seeding rate experiment. There were two

replications at each location.

The average yields for the four treatments are reported on the first

line of the partial budget (see Table 3.2)

Adjusted Yield

The next step is to consider adjusting the average yields.

Experimental yields, even those planted on-farm under representative

conditions, are often higher than the yields that farmers could expect

using the same treatments. There are several reasons for this:

1. Management. Researchers can often be more precise and timely than

farmers in applying a particular treatment, such as plant spacing,

timing of planting, fertilizer application, or weed control. Further

Table 3.1

Yields in kg per Hectare for Weed Control by Seeding Rate Experiment in One Recommendation Domain

Treatment 1

No weed control
120 kg seed/ha
(farmer practice)

Rep. 1 Rep. 2 A

2,180 2,220 2

2,800 2,640 2

1,720 1,880 1

2,680 2,620 2

530 670

Average yield


Treatment 2

herbicide (2 It./ha)
120 kg seed/ha








Rep. 1



















Treatment 3

No weed control
160 kg seed/ha

Rep. 1






Rep. 2 Av.

2,180 2,310

3,010 2,900

1,680 1,750

2,770 2,860

500 600


Treatment 4

herbicide (2 it/ha)
160 kg seed/ha

Rep. 1






Rep. 2













* Affected by drought-
^ /I







-/90s -1 ,L3

-32- p--


-4 ) l-O/CEL

t ~v

31 3-~

bias may be introduced if researchers manage some of the non-

experimental variables.

2. Plot size. Yields estimated from small plots often overestimate the

yield of an entire field because of errors in the measurement of the

harvested area and because the small plots tend to be more uniform

than large fields.

3. Harvest date: Researchers often harvest a crop at physiological

maturity, whereas farmers may not harvest at the optimum time. Thus

even when the yields of both researchers and farmers are adjusted to a

constant moisture content, the researchers' yield may be higher,
because of fewer losses to insects, birds, rodents, ear rots, or


4. Form of harvest: In some cases farmers' harvest methods may lead to

heavier losses than researchers' harvest methods. This might occur,

for example, if farmers harvest their fields by machine while

researchers carry out a more careful manual harvest.

Unless some adjustment is made for these factors, you will

overestimate the returns that farmers are likely to get from a particular

treatment. One way to estimate the adjustment required is to compare the

yields obtained in the experimental treatment which represents farmer

practice with the yields from check plots in the farmers' fields. Where this

is not possible, it is necessary to review each of the above four factors

and assign a percentage adjustment. As a general rule, total adjustments

between 5 and 30% are normally used. (A yield adjustment above 30% would

indicate that the experimental conditions are much different from those of

the farmers, and that some changes in experimental design or management

might be in order.)

For example, in the case of the weed control by density experiment in

wheat, researchers estimated that their methods of seeding and of herbicide

application were more precise than those of the farmers, and so estimated a

yield adjustment of 10% due to management differences. Plot size as well

was judged to be a factor, and a further 5% adjustment was suggested. The.

plots were harvested at the same time as the farmers, so no adjustment was

needed. Finally, the plots were harvested with a small combine harvester,

while the farmers use larger machines, and the difference in harvest loss

was judged to be about 5%. Thus the total yield adjustment for this

experiment was estimated to be 20%. The second line of the partial budget

(Table 3.2) thus adjusts the average yields.downwards by 20%. For instance,

the average yield for treatment 1 is 1,994 kg/ha and the adjusted yield is

80% x 1,994 or 1,595 kg/ha.

It is obvious that this type of adjustment is not precise, nor does it
pretend to be. The point is that it is much better to estimate the effect

of a factor than to ignore it completely. As researchers gain more

experience in an area they will have better estimates of the differences

between farmers' fields and their experiments, and yield adjustments will

become more accurate. -The yield adjustment, although approximate, should

not be looked upon as a factor to be applied mechanically. Each type of

experiment, each year, should be reviewed before deciding on an appropriate

adjustment. If this is done, researchers will be able to make decisions

about new technologies with a realistic appreciation of farmers'


Field Price of the Crop

The field price of the crop is defined as the value to the farmer of

an additional unit of production in the field, prior to harvest. It is

calculated by taking the price that farmers receive (or can receive) for

the crop when they sell it, and subtracting all costs associated with

harvest and sale that are proportional to yield, that is, can be expressed

per kg of crop.

The place to start is the sale price of the crop: This is estimated by
finding out how the majority of the farmers in the recommendation domain

sell their crop, to whom they sell it, and under what conditions (such as

discounts for quality). Unless all farmers store their crop for some time

before sale, it is best to use the price at harvest time. It is the amount

that the farmer actually receives, rather than the official or market price
of the crop, that is of interest.

Next, subtract the costs of harvest and marketing that are

proportional to yield. These may include the costs of harvest, shelling,

threshing, winnowing, bagging, transport to point of sale, or storage (if

the crop is stored for a period before sale). These costs will have to be

estimated on a per kg basis. In the case of harvesting or shelling, for

instance, this will require collecting data on the average amount of labor

necessary to harvest a field of defined size and yield, or shell a given

quantity of grain. Again, these may be cash costs or opportunity costs. If

farmers sell maize to traders for $6.00/kg, and they incur harvesting costs

of $0.30/kg, shelling costs of $0.20/kg, and transport costs of $0.20/kg,
then the field price of maize is $6.00 -($0.30 + 0.20 + 0.20) = $5.30/kg.

It must be emphasized that these costs must be accounted for because

they are proportional to the yield; the higher the yield of a particular
treatment, the higher the cost (per hectare) of harvesting, shelling and

transport. That is, the cost of harvesting,.shelling, and transporting 2

tons is almost exactly twice the cost of performing the same activities for

a harvest of 1 ton. As these costs will differ across treatments (because

the treatment yields are different), they must be included in the analysis.

Estimating these costs as part of the field price simplifies the partial

budget, because these costs do not then have to be estimated for each

individual treatment. These costs are treated separately from the costs

that vary described in Chapter 2 because, although they do vary across

treatments, they are incurred at the time of harvest and thus do not enter
into the marginal analysis of the returns to resources invested. That is,

farmers have to wait perhaps 5 months to recover their investment in

purchased inputs, but only a few days to recover harvest-related costs.

If there are costs associated with harvest or sale that do not vary

with the yield, then these should not be included in the field price, nor

in the partial budget. In the example of the weed control by seeding rate

experiment, the farmers sell their wheat in town for $9/kg. The harvesting

is done by combine, and the operators charge on a per hectare basis

(regardless of yield), so harvest cost is not included in the

calculation of field price. There is a bagging charge of $0.10/kg, a
transport charge of $0.50/kg, and a market tax of $0.40/kg, so the field

price of the wheat is $9 ($0.10 + 0.50 + 0.40) = $8/kg.

The gross field benefit for each treatment is calculated by

multiplying the field price by the adjusted yield. Thus the gross field

benefit for Treatment 1 is 1,595 kg/ha x $8/kg = $12,760/ha.

The above calculations for field price'assume that the majority of

farmers of the recommendation domain sell some of their crop. If they do
not sell any of their crop, then an opportunity field price for the crop

can be used. This is the money price which the farm family would have to

pay to acquire an additional unit of the product for consumption. This

should be the price in the market (plus transport costs) at the time of the

year when farmers are most likely to be buying food, and may be much higher
than official prices.

Net Benefits

The final line of the partial budget is the net benefit. This is

calculated by subtracting the total costs that vary from the gross field

benefits for each treatment. Table 3.2 is a partial budget for the weed

control by seeding rate experiment.

Including All Gross Benefits in the Partial Budget

The examples discussed above have assumed that a single product is the

only thing of value to the farmers from their fields. This is often not the

case. In many regions crop residues have fodder value, for instance. The

Average Yi

Adjusted Y

Gross Fiel

Table 3.2
Partial Budget. Weed Control by Seeding Rate Experiment

Treatment Treatment Treatment Treatment

1 2 3 4
eld (kg/ha) 1,994 2,444 2,084 2,600

ield (kg/ha) 1,595 1,955 1,667 2,080

d Benefit ($/ha) 12,760 15,640 13,336 16,640

Costs that Vary

Seed ($/ha)

Herbicide ($/ha)



Labor to apply herbicide

Labor to haul water ($/ha)

Sprayer ($/ha)

Total Costs that Vary

Net Benefits ($/ha)














* 3,200




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 (by treatment) and adjust the average yields. Then calculate a

field price. Of course "harvesting" becomes "collecting", "shelling"

becomes "baling", etc. It is important to consider each activity that is

carried out (is "chopping" done on maize fodder for instance?). Multiplying

the field price of the fodder by the adjusted fodder yield gives gross

field benefit from fodder, and this should be added to the gross field

benefit from grain.

Another important example is that of intercropping. If the majority of

farmers in the recommendation domain intercrop, then experiments should

reflect that practice. It may be that the experimental variables affect

only one crop, but if farmers intercrop their maize and beans (for

instance), then a fertilizer experiment on maize should include beans, or a

disease control experiment on beans should be planted with maize. The

yields of both crops should be measured. The partial budget would then have

two average yields, two adjusted yields, and two gross field benefits. The

total costs that vary would be subtracted from the sum of the two gross

field benefits to give the net benefits. Table 3.3 gives an example.

Table 3.3
Partial Budget for an Experiment on Bean Density and
Phosphorus Application in a Maize-Bean Intercrop

1 2 3 4
Bean Density .40,000 60,000 80,000 80,000
Phosphorus. rate 30 kg/ha 30 kg/ha 30 kg/ha 60 kg/ha
Average bean yield (kg/ha) 650 830 890 980
Average maize yield (kg/ha) 2,300 2,020 1,700 1,790
Adjusted bean yield (kg/ha) 553 706 757 833
Adjusted maize yield (kg/ha) 1,955 1,717 1,445 1,522
Gross field benefits beans ($/ha) 17,143 21,886 23,467 25,823
Gross field benefits maize ($/ha) 14,663 12,878 10,838 11,415
Total gross field benefits ($/ha) 31,806 34,764 34,305 37,238
Cost of bean seed ($/ha) 900 1,350 1,800 1,800
Cost of labor for planting
beans ($/ha) 450 675 900 900
Cost of fertilizer 1,050 1,050 1,050 2,100
Total costs that vary ($/ha) 2,400 3,075 3,750 4,800
Net Benefits ($/ha) 29,406 31,689 30,555 32,438





In the previous chapter a partial budget was developed in order to

calculate the total costs that vary and the net benefits for each treatment

of an experiment. This chapter describes a method for comparing the costs

that vary with the net benefits. This comparison is important to farmers,

because they are interested in seeing the amount of increase in costs

required to obtain a given increase in net benefits. The best way of

illustrating this comparison is with a net benefit curve, which plots the

net benefits of each treatment versus the total costs that vary. The net

benefit curve is useful for visualizing the changes in costs and benefits

in passing from one treatment to the treatment of next highest cost. The

net benefit curve also makes clear the reasoning behind the calculation of

marginal rates of return, which compare the increments in costs and

benefits between such pairs of treatments. Before proceeding with the net

benefit curve and the calculation of marginal.rates of return, however, an

initial examination of the costs and benefits of each treatment, called

dominance analysis, may serve to eliminate some of the treatments from

further consideration and thereby simplify the analysis.

Dominance Analysis

Table 4.1 lists the total costs that vary and the net benefits for

each of the treatments in the weed control by seeding rate experiment from

the previous chapter.

Table 4.1

Dominance Analysis

Weed Seeding
Treatment Control Rate Total Costs That Vary 'Net Benefits

#1 None .120 kg $2,400 $10,360

#3 None 160 kg $3,200 $10,136 D

#2 Herbicide 120 kg $3,875 $11,765

#4 Herbicide 160 kg $4,675 $11,965

Notice.that the treatments are listed in order of increasing total

costs that vary. The net benefits also increase, except in the case of

Treatment 3, whose net benefits are below that of Treatment 1. No farmer

would choose Treatment 3, in comparison with Treatment 1, because Treatment

3 has higher costs that vary, but lower net benefits. Such a treatment is

called a dominated treatment (marked with a "D" in Table 4.1), because

there is another alternative with a higher net benefit and lower costs that

vary. Such treatments can be eliminated from further consideration.

This example illustrates that it is important to pay attention to net

benefits, rather than yields, if you are interested in improving farmers'

incomes. Notice (from Table 3.2) that the yields of Treatment 3 are higher

than those of Treatment 1, but the dominance analysis shows that the value

of the increase in yield is not enough to compensate for the increase in

costs. Farmers would be better off using the lower seed rate, provided they

are not using herbicide.

Net Benefit Curves

The dominance analysis has allowed you to eliminate one treatment from

consideration because of its low net benefits, but it has not provided a

firm recommendation. It is possible to say that Treatment 1 is better than

Treatment 3, but in order to compare Treatment 1 with Treatments 2 and 4

further analysis will have to be done. For this analysis, a net benefit

curve is useful.

Figure 4.1 is the net benefit curve for the weed control by seeding

rate experiment. Each of the treatments is plotted according to its net

benefit and total costs that vary. The alternatives that are not dominated

are connected with a line. The dominated alternative (Treatment 3) has been

graphed as well, to show that it falls below the net benefit curve.

Marginal Rate of Return

The net benefit curve in Figure 4.1 shows the relation between the

costs and benefits for the three non-dominated treatments. The slope of the

line connecting Treatment 1 to Treatment 2 is steeper than the slope of the

line connecting Treatment 2 to Treatment 4. Calculation of the marginal

rate of return allows you to quantify this difference.

The purpose of marginal analysis is to reveal just how the net

benefits from an investment increase as the amount invested increases. That

is, if farmers invest $1,475 in herbicide and its application, they will

recover the $1,475 (remember, the costs that vary have already been

subtracted from the gross field benefits), plus an additional $1,405.

Nk* i -- xi~~

N4u?- eBw;s

C1 I"h )

-Ia,ooo -



10JO00 -

4 / I.r


Thr Vary CI|k)





~ ~SbD

o10\ Cosi,

An easier way of expressing this relationship is by calculating the

marginal rate of return, which is the marginal net benefit (i.e., the

change in net benefits) divided by the marginal cost, expressed as a

percentage. In this case, the marginal rate of return for changing from

Treatment 1 to Treatment 2 is:

11,765 10,360 $1405 95 95%
3,875 2,400 1475 -

This means that for every $1.00 invested in herbicide and its application,

farmers can expect to recover the $1.00, and obtain an additional $0.95.

The marginal rate of return for going from Treatment 2 to treatment 4

is calculated in a similar fashion.

11,965 11,765 $200 25 = 25%
4,675 3,875 T- -5 '

Thus for farmers who are using herbicide, and who plant at a rate of

120 kg seed/ha, investing in the higher seed rate would give a marginal

rate of return of 25%; for every $1.00 invested in the higher seed rate,

they will recover the $1.00 and.an additional $0.25.

The two marginal rates of return confirm the visual evidence of the

net benefitcurve; the second rate of return is lower than the first. It'is

possible to do a marginal analysis without reference to the net benefit

curve itself. Table 4.2 shows how tnis is done.

Table 4.2

Marginal Analysis

Costs that Marginal Net Marginal Marginal Rate
Treatment Vary Costs Benefits Benefits of Return
# 1 $2,400 $10,360
$1,475 $1,405 95% \
# 2 $3,875 $11,765
$ 800 $ 200 25% /
# 4 $4,675 $11,965

Note that the marginal rates of return appear in between the two

treatments. This is important, because it makes no sense to speak of the

marginal rate of return of a particular treatment; rather, the marginal

rate of return is a characteristic of the change from one treatment to

another. Also note that because dominated treatments are not included in

the marginal analysis the marginal rate of return will always be positive.

The marginal rate of return thus indicates what farmers can expect, on

the average, in return for their investment when they decide to change from

one practice (or set of practices) to another. In the present example,

adopting herbicide implies a 95% rate of return, and then increasing seed

rate implies a further 25%. As the analysis in this example is based on

only five experiments in one year, it is likely that the conclusions will

be used to select promising treatments for further testing, rather than for

immediate farmer recommendations. Nevertheless, a decision cannot be taken

regarding these treatments without knowing what rate of return is

acceptable to the farmers. Is 95% high enough? What about 25%? The next

chapter explains how to estimate a minimum rate of return.


In order to make farmer recommendations from a marginal analysis, it
is necessary to estimate the minimum rate of return acceptable to farmers
in the recommendation domain. If farmers are asked to make an additional

investment in their farming operations, they are going to consider the cost

of that investment. This is a cost that has not been considered in previous

chapters. Because of the critical importance of capital availability it is
treated separately. Working capital is 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. The cost of working capital (or simply,

the cost of capital) is the benefits given up by the farmer by having the
working capital tied up in the enterprise for a period of time. This may be

a direct cost, as in the case of a person who borrows money to buy

fertilizer and must pay an interest charge on it. Or it may be an

opportunity cost, the earnings of which are given up by not using money, or

an input already owned, in its best alternative use.

In addition to estimating the cost of capital, it is also necessary to

estimate the level of additional returns which will satisfy farmers, beyond

the cost of capital, to make their investment worthwhile. After all,
farmers are.not going to borrow money at 20% interest, for instance, to
invest in a technology that only returns 20%, and leaves them with nothing

to show for their investment. In estimating a minimum acceptable rate of

return, something must be added to the cost of capital to repay the farmer
for the time and effort spent in learning to manage a new technology.

There are several ways of estimating a minimum rate of return for a

particular recommendation domain.

A First Approximation of the Minimum Rate of Return

Experience and empirical evidence have shown that for the majority of

situations the minimum rate of return acceptable to farmers will be between

50% and 100%. If the technology is new to the farmers (e.g. chemical weed

control where farmers currently.practice hand weeding) and thus requires

that the farmers learn some new skills, a 100% minimum rate of return is a

reasonable estimate. If a change in technologies offers a rate of return

above 100% (which is equivalent to a "2 to 1" return, which farmers often
speak of) it would seem safe to recommend it in most cases.

If the technology simply represents an adjustment in current farmer

practice (such as a different fertilizer rate for farmers that are already

using fertilizer) then a minimum rate of return as low as 50% may be

acceptable. Unless capital is very easily available and learning costs are

very low, it is unlikely that a rate of return below 50% will be accepted.

This range of 50% to 100% is not as crude as it appears, and it should

always be remembered that the other agronomic and economic data used in the

analysis will be estimates or approximations as well. This range should

serve as a useful guide in most cases for the minimum rate of return

acceptable to farmers. This range represents an estimate.for crop cycles of

4-5 months. If the crop cycle is longer, the.minimum rate of return will b6
correspondingly higher. In areas where the inflation rate is very high,

this range should be adjusted upward by the rate of inflation over the

period of the crop cycle as well.

The Informal Capital Market

An alternative way of estimating the minimum rate of return is through

an examination of the informal capital market. In many areas, farmers do

not have access to institutional credit. They must either use their own

capital, or take advantage of the informal capital market (such as village

moneylenders). The interest rates charged in this informal sector provide a

way of beginning to estimate a minimum rate of return. Informal

conversations with several farmers of the recommendation domain should give

researchers a good idea of the local rates of interest. "If you need cash

to purchase something for the farm, to whom do you go?" and "How much does

this person charge for the loan of the money?" are examples of relevant


If it turns out that local moneylenders charge 10% per month, for

instance, then a cost of capital for a five month period would be 50%. To

estimate the minimum rate of return in this case, an additional amount

would have to be added to represent what the farmers expect to repay their

effort in learning about and using the new technology. This extra amount

may be approximated by doubling the cost of capital (unless the technology

represents .a very simple adjustment in practices). Thus in this example,

the minimum rate of return would be estimated to be 100%. Again, it should

be emphasized that this is simply a way of deriving a rough estimate of the

level of returns that farmers will require.

The Formal Capital Market

It is also possible to estimate a minimum rate of return using

information from the formal capital market. If farmers have access to loans

through private or government banks, cooperatives, or other agencies

serving the agricultural sector, then the rates of interest charged by

these institutions can be used to estimate a cost of capital. But this

calculation is relevant only if the majority of the farmers in fact have

access to institutional credit. If they do not, then they will probably

face a cost of capital different from that offered through relatively cheap

institutional credit. In some cases, it may be that farmers with otherwise

similar circumstances must be divided into two groups according to their

access to this type of credit. These two groups of farmers would face

different minimum rates of return and may well represent two separate

recommendation domains.

In other cases, institutional credit may be available to farmers, but

only for certain crops or rigidly defined credit packages. If such credit

is not likely to be available for the recommendations that are being

considered, then the cost of capital in these credit programs is not

relevant to the estimation of a minimum rate of return. This is another

example where on-farm research can provide information to policy-makers, in

this case by interacting with credit institutions to assure that their

services are directed to farmers in as efficient a manner as possible.

If farmers do have access to institutional credit, the cost of capital

can be estimated by using the rate of interest charged over the

agricultural cycle. That is, the rate of interest should cover the period

from when the farmers receive their cash (or inputs) to when they sell

their harvest and repay the loan. In addition, it is necessary to include

all charges connected with the loan. There are often service charges,

insurance fees, or even farmers' personal expenses for things like

transport to town to arrange the loan, that must be included in the

estimate of the cost of capital.

Once the cost of capital on the formal market has-been calculated, an

estimate of the minimum rate of return can be obtained by doubling this

rate. This will provide a rough idea of the rate of-return that farmers

will find acceptable if they are to take a loan to invest in a new



It is necessary to estimate a minimum rate of return acceptable to the

farmers of a recommendation domain. In the majority of cases it will not be

possible to provide an exact figure, but experience has shown that the

figure will rarely be below 50%, even for technologies that represent only

simple adjustments in farhier practice, and is most often in the

neighborhood of 100%, especially when the proposed practice is new to

farmers. Where farmers have access to credit, either through the informal

or formal capital markets, it is possible to estimate a cost of capital (or

an opportunity cost of capital) and then use this to estimate a minimum

rate of return. But even in these cases, it must be remembered that the

figure will be an approximate one. The next chapter explains how to use the

estimates of the minimum rate of return to judge which changes in

technology will be acceptable to farmers.


/ Chapter 4 demonstrated how to develop a net benefit curve and to

calculate the marginal rate of return between adjacent pairs of treatments.

Chapter 5 discussed methods for estimating the minimum rate of return

acceptable to farmers. The purpose of this chapter is to describe how to
compare the marginal rate of return with the minimum rate of return in

order to help decide which treatments represent reasonable alternatives for

farmers. It should be emphasized again that this type of analysis is useful

both for making recommendations to farmers, where there is sufficient

experimental evidence, and for helping select treatments for further


An Introductory Example

It might be best to start by returning to the example of the weed
control by seeding rate experiment summarized in Figure 4.1. After the

dominance analysis there were only three treatments left for consideration,

and the marginal rates of return were .calculated. If treatment 1 represents

the farmers' practice, will farmers be willing to adopt Treatment 2 or

Treatment 4?

Farmers will be willing to change from one treatment to another if the
marginal rate of return of that change is greater than the minimum rate of

return. In this case, if the minimum rate of return were 100%, the farmers
would probably not be willing to change from their practice of no weed

control, represented by treatment (1), to the use of herbicide, represented

by treatment 2, because the marginal rate of return (95%) is below the

minimum. If theminimum rate of return were 50%, then farmers would be
willing to change to treatment (2). Only if the minimum rate of return were

below 25% (which is very unlikely) would the farmers be willing to change

from (2) to (4). As long as the marginal rate of return between two

treatments exceeds the minimum acceptable rate of return, the change from
one treatment to the next should be attractive to farmers. If the marginal
rate of return falls below the minimum, on the other hand, the change from

one treatment to another will not be acceptable.

Two more examples follow:

A Fertilizer Experiment
.Figure 6.1 shows the results of a nitrogen experiment in maize. Table

/6.1 gives details on the design and costs that vary for the experiment. The
yield data are the average of 20 locations from three years of
experimentation. Table 6.2 is a partial budget for the experiment. Figure

6.2 shows the net benefit curve and Table 6.3 shows the marginal analysis
(one of the treatments is dominated).

In the recommendation domain where these experiments were planted
researchers estimated that the minimum rate of return for the crop cycle
was 100%. With 20 experiments over three years, researchers felt that they
were ready to make a nitrogen recommendation to farmers, who are currently
not using nitrogen fertilizer on their crop. What should be the
recommendation? Or in other words, if farmers are considering investing in.
nitrogen fertilizer (and the labor to apply it), what should be the
recommended level of investment?

y\ zk sc k

~::2c~f\Jdc~i hA-a.

4.. cw

N iO. (

3, -oo



Data on

Treatment (kg/ha)
A (Farmers' practice) 0
B 40
C 80
D 120
E 160

Table 6.1

Nitrogen Experiment

Number of Average Yield for
applications of N 20 locations (kg/ha)
0 2,222
1 2,867
2 3,256
2 3,444
2- 3,544

Economic Data

Field price of N = $0.625/kg

Field price of maize = $0.20/kg

Cost of one.fertilizer application = $5.00/ha

Yield adjustment = 10%

Minimum rate of return = 100%

Table 6.2

Partial Budget for Nitrogen Experiment


0 N 40 N 80 N

Average yield (kg/ha) 2,222 2,867 3,256

Adjusted yield (kg/ha) 2,000 2,580 2,930

Gross field benefit ($/ha) 400 516 586

Cost of nitrogen ($/ha) 0 25 50

Cost of labor ($/ha) 0 5 10

Total costs that vary 0 30 60

Net benefits 400 486 526

D .

120 N









160 N









Io N

8D N

I0 N

Tok\A Co&sA Tkc(

Cur e NtoA O ByPl .ic. A



(N 9A



vw )


Vet iV'penei

Table 6.3

Marginal Analysis. Nitrogen Experiment

Total Costs
Treatment that Vary Net Benefits
A. 0N 0 $400

B. 40 N $ 30 $486

C. 80 N $ 60 $526.

D. 120 N $ 85 $535

E. 160 N $110 $528 D*

* Treatment E is dominated

Marginal Rate
of Return




This analysis should always be done in a stepwise manner, passing from
the treatment with the lowest costs that vary to the next. If the marginal

rate of return of the change from the first to the second treatment is

equal to or above the minimum rate of return, then the next comparison can

be made, between the second and third treatments (not between the first and

third). These comparisons continue (i.e., increasing investment) until the

marginal rate of return falls below the minimum rate of return. If the

slope of the net benefit curve continues to fall, then the analysis can be

stopped at the last treatment which has an acceptable rate of return

compared to the one of next lowest cost. If the net benefit curve is

.irregular, then further analysis must be done. (See the following example).

In the nitrogen experiment, the marginal rate of return of the change

from 0 N to 40 N is 287%, well above the 100% minimum. The marginal rate of

return from 40 N to 80 N is 133%, also above 100%. But the marginal rate of

return between 80 N and 120 N is only 36%. So of the treatments in the

experiment, 80 N would be.the best recommendation for farmers.

There are a couple of things to notice about this conclusion. First,
the recommendation is not (necessarily) based on the highest marginal rate

of return. For farmers who use no nitrogen, investing in 40 N gives a very

high rate of return, but if farmers stopped there, they would miss the

opportunity for further earnings, at an attractive rate of return, by

investing in an additional 40 kg of nitrogen. Farmers will continue to

invest as long as the returns to each extra unit invested (measured by the

marginal rate of return) are higher than the cost of the extra unit

invested (measured by the minimum acceptable rate of return).

The second thing to notice is that the recommendation is not

(necessarily) the treatment with highest net benefits (120 N). If instead

of a step-by-step marginal analysis, an average analysis is carried out,

comparing 0 N with 120 N, the rate of return looks attractive (i.e.

535-400/85-0 = 159%), but this is misleading. The average rate of return of

159% hides the fact that most of the benefits were already earned from

lower levels of investment. This average rate of return lumps together the

profitable and the unprofitable segments of the net benefit curve. The

marginal analysis indicates acceptable rates of return up to 80 N. If the

farmers are to apply 120 N, the analysis shows they would only get a

marginal rate of return of 36% on their investment of the last $25. It is

likely that they would be willing to invest their money in nitrogen up to

80 N, and then ask if there is not some other way of investing that final

$25 (a little extra weeding, fencing for animals, etc.) that would give a

better rate of return than 36%.

This type of marginal (or stepwise) analysis is thus a more accurate

representation of farmer decision-making than alternative methods of
analysis. A benefit/cost analysis that compared each alternative directly

with farmer practice would have concluded that 120 N should be the

recommendation. Similarly, an analysis based on highest net benefit would

also have selected 120 N. The reasoning above shows that it is unlikely

that farmers would accept such a recommendation, because the cost of

capital is neglected.

Analysis Using Residuals

Another way of looking at the same problem is to use the concept of

"residuals". Residuals (as the term is used here) are calculated by

subtracting the cost of the investment (the minimum rate of return

multiplied by the total costs that vary) from the net benefits. Table 6.4

illustrates this method.

The treatments are listed, as usual, in order of total costs that

vary. Column (1) gives the total costs that vary and column (2) gives the

net benefits. Column (3) is the minimum acceptable rate of return

multiplied by the costs that vary, and is the cost (in terms of capital and

management) that the investment represents to the farmers. For instance, if

40 N has costs that vary of $30/ha, and if the minimum rate of return is

100%, this means that farmers would ask for returns of at least an

addi-tional $30/ha before investing in 40 N. Finally, the residual, column

4, is the difference between net benefits, column (2), and the cost of the

investment, column (3). Of course this residual is not the profit, and it

it the comparison between the residuals, rather than their absolute value,

that is of interest.

Farmers will be interested in the treatment with the highest residual,

that is, the treatment that has most benefits after accounting for the cost

of the investment. In this case, the treatment with the highest residual is
80 N, which is the same conclusion that was reached in the previous

analysis. Stopping at 40 N denies the farmers the possibility to earn more

money per hectare. Going on to 120 N implies actually losing some money,

after accounting for the cost of the investment.

Table 6.4
Analysis of Nitrogen Experiment Using Residuals

(1) (2) (3)

Total Costs Net Cost of Investment
Treatment that Vary Benefits 100% x (1)
A. 0 N 0 400 0
B. 40 N 30 486 30

C. 80 N 60 526 60

D. 120 N 85 535 85


Residual after
accounting for
minimum return
(2) (3)



This method of calculating and comparing residuals will always give

the same conclusion as the graphical method of marginal analysis shown

earlier. The method of using residuals, however, requires an exact figure

for the minimum rate of return, while the graphical method allows

comparison of the marginal rates of return with various assumptions about

the minimum rate of return. Thus it is advisable to use the graphical

method first and then, if necessary, check the conclusions with respect to

a particular minimum rate of return by calculating residuals.

In summary, the recommendation is not necessarily the treatment with

the highest marginal rate of return compared to that of next lowest cost,

nor the treatment with the highest net benefit, nor the treatment with the

highest yield. The identification of a recommendation requires a careful

marginal analysis using an appropriate minimum rate of return.

A Second Example: An Insect Control Experiment

A second example will illustrate some further aspects of marginal

analysis and the selection of recommendations. Figure 6.3 presents yield

data from an insect control experiment in maize. Table 6.5 gives details of

the design and the costs that vary. The yield data are the average of 6

locations from one year of experiments. Table 6.6 shows the partial budget.

Figure 6.4 shows the net benefit curve and Table 6.7 shows the marginal



4 lYx)





*-3 cl

i1f-erj AetAC


Table 6.5

Insect Control Experiment

1. Insecticide A
(Farmer practice)

2. Insecticide B

3. Insecticide A

4. Insecticide C

Type of Application
One sprayer appli-

Granular application

Two sprayer appli-

Granular application

Yield for 6
Quantity of locations
Insecticide (kg/ha)
2 It/ha 3,797

10 kg/ha

x 2 .t/ha

10 kg/ha




Economic Data

Field price of

Field price of

Field price of

insecticide A

insecticide B

insecticide C

Cost of applying insecticide A (including carrying water)

Rental of Sprayer

Cost of applying insecticide B or C

Field price maize = $0.08/kg

Yield adjustment = 20%

Minimum Rate of Return = 50%







Table 6.6
Partial Budget. Insect Control Experiment

1 2 3 4
Average yield (kg/ha) 3,797 4,344 4,500 4,609
Adjusted yield (kg/ha) 3,038 .3,475 3,600 3,687
Gross field benefit ($/ha) 243 278 .288 295
Cost of insecticide ($/ha). 10 20 20 32
Cost of labor ($/ha) 4 2 8 2
Cost of sprayer rental ($/ha) 1 0 2 0
Total costs that vary ($/ha) 15 22 30 34
Net benefits ($/ha) 228 256 258 261

)e, en3~ ;\ Cr.

~xseat CcA{n\ ~Tho~X e,'~


2So +

-l --H 1 I [i
Is 80 30 35-


l T T q tn l

Table 6.7

Marginal Analysis. Insect Control Experiment


Total Costs That Marginal Rate of
S Treatment Vary Net Benefits Return
1 15 228
2 22 256
3 30 258 42%
4 34 261

First, it should be noted that this insecticide experiment is
different from the nitrogen experiment in that it tests four distinct

treatments, rather than the continuous increase of one factor. (Notice that

the experimental results of the nitrogen experiment are represented on a

curve, while the results of the insecticide experiment are represented on

-a bar graph). It is impossible to use 80 kg N without using 40 kg N, but

using insecticide B does not require first using insecticide A. There are
& four different options, arranged on the net benefit curve in order of

increasing costs. The marginal analysis calculates the returns to

increasing investment and indicates which treatment should be recommended.

There is no implication that to choose one insecticide, the farmer must use

the insecticides of lower cost.

Second, the situation is a bit different from the previous example in

that only 6 locations from one year areavailable. Thus the analysis will

be used to help plan further experiments, rather than to make farmer

recommendations. In addition, the minimum rate of return in this

recommendation domain is lower 50% because farmers are already

controlling insects and the experiment is simply looking at alternative


Finally, the shape of the net benefit curve is different from the

previous example. The marginal rate of return in going from Treatment 1 to

Treatment 2 is 400%, well above the minimum. Therefore treatment 2 is

certainly a worthwhile alternative to the farmers' practice. Next, the

marginal rate of return in going from Treatment 2 to Treatment 3 is 25%, P
below the minimum. Treatment 3 can therefore be eliminated from .

consideration. But the marginal rate of return between Treatments 3 and 4

is 75%, above the minimum rate of return. In cases such as this, where the

marginal rate of return between two treatments falls below the minimum, but

the following marginal rate of return is above the minimum you must

eliminate the treatments) that are unacceptable and recalculate a new

marginal rate of return. In this example, it is necessary to calculate a

marginal rate of return between treatment 2 and treatment 4. The result

is 42% (261-256 42%), which is below the minimum rate of return. Thus
34-22 -
Treatment 4 is also rejected. If this last marginal rate of return had been

above 50%, however, Treatment 4 would have been the best treatment.

In this case researchers should continue to experiment with
insecticide B, which seems to be a promising alternative to the farmers'

practice of insecticide A. Treatments 3 and 4 give higher yields, but their

costs are such that they do not provide an acceptable rate of return. They

should be eliminated from future experimentation, unless agronomists

propose more effective dosages or application techniques, or there is

evidence that insect attack during this year was well below normal and that

another year of testing is worthwhile.

If residuals are calculated for this experiment, the same conclusion
will be reached. Table 6.8 shows the results; Treatment 2 is the one with

the highest residual.





Table 6.8
Analysis of Insect Control Experiment Using Residuals

(1) (2) (3) .(4)
Total Costs Net Cost of Investment Residual
that Vary Benefits 50% x (1) (2) --(3)
15 228 7.5 220.5

22 256 11 245

30 258 15 243

34 261 17 244


1. Is marginal analysis the "last word" for making a recommendation?

The marginal analysis is an important step in assessing the results of
on-farm experiments before making recommendations. But you should recall

that agronomic interpretation and statistical analysis are also part of the

assessment, as well as farmer evaluation. As researchers conduct on-farm

experiments, they must constantly solicit farmers' opinions and reactions.
Alternatives that seem to be promising both agronomically and economically

may have other drawbacks that only farmers can identify. To the extent

possible, this screening of treatments for compatibility with the farming

system should take place before experiments are planted. But farmer

assessment of the experiments is also essential. It is the farmers who

have the last word.

2. How precise is the marginal rate of return?

It is important to bear in mind that the calculation of the marginal

rate of return is based on yield estimates derived from agronomic

experiments and on estimates of various costs, often opportunity costs. As

well, the marginal rate of return is compared to a minimum rate of return

which is only an approximation of the goals of the farmers. Discretion and

good judgement must always play an important part in interpreting these

rates and in making recommendations. If the marginal rate of return is

comfortably above the minimum, the chances are good that the change will be

accepted. If it is close to the minimum rate of return then caution must be

exercised. In no case can one apply a mechanical rule to recommend a change

that is a few percentage points above the minimum rate, or reject it if it

is a few points below. Making farmer recommendations requires a thorough

knowledge of the research area and the problems that farmers face, a

dedication to good agronomic research and the ability to learn from

previous experience. Marginal.analysis is a powerful tool in this process,

but it must be seen as only a part of the research strategy.

3. Can the marginal rate of return be interpreted if the change in costs

that vary is small?

Certain experiments, such as those that look at different varieties or

perhaps modest changes in seeding rate, involve changes in costs that may

be quite small, perhaps equivalent to less than one hundred kilos of grain.

If the yield differences are at all substantial, the resulting marginal

rate of return can be very large, sometimes in the thousands of percent. In

these cases the marginal rate of return is of little use in comparing

treatments. Thus it is usually not worthwhile calculating marginal rates of

return for variety experiments, unless there are significant differences in

cost between varieties (e.g. local maize variety versus a hybrid).

4. Is it really possible to make recommendations, using marginal analysis,

without considering all the costs of production?

Remember that the starting point in on-farm research is the assumption
that it is much better to consider relatively small improvements in

farmers' practices, rather than propose large-scale changes. The idea is

thus to ask what changes can be made in the present system, and to compare
the change in benefits with the change in costs. Because the focus is on

the differences between two treatments, rather than their absolute values,

costs that don't vary between treatments will not affect the calculation of

the marginal rate of return. Table 6.9 shows two cases, both using the same

yields and costs that vary. In case A, the marginal rate of return is

calculated in the usual way. In case B, all of the costs of production are

included in the budget; they are of course constant ($300/ha) for each

treatment. When the marginal rate of return is calculated using benefits

and total costs, the result is the same.

Table 6.9

Marginal Analysis Using a Partial Budget and a Complete Budget

Case A Case B

1 2 1 2
Gross field benefits 500 650 Gross field benefits 500 650

Total costs that vary 100 200 Total costs that vary 100 200

Net benefits 400 450 Total of costs that don't vary 300 300

Total costs 400 500

Benefits 100 150

Marginal rate 450 400 50
of return 200 100
Marginal rate 150 100 50%
of return 500 400

5. Is the correct strategy always to consider small changes in the farmers'


Experience has shown that farmers are much more likely to adopt new

practices in small steps rather than in complete packages. But in following

this strategy it should be realized that farmers can (and do) arrive at an

adoption of a new set of practices, over a period of several years of

testing, rather than all at once. The size and.complexity of the individual

steps depends on the nature of the agronomic interactions of the elements

being tested and the resources available to farmers.

It is often possible to take advantage of this sequential adoption

pattern in making recommendations. Initial recommendations may be

intermediate between farmer practice and the recommendation that would be

selected by marginal analysis. Figure 6.5 is the net benefit curve for a

N -- T--QA4 -- rirue-L --

(O, W o)

) D0 )

o 'Rekrn = looWk

To~ \ CcssA -T V&i



*I Kjc

\ vre G. C

IV&t c~~enzrC11.~ Cyr~N

N As to

L voscir^ a Ep- Frimts

~~EiD, o)

nitrogen by phosphorus experiment. The curve shows that treatment (80, 40)

should be the recommendation.

Nevertheless, it is possible to first promote an intermediate

recommendation of nitrogen only (80,0) and then add phosphorus. The net

benefit curve gives assurance that farmers would be able to profitably

follow this stepwise adoption strategy.

More complex changes, such as the introduction of new crops or

cropping patterns, are of course possible as well. But such changes require

extremely careful planning and study and are beyond the scope of this


6. Why not include the minimum rate of return in the partial budget?

The partial budget includes all of the costs that vary across

treatments. The field price of the crop (used to calculate gross field
benefits) includes the harvesting and marketing costs that vary with the

yield. Why leave the cost of capital and the other returns on investment

that the farmers require (the minimum rate of return) until the last? It

would be possible, after all, to include the minimum rate of return on the

costs that vary as another element in the partial budget. The treatment

with the highest "net benefit" would than be the appropriate

recommendation. (This is what in fact is done with the calculation of

residuals). One reason for first calculating the marginal rate of return on

the additional costs that vary, and then comparing it to a minimum rate of

return, is the uncertainty regarding the minimum rate of return. It is

easier to calculate a marginal rate of return and compare it to several

possible minimum rates of return, rather than to calculate several

different residuals, each with a different minimum rate of return. A second

reason for leaving the cost of capital until the last is the fact that it

emphasizes the importance of the capital constraint for most farmers.

7. What is the difference between a marginal analysis and a continuous

analysis of data?

Agronomists often estimate response functions for factors such as

nutrients, and economists use similar continuous functions to select

economic optima. Yet the methodology of this manual uses a marginal

analysis comparing pairs of individual responses. There are three reasons

for emphasizing this latter method. First, marginal analysis, using

discrete points, can be used for any type of experimentation, while

continuous analysis is only applicable to factors that vary continuously,

such as fertilizer rates or seed rates. Second, the computational skills

and facilities necessary for estimating response functions are not always

available. Finally, farmer recommendations (e.g. for fertilizer levels)

will always be adjusted by farmers to their individual conditions.

A continuous economic analysis may be very useful in certain

situations, however. But if it is carried out, it requires the same degree

of care in estimating the benefits and costs that farmers face tnat has

been emphasized here for the construction of a partial budget and marginal

analysis. Sophisticated analyses done with uninformed assumptions about

yields, prices, or minimum rate of return will not give useful conclusions.

8. Does the marginal analysis assume that capital is the only scarce factor

for farmers?

In the marginal analysis all factors are expressed in terms of value,

estimated by a currency.equivalent. This is done not with the idea that

cash is necessarily the limiting factor, but simply to provide a common

unit for comparison. In an extreme case, for instance, marginal analysis

may be used in an experiment which compares treatments which differ only in

the amount of family labor utilized, for a crop which is not sold. In order

to decide whether extra amounts of labor would be effectively invested to

.produce extra amounts of the crop, opportunity costs and prices can be
assigned and the comparison made.

Nevertheless, in cases where family labor is the predominant source of

labor, and experimental treatments involve significant changes in labor
use, care must be taken in thinking about labor costs. If, for instance, a

change from one treatment to another implies a reduction in family labor
and an increase in cash expenditure, a modest increase in total costs that

vary may in fact represent a significant increase in cash outlay (balanced

to some extent by a reduction in labor "costs"). In cases where family

labor is a particularly important factor in farmer decision-making
regarding new technologies, a careful analysis must be undertaken. This is

complicated by the fact that the opportunity cost of labor is often

difficult to estimate. Different members of the household (men, women,
children) will likely have different opportunity costs of labor, and the

time of the year (slack season, peak season) will also affect the estimate.

One possibility is to do a sensitivity analysis (Chapter 9), which

involves doing several marginal analyses using different.estimates of the

opportunity cost of labor. Another technique involves estimating the

returns to labor for the treatments and comparing the marginal returns to

labor between two treatments with various estimates of the opportunity cost

of labor. Although beyond the scope of this manual, this is a reminder that

there are often alternative analytical techniques which may be useful for

helping to make decisions about the appropriateness of a particular


9. Can the concept of marginal analysis be used for planning


It is common to consider a change in farmer practice by doing a quick

calculation of how much additional yield would be needed to pay for the

extra costs of the new practice. For instance, if an extra 100 kg of

fertilizer costs $1000, and wheat is selling for $5/kg, then the estimate

might be that the farmers would need an extra 200 kg of wheat ($1000/$5) in

order to "repay the fertilizer". There are three errors in this kind of

calculation, however.

The first is in using market prices for fertilizer and wheat, rather
than field prices. The second is not including the labor or machinery costs

associated with the use of fertilizer. The third is in not including the

minimum rate of return. The following formula corrects those errors, and

provides a useful way for helping to consider practices that are proposed

for experimentation.

AY._ ATCV (1 + M)
Where AY = minimum change in yield required

ATCV = change in total costs that vary

P = field price of product

M = minimum rate of return

In the above example, if the additional fertilizer, plus the labor to

apply it, is worth $1,200, the field price of wheat is $4/kg, and the
minimum rate of return is 50%, then:

.AY $1,200 (1 + .5)
= 450 kg of wheat

Thus given current prices, the minimum yield increase required by
farmers from the addition of an extra 100 kg of fertilizer is 450 kg of

*wheat, not the 200 kg in the original calculation.

10. Can marginal analysis be used when yields are variable or prices


Yields in agronomic experiments are usually quite variable, and prices

often change. Methods for accommodating this kind of variability to

marginal analysis are discussed in Chapters 7, 8 and 9.






Marginal analysis for a particular experiment should be carried out on

the pooled results of at least several locations, over one or more years.

In order to prepare the experimental results .for this type of analysis,

several steps must be taken. First, researchers must review the purpose of

the experiment, in order to decide whether the results of the analysis are

to be used for making recommendations for farmers or for guiding further

research. Second, a review of the results from the different locations will

indicate whether all of the locations belong to the same recommendation

domain and can therefore be analyzed together. Finally, a combination of

agronomic judgement and statistical analysis will lead to a judgement

regarding the yield-differences among treatments in the experiment. If

researchers have little confidence that there are real differences in

yields, then the total variable costs of each treatment can be compared;

the treatment with the lowest costs will generally be preferred. If, on the

other hand, there is confidence that the differences observed represent

real differences among treatments, then a marginal analysis should be

carried out.

Reviewing the Purpose of the Experiment

Each experimental variable in an experiment has a purpose, and before
thinking about an economic analysis, researchers should review the

objectives of the experiment. Some experimental variables are of an

exploratory nature; they are meant to provide answers regarding response

(e.g. is there a response to phosphorus?) or to elucidate particular

production constraints that have been observed (e.g., is the low tillering

observed in the wheat crop due to a nutrient deficiency or to the

variety?). These variables are meant to provide information that can be

used in specifying production problems and designing solutions for them.

The treatments in these exploratory types of experiments are chosen in
order to provide the possibility of clear responses, and thus do not always

represent economically viable solutions to a particular problem.
Researchers must bear this.in mind when considering the economic analysis

of experiments with this type of exploratory variable. If the experimental
results provide clear evidence for a particular production problem, the

economic analysis may help to select possible solutions for testing. If a
high level of an insecticide in an exploratory experiment provided evidence

of a-response, but if the marginal analysis then showed an unacceptable
rate of return, researchers would want to examine lower levels or less

expensive insect control methods in subsequent experimentation.

Other experimental treatments test possible solutions to well defined
production problems. The solutions will have been selected for testing not
only because they offer promise of economically acceptable returns, but

because they are compatible with the farming system and do not represent

special risks to farmers. When there are yield differences among treatments

in these cases, the marginal analysis should be more rigorous, because a
recommendation may be made to farmers.

The marginal analysis should be done on the pooled results of a number
of locations, usually over more than one year. No strict rules can be given

here, but the number of locations should be sufficient to give researchers

confidence that the results fairly represent the conditions faced by

farmers in the recommendation domain. A very rough rule of thumb might be

to include at least a total of 20 locations over two years in the

recommendation domain. The exact amount of data required will depend on the

variability (across sites and across years) in the recommendation domain

and on the technology being tested. For instance, fertilizer

recommendations will usually require a fairly large number of locations, to

adequately sample the range of response by soil type, rotation, etc. Insect

control recommendations may require several years of evidence, in order to

sample year to year variability in insect populations.

Once recommendations are derived, they are often shown to farmers in
demonstrations, which may involve a single large plot with the

recommendation next to a similar plot with the farmers' practice. As a way

of following up on the recommendation the results of these demonstration
plots should also be subjected to an economic analysis.

Tentative Recommendation Domains i i

Whether the experiments are of an exploratory nature or are testing
possible solutions, they should be planted in locations that represent the

tentative definition of the recommendation domain. Recall that a

recommendation domain is a group of farmers whose circumstances are similar

enough that they are eligible for the same recommendation.

An example may help. In a particular research area there is

experimental evidence of a response to nitrogen in maize. Farmers currently

use no fertilizer, and an experiment is designed to test various levels of

nitrogen. The majority of the farmers plant maize under rainfed conditions,

although a few have access to irrigation. Because the response to nitrogen

with irrigation may be different from that under rainfed conditions, and

because of the small number of farmers with irrigation, only farmers with

rainfed fields are considered. (If there were more farmers with irrigation,

experiments might be planted with them as well; they would almost certainly

be a separate recommendation domain, however.) Most of these farmers have

sandy to sandy-loam soils. Locations are chosen to represent this range of

soil types, and careful note is taken in the field book of the soil type of

the location. The tentative definition of the recommendation domain

includes the range of soil types, but the experimental results may

distinguish separate domains. Non-experimental variables, such as variety,

planting date, and weed control are left in the hands of the farmers. A

certain range in these practices is present in the recommendation domain,

and the actual practices at each location are noted in the fieldbook. The

researchers do their best to reject locations that represent very unusual

practices or conditions (such as a few farmers who plant a special maize

variety that is used for sale as green maize.)

The tentative definition of the recommendation domain for the

fertilizer experiment is thus: "All farmers in the area who plant maize

under rainfed conditions on sandy to sandy-loam soils". This definition

allows for much variability in conditions and practices, and the selection

of experimental sites tries to represent this range, but avoids obvious


Notice that the recommendation domain is defined for the particular

experimental variable. A different experimental variable (say, a disease

resistant variety) might be tested in a domain of a different definition.

In this case, the variety might be tested on both irrigated and rainfed

fields, if no difference in its disease resistance capacity were expected.

Reviewing Experimental Results

The results of each experiment at each location in the tentative

recommendation domain must be reviewed. Inconsistencies in results between

locations can be due to'one of three causes:

1. Redefinition of the recommendation domain. In the above example, soil

type was being considered as a possible means of subdividing the

recommendation domain. If the responses are very different at

locations with sandy soils and those with sandy loam soils, then there

may be two separate recommendation domains (and two separate economic

analyses). Or it may be that an unexpected characteristic is of

importance. Suppose, in this same example, that some farmers plant a

maize-maize rotation, while others rotate their maize with a fallow.

If the responses to nitrogen are different on these two types of

fields, the original recommendation domain may be refined (by

eliminating the rotation that represents a minority of the farmers) or

divided (by rotation, if both rotations are of importance in the


The important point is that researchers.must have a clear and

consistent definition of the recommendation domain whose experiments

will be submitted to economic analysis. Domain definitions are

reviewed and refined during the experimental process, but usually only

when there has been a careful selection of experimental locations. The

number of possible defining characteristics for domains is greater

than the number of locations to be planted, so one cannot hope to let

a random selection of sites turn up characteristics for domain

definition. If there is a suspicion that a particular criterion may be

used to refine the definition of a domain, locations should be chosen


Finally, it should be very clear that a particular location cannot be

eliminated from the economic analysis, or assigned to a different

domain for separate analysis, simply because the experimental results

do not meet the expectations of the researchers. The change can be

made only if the results at that particular location can be explained

by a characteristic that the extension agent who is to make the

recommendation can recognize as a circumstance of particular farmers.

2. Experimental management. At times the results at a location may differ

from the others because of problems in experimental management. This

may include errors by the researchers, such as applying the wrong

dosage of a chemical, or factors related to the farmer, such as an

animal destroying part of the experiment or the farmer failing to weed.

(if this non-experimental variable were to be managed by the farmer).

In such cases the location can be eliminated from the analysis and the

researchers will gain a bit more experience in the management of

chemicals, in locating experiments where there is little chance of

animal damage, or in carefully.discussing with farmers their

responsibilities in the management of an experiment. Part of

experimental management includes the selection of locations. If

locations have to be eliminated because they have characteristics well
outside the normal range of the recommendation domain (such as very

late planting dates) this too is an indication of the necessity to

improve experimental management.

3. Unexplained or unpredictable sources of variation. After having

eliminated locations from the analysis because they do not represent

the recommendation domain, and sites where the management of the

experiment is responsible for unrepresentative results, there may

still be considerable variation in the results from the remaining

locations. This may be due to factors that are not understood (and may

be the focus of further agronomic investigation), or due to factors

that are understood but not predictable, and hence not eligible for

defining a recommendation domain, like drought or frosts. These sites

must be included in the economic analysis, unless researchers are able

to identify particular areas where the factor is more likely to occur.

It may be, for instance, that the research area can be divided into

more and less drought-prone domains. But if the drought (or frosts or

insect attack) cannot be associated with particular areas, then the

results of the affected locations must enter the analysis. More will

be said about treating these risk factors in Chapter 8, but it is

important to emphasize that locations that have been affected, or even

abandoned, because of these factors must be included in the marginal


Recommendation Domains Defined by Socioeconomic Criteria

The previous discussion of recommendation domains has focused on cases

where the experimental results from a set of locations may be regrouped

according to some characteristic of the locations or of their management.

There is also the possibility of using all the results from a set of

locations to do two separate economic analyses, according to a socio-

economic characteristic which may affect the economic viability of the

treatments. One example is-that of tenant farmers. Although forms of

tenancy vary widely, it is common for the landlord and the tenant to share-
the crop according to some formula. In some cases, the landlord and tenant

also share the cost of the inputs, while in other cases the tenant is

responsible for all cash inputs. In the latter case, this arrangement may

exert a very significant influence on the.choice of practices. In the case

of the nitrogen experiment in Table 6.2 if the tenants were to receive half

of the harvest but were responsible for all of the costs of fertilizer and

labor, Table 7.1 shows that the difference in marginal rate of return for

40 kg of nitrogen would drop from 287% for owners to 93% for tenants.

These two analyses could be done on a single set of experiments (a

combination of owners' and tenants' fields), provided that there was no

evidence of any differences (in management, soil type, etc.) between the
fields of owners and those of tenants.

Similar cases might be found for farmers who have the same

circumstances but pay very different transport costs (due to distance from

markets) or who face different opportunity costs of labor due to different

off-farm employment opportunities.

Table 7.1

Marginal Analysis for Two Recommendation Domains Defined by Tenancy

Owners Tenants

0 N 40 N 0 N 40 N

Adjusted yield 2,000 2,580 Adjusted yield 1,000 1,290

Gross field benefits 400 516 Gross field benefits 200 258

Total costs that vary 0 30 Total costs that vary 0 30

Net benefits 400 486 Net benefits 200 228

MRR 486 400 % MRR 228 200
S48 287% = 93%
30 0 30 0

Similar cases might be found for farmers who have the same

circumstances but pay very different transport costs (due to distance from

markets) or who face different opportunity costs of labor due to different

off-farm employment opportunities.

Statistical Analysis

In Chapter 3 it was pointed out that the economic analysis of an

experiment should only be done after reviewing the results of the

statistical analysis. If researchers have no confidence that there are

differences between treatments, then a marginal analysis is inappropriate.

Statistical criteria for making this judgement cannot be offered in this

manual, but it is important to remember that statistical analysis is only a

tool to help the agronomist assess the results of the experiment.

Conventional significance levels such as 5% are not relevant here. In the

final analysis, it is the judgement and the experience of the agronomist

that decides whether the differences between treatments in the experiment

are such that they warrant an economic analysis.

If the judgement is that there is no evidence of differences between

two treatments, then the treatment with the lowest total costs that vary

should be chosen, and no partial budget need be constructed. (There are

probably exceptions to this rule. For example, one fertilizer may cost a

bit less than another and give equivalent results in experiments, but if

agronomists know that this.fertilizer may contribute to soil acidification

which would cause problems within a few years, then the other fertilizer

should be considered).

Cases where no significant yield differences exist and no marginal

analysis is required are not necessarily trivial. If experimentation leads

to recommendation of a practice that lowers the costs of production while

maintaining yields, the gains in productivity of farmer resources are as

legitimate as those from a higher yielding (and higher cost) treatment. One

common example is that of substituting some form of reduced tillage for

mechanical tillage. This often results in considerable cost savings,

although yields may not be affected.

In on-farm experiments there are often cases where there is evidence

of differences between certain treatments, but not between others. Table

7.2 shows the results of a weed control experiment in which the farmers'

method of hand weeding was tested against several alternatives. A test of
mean separation indicates that there is little likelihood of any real

differences between treatments (1) and (2). It would be best to select the

lowest cost treatment of these two and compare it to treatments (3) and (4)

in a marginal analysis.

You should keep in mind that statistical tests are only an aid in

deciding if there are responses to the various treatments, and both the

statistical and economic analysis of the experiments are complements to the

agronomic assessment in the exploratory stage of experimentation. If these

results are from a few locations in one year, then several of the

treatments may warrant further testing the next year.

Table 7.2
Yield Data for a Weed Control Experiment

Treatment (kg/ha)

1. Two handweedings (Farmers'
practice) 2,275

2. Herbicide A 2,420

3. Herbicide B 2,990

4. Herbicide B + one hand weeding 3,350

A second case is that of factorial experiments. If an experiment looks

at two or more factors, and if the statistical analysis shows that one

factor is responsible for significant yield differences while the other is

not, then the average yields of the significant factor across those of the

other factor will enter the partial budget. Table 7.3 shows such a case, in

a nitrogen by insecticide experiment. There is a significant response to

nitrogen, but not to insecticide. The insecticide to be chosen for further

experimentation is the one which costs less. The partial budget for such an

experiment'will then have only two columns, corresponding to the two

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 - - mvs