Gaps Analysis & Improved
Methods for Assessing
Post-Harvest Losses
April 2017
Working Paper No. 17
Global Strategy Working Papers
Global Strategy Working Papers present intermediary research outputs (e.g.
literature reviews, gap analyses etc.) that contribute to the development of
Technical Reports.
Technical Reports may contain high-level technical content and consolidate
intermediary research products. They are reviewed by the Scientific Advisory
Committee (SAC) and by peers prior to publication.
As the review process of Technical Reports may take several months, Working
Papers are intended to share research results that are in high demand and should
be made available at an earlier date and stage. They are reviewed by the Global
Office and may undergo additional peer review before or during dedicated
expert meetings.
The opinions expressed and the arguments employed herein do not necessarily
reflect the official views of Global Strategy, but represent the author’s view at
this intermediate stage. The publication of this document has been authorized
by the Global Office. Comments are welcome and may be sent to
Gaps Analysis & Improved
Methods for Assessing
Post-Harvest Losses
Kebe, M. 2017. Gaps analysis & improved methods for assessing post-harvest losses. Working Paper No. 17. Global Strategy Working Papers: Rome
Table of Contents
A – Introduction..................................................................................................... 1 B – Gap analysis with methodological options..................................................... 7 B1. Concepts and definitions................................................................................. 8 B2. PHL studies as part of an integrated statistical programme within the national statistical system...............................................................................
12
B3. Scope of PHL studies........................................................................................ 16 B4. Statistical units of PHL studies......................................................................... 17 B5. Coverage of PHL studies.................................................................................. 19 B6. Frequency of PHL studies................................................................................. 20 B7. Time Reference of PHL studies........................................................................ 21 B8. Enumeration/survey period of PHL studies..................................................... 21 B9. General Statistical methodology of PHL studies.............................................. 21 B10. Cost-effectiveness aspects of sample survey PHL studies............................. 25 B11. Measurement of Harvesting Losses in PHL studies........................................ 26 B12. Measurement of Storage Losses in PHL studies............................................ 27 C – Methodological options for testing................................................................ 32 D – Summary & concluding remarks.................................................................... 35 E – Annex I: Modified count and weigh method.................................................. 46 E1. Introduction..................................................................................................... 46 E2. The methods.................................................................................................... 47 F – Annex II: Value chain/food supply chain approaches.................................... 50 G – Annex III: Country papers on PHL surveys...................................................... 56 G1. Introduction..................................................................................................... 56 G2. Malawi............................................................................................................. 57 G3. Indonesia......................................................................................................... 63 H – Annex IV: General survey framework (protocol) for field testing.............................................................................................................. 67 H1. Survey related issues....................................................................................... 67 I – Annex V: Ghana 2016-2017 field testing.......................................................... 97 I1. Objectives of the study..................................................................................... 97 I2. Institutional arrangements............................................................................... 99 I3. Coverage........................................................................................................... 100 I4. Scope................................................................................................................. 100 I5. Sampling design and selection of domains/strata............................................ 110 I6. Sample size and sampling procedure................................................................ 110 I7. Data collection method..................................................................................... 112 I8. Time reference, survey period and field enumeration..................................... 115 I9. Farm crop cutting technique............................................................................. 116 I10. Additional sampling considerations................................................................ 120 J- Annex VI: Summary of the desk study of the Malawi 2009-2010 PHL survey data......................................................................................................... 126 K - Annex vii: Selected bibliography...................................................................... 132
5
A Introduction In line with the current policy debate on agriculture and rural development, the
improvement of methods for estimating post-harvest losses was identified by
FAO member countries as a priority research topic to be included in the
Research Programme of the Global Strategy to Improve Agriculture and Rural
Statistics (GS). The objective of this research is to develop cost-effective
statistical methods for measuring post-harvest losses. This research has been
undertaken by an international senior consultant, Mr. Mbaye Kebe, under the
supervision and with the support of statisticians of the Global Office.
Activities started with a broad literature review that included previous FAO
publications, manuals, methodologies and guidelines on estimating post-
harvest losses as well as publications by other institutions, international
organizations, and relevant country experiences. Other technical reports have
been produced focusing on Definitions and Concepts, provided a synthesis of
Methods and Techniques for assessing post-harvest losses, identified key issues
and crafted the basic foundational elements for developing improved
methodologies to measure post-harvest losses. This material has been reviewed
by a small group of FAO experts and has been presented at a Technical
Workshop held in Washington DC in February 2015 and attended by the main
US and international institutions active in this field of research. The final
output of all these efforts has been published in the form of a working paper
(Working Paper No. 2) in September 2015 on the Global Strategy’s website
(Review of methods for estimating grain post-harvest losses) under the title
“Improving Methods for Estimating Post-harvest Losses - A Review of
Methods for Estimating Grain Post-Harvest Losses”.
During an Expert Group Meeting (EGM) held in Rome in April 2015, concepts
and definitions of post-harvest losses were further clarified and discussions
focused on gaps analysis and on methodological options and related field test
designs for countries to fill those gaps. Observations and recommendations
made during that meeting have been incorporated into the present report.
6
Following the proposed field test designs, a field test study of on-farm post-
harvest losses was conducted in Ghana during the period October 2016 –
March 2017. It was undertaken as part of the AGRIS field-testing activities and
labeled “Act. 1.3 - Experiments on crop production and post-harvest losses”. In
parallel, a desktop study was done on the basis of data provided by Malawi for
its 2009/2010 Post-Harvest Loss Survey: the micro-data was analyzed and an
attempt to identify the most important explanatory factors of post-harvest
losses in that country.
The present report provides additional relevant information and experience
gained from that Ghana field-test and the Malawi desktop study. However,
more detailed general and analysis reports are under preparation to provide
more insights on these undertakings.
7
B Gap Analysis with
Methodological Options The term gap analysis is used in the management literature to signify the
comparison between actual performance and potential or desired performance.
In this context of post-harvest loss (PHL) studies however, it is used as a tool
to contrast the current situation of methods and techniques for estimating post-
harvest losses in developing countries with the future state that the GS wants to
achieve (improved statistical methods for estimating losses) once the project is
complete. By so doing, methodological improvements in collecting, storing,
processing, analyzing and reporting/disseminating data on post-harvest losses
are identified to “bridge the gap”. This is the main objective and ambition of
this research project of the GS.
In this report, “bridging the gap” is then systematically considered for the
following data processes:
1. Collection
2. Storage
3. Processing, tabulation, and analysis
4. Reporting and dissemination
5. Organizing into an information system
These processes are the key ingredients of the balanced statistical programmes
advocated by the UN Statistical Office and FAO since 1966 as integral and
inseparable parts of national statistical systems for food and agriculture.
8
B1 CONCEPTS AND DEFINITIONS
Statistics must have a sound conceptual base, along with definitions and
concepts properly operationalized, to ensure that the resulting collected data is
interpreted and analyzed in order to turn it into useful information for decision
makers.
The literature on post-harvest losses abounds with a number
definitions/concepts; the most relevant are reviewed here and are classified as
follows:
High level/global
Operational
High level/global definitions/concepts
In February 2014, FAO released a definitional framework of food loss, as a
global reference for stakeholders, to use within their context of operations. In
essence, the terms and concepts in that FAO definition are:
Food waste is a part of food loss, however not sharply distinguished;
the term “food loss and waste” (FLW) is nevertheless maintained in
regular communication.
‘Intended for human consumption’ (already embedded in the Codex
definition of ‘Food’).
Plants and animals produced for food contain ‘non-food parts’ which
are not included in FLW.
Food redirected to non-food chains (including animal feed) is food loss
or waste.
Quantitative FLW = the mass (kg) reduction.
Qualitative FLW = reduction of nutritional value, economic value, food
safety and/or consumer appreciation.
For the World Resources Institute (WRI) “Food loss and waste” refers to the
edible parts of plants and animals that are produced or harvested for human
consumption but that are not ultimately consumed by people.
More specifically, “food loss”, refers to food that spills, spoils, incurs an
abnormal reduction in quality such as bruising, or wilting. Food loss
accordingly, is then the unintended result of an agricultural process or technical
limitation in storage, infrastructure, packaging or marketing. “Food waste”, on
the other hand refers to food that is of good quality and fit for human
9
consumption but that does not get consumed because it is discarded-either
before or after it spoils. Hence, food waste is the result of negligence or a
conscious decision to throw food away.
Operational definitions/concepts
These are more detailed and operational definitions found in the literature on
food loss assessments studies. The most commonly encountered terms (based
mainly on previous FAO work, and Boxall, R.A. (1986) are the following:
Grain: Is used in a broad sense and includes cereals and pulses; it includes
cereals on the head, ear or cob, and after threshing or shelling, and pulses both
shelled and in pod.
Food: Commodities which people normally eat: the weight of wholesome
edible material, measured on a moisture-free basis that would normally be
consumed by humans. Inedible portions of the crop, such as stalks, hulls and
leaves, are not food. Crops for consumption by animals are not considered
food.
Harvest: It refers to the act of separating the food material from the site of
immediate growth or production.
Post-harvest: This is the period after separation from the site of immediate
growth or production. It begins at cutting and ends when the food enters the
mouth. For most post-harvest loss studies, the end point is reached when the
grain or grain product is finally prepared for future consumption.
Loss: This is the measurable decrease of food grain which may be quantitative
or qualitative.
Damage: Is referring to the superficial evidence of deterioration, for example,
holed or broken grains, from which lass may result.
Grain loss: This is the loss in weight of food grain that would have been eaten
had it remained in the food chain.
Food loss: Any change in the availability, edibility, wholesomeness, or quality
of food that reduces its value to humans. Food grain losses may either be
characterized as direct or indirect.
10
Direct loss: This is the disappearance of food by spillage or consumption by
insects, rodents and birds.
Indirect loss: This is the loss caused by a lowering of quality leading to its
rejection as food.
Losses of crop product: Crop products may be lost from the food chain, at any
or all of the periods between planting and preparation for immediate
consumption. Three general periods have been identified.
Pre-harvest losses. They occur before the harvesting process begins and may
be due to such factors as insects, weeds, or diseases afflicting crops.
Harvest losses. They occur during the harvesting process and may be due to,
for example to shattering and shedding of the grain from the ears to the ground.
Post-harvest losses. They occur during the post-harvest period.
Post-production losses: The losses consist of the combined harvest and post-
harvest losses.
Quality loss: Local population and concerned traders assess quality of produce
in different ways according to the factors they consider important. General,
quality is assessed and products graded on the basis of appearance, shape, size,
and sometimes smell and flavor.
Nutritional loss: Reflects the product of the quantitative and qualitative losses;
more specifically, it is the loss in terms of nutritional value to the human
population concerned. Weight loss during storage, excluding loss of moisture
by definition, is a measure of food loss; however, the nutrient loss may be
proportionately larger due to selective feeding by pests.
Loss of seed viability: The loss of seed viability is related to the loss in seed
germination, which is important because of its effects on future food supplies.
This report will deal mainly with Post Harvest Loss (PHL) and not Food Loss
and Waste.
Observed gaps on definitions/concepts
It has been observed during the review of the literature that even between loss
assessment surveys, no uniform concepts, definitions, and classifications have
11
been systematically used. Different notions/concepts have been used by various
studies, in complete isolation from each other.
This has at least two implications; (a) terminologies, concepts and definitions
specific to loss assessment surveys should be uniform at least within a region
or country, and (b) loss assessment surveys making use of concepts like
household, holding, holder, should ensure that they conform to the standard
international definitions of those concepts.
On a more specific note, a formal definition of food supply chain or value was
almost never mentioned by the PHL assessment studies although that approach
was heavily used by a number of them.
The concept of harvest is mainly described as the act of separating the
foodstuff from its growth medium (reaping, uprooting, cutting, picking, etc.).
In developing country context however, harvesting may be and is often carried
out within a certain time frame and may comprise of additional crop dependent
activities performed by the farmers in the field that might include drying,
stacking, etc. It is only after these activities are carried that the product leaves
the farmers field for the next food supply chain level.
Options to improve on definitions/concepts
The concepts related to yield of crops should be properly defined and
harmonized because of its profound impact on measuring amount of food grain
loss through crop cutting. PHL studies literature discusses potential yield and
actual yield or obtained yield (which is the yield achieved by the farmer);
potential yield is defined as the yield which might have been achieved if there
had been no losses of any kind. In a document entitled “Rice and the yield gap
2004”, FAO distinguishes between i) theoretical potential yield, ii) experiment
station yield for which scientist conceive and breed potential varieties, iii)
potential farm yield, and iv) actual farm yield.
In general, concepts, definitions, classifications, in loss assessment activities
should be harmonized with those in other data sources within the national
system.
In relation with the specifics mentioned earlier, it is recommended to provide a
formal definition of value chain or food supply chain since this approach is
used in many PHL studies (what it is and what it comprises of).
12
The concept of harvest should be refined to include a temporal and component
dimension.
The guidelines on international classifications for agricultural statistics (GS,
2015), published on the Global Strategy’s website, should be used to improve
on the present situation.
B2 PHL STUDIES AS PART OF AN INTEGRATED
STATISTICAL PROGRAMME WITHIN THE NATIONAL
STATISTICAL SYSTEM
FAO Working Paper 2 “Improving Methods for Estimating Post-harvest
Losses - A Review of Methods for Estimating Grain Post-Harvest Losses”,
has elaborated on PHL assessment studies that have been mainly one or a
combination of the following types:
General baseline surveys
Probability sample surveys
Experimental designs - Field trials
Multivariate Linear Regression Fitting
Observed gaps on the approach of PHL studies
The general baseline surveys have been used in combination with the
probability sample survey as a recommended best practice.
Area-wide probability sample surveys were common in some PHL studies;
however they tended to cover a limited geographical area of the country; in
addition, during the measurement stage of the survey, they almost exclusively
relied on farmers’ declarations, ignoring the objective methods of
measurement. Furthermore, most of these studies focus on PHL at the storage
level without taking into account the whole supply chain (for example between
harvest and storage). This is the case, for example, of the PHL survey in
Malawi.
The majority of the loss assessment systems that have been investigated did not
make an explicit and clear-cut reference to an integration and/or linkage with
other statistical operations/studies/surveys within the national information
system of any given country;
13
Boxall R.A. (1986) had already observed that: “There has been a somewhat
piecemeal approach to post-harvest loss assessment. Studies of discrete parts
of the post-harvest system have been undertaken, but rarely have losses at the
different stages been considered in relation to each other...”
This state of affairs has remained mostly unchanged in recent times; in
addition, most studies have also been one-off, and in complete isolation with
each other and with the other surveys of any country's national statistical
system.
Options to improve on the integration of PHL studies
FAO publications in agricultural statistics since 1966, including the more
recent ones like “The World Programme for Census of Agriculture 2010”
(which contains developments on integrated systems), the global strategy to
improve agricultural and rural statistics, the guidelines on sectoral plans for
agricultural and rural statistics, all focus on integrated agricultural and rural
statistics systems, including an integrated survey framework and medium to
long term well-articulated and user driven survey programmes as proposed by
Global Strategy in the AGRIS project.
By emphasizing the linkages between these data sources, countries would
achieve the following:
1. conceptual and classification uniformity
2. optimize the use of scarce available statistical resources
3. prevent overloading any statistical operation/survey/inquiry with too
many items
4. prevent publishing conflicting statistics
5. provide all data needed for total analysis and finally
6. ensure full processing, analysis and user availability of collected data.
In other words, countries are encouraged to prepare long term integrated
statistical programs, as part of their national information system for decision-
making in food, agriculture and rural development.
Hence, national loss assessment/measurement exercises should also be
designed and implemented within these frameworks that factor in the
developing countries context and FAO's experience in that respect.
14
The FAO manual on assessment and collection of data on post-harvest food
grain losses (1980) states very clearly that: “To economize the collection of
data on food-grain losses, it will be desirable to link such surveys with some
other agricultural surveys such as crop-cutting surveys for the estimation of
total food production, food consumption surveys, etc... The estimation of post-
harvest losses will involve multidisciplinary approach and therefore in
planning such inquiry, specialists such as statisticians, plant pathologists,
entomologists, agricultural engineers and other technologists should be
associated.”
Every developing country has an agricultural statistics system to start with and
hence, they are encouraged to use a three-pronged strategy that gets data from
primary sources, administrative records, and technological research
information data sources. Primary sources data will be provided through the
conduct of proper survey making use of sampling techniques. Administrative
records are secondary data from government and other institutions like for
instance rainfall, temperature, humidity information, etc. These are data on
factor variables that impact loss variables. By technological research data, it
is meant the data generated by the conduct of agronomic trials or experimental
designs, as usually carried out by agronomists in government managed
research stations, or agricultural colleges.
There are a number of ways that the data sources within the system can be used
to provide relevant PHL data that is meaningful at various administrative levels
within the country. For countries following the FAO recommendations, the
most relevant data sources are:
the annual agriculture production survey (to be conducted every
year to provide crop production estimates); with crop forecasting
considered as a complementary activity; the statistical unit for the
agricultural production survey is the agricultural holding, and the
population from which the sample is drawn, should, if possible, cover
all holdings in the country.
the population census (to be conducted once every ten years in most
cases and sometimes once every five years to provide a frame and
population information). This source does not directly provide PHL
data but can be useful in indirectly providing frame and population
figures to be used by PHL surveys/studies.
the census of agriculture (to be conducted once in a ten year period)
can provide data on a) the structure of agriculture (number of farms,
15
their sizes, etc.) and b) provides a frame for other agriculture surveys.
The statistical unit is the agricultural holding.
the farm management survey (to be conducted twice in a ten year
period) collects detailed data on key aspects of decision making on
holdings like assets, organizational structure, allocation of resources,
input-output relations; the agricultural holding is hence the statistical
unit.
the post-harvest loss survey (to be conducted once in a ten year
period) aims to measure losses as consequence of waste and spoilage of
agricultural products between field and table, particularly for food
grains, the major part of agricultural production destined for human
consumption. Since the survey covers losses at various stages of
marketing, transportation and storage, the statistical units for the
corresponding phases of the survey are agricultural holdings,
intermediaries, and warehouses respectively.
the food consumption and nutrition survey (to be conducted twice in
a ten year period) focuses on the nutrition of individuals; data is
collected on the intake of food and its nutritive value from individuals;
for technical reasons however, the statistical unit is the household.
the household income/expenditure survey (hies, lsms, hbs, to be
conducted twice in a ten year period) with the aim of obtaining data on
the income of rural households derived from all sources and on the
expenditure patterns of those households. The statistical unit in this case
is the household.
other special agricultural surveys are the livestock censuses/surveys
collecting data on both animals and holdings with livestock with broad
enough scope to include all animal husbandry systems, including
nomadic ones; the agricultural holding is the statistical unit for this
survey. Price statistics (prices received by farmers, wholesale prices,
retail prices, export prices, import prices, and prices paid by farmers),
often collected through specialized price surveys, and may be collected
through other agricultural surveys. The statistical unit for collecting
price statistics will vary according to the type of price and the
prevailing marketing systems in the country.
administrative records generated and kept by government agencies for
their internal use, may serve some general purposes of PHL studies if
made available in statistical form. Records of prices at wholesale
markets, of imports and exports (food balance sheets), of livestock
slaughter houses, and other government projects are all useful sources
of data relevant to PHL concerns.
16
international statistical publications and databases of other
countries and of the international agencies are useful are useful for
cross-country PHL comparative studies; a country lacking statistics of
its own on a particular topic (PHL of maize), may adapt the statistics of
other countries having similar conditions with respect to that topic;
these international statistics are much more useful if the concepts
definitions used and the classification schemes adopted were in line
with international recommendations.
technological research information in food and agriculture is
generated from research performed in laboratories, universities and
research institutes; they involve experiments with food crops,
agronomic practices, harvesting techniques, storage technologies,
animals, and agricultural inputs; the main aim is to discover causal
relationships between various indicators (variables), in contrast to other
data sources which collect primarily descriptive data; because of its
nature technological research and other data sources complement each
other; statistics obtained from censuses and surveys, may help in
formulating hypotheses to be tested by technological research; research
results on the other hand may suggest the need for additional
information to be provided by censuses/surveys. A good example of
institutional set-up is found in Malawi, where the National Statistical
Office and the Ministry of Agriculture closely collaborate with
Agricultural Research Stations (Bvumbwe and Chitedze) for the design
and implementation of the PHL survey.
The country statistical programme will ensure proper scheduling and timing
between these operations so that information collected in one data source can
be used in the design of another.
PHL data collection can be piggy-backed on any of these surveys in the form
of specifically designed modules to augment their scope/coverage. In the case
of the annual production survey, modifications can be made to collect loss data
during the setting up of yielding plots for crop cutting.
B3 – SCOPE OF PHL STUDIES
The scope of the study concerns the data to be collected, the indicators that
constitute the measured facts, along with their respective dimensions (area,
yield, production and losses for the three most important food grain crops for
instance). Well-designed survey tabulation plan should provide a clear
indication of study scope. In general, the objectives of the study determine to a
17
large extent any given scope; hence the importance of clearly stated objectives
and tabulation plan when planning the study.
Observed gaps on the scope of PHL studies
Loss assessment studies have almost exclusively concentrated upon the
measurement of % weight loss and related quantities of loss along the value
chain, overlooking other relevant indicators like value of loss (integrating price
data), and loss indices (comparison of the total value of loss between different
time periods for instance), etc. The issue of price/economic related post-harvest
losses although very relevant will not be an area of focus in this paper.
In addition, scopes have varied largely between studies making it difficult to
circumvent a set of common data/information items that can be used for inter-
regional comparison. In most cases, the studies have covered only certain
points of the value-chain.
Options to improve on the scope of PHL studies
Ideally there should be a systematic approach of deriving a core set of relevant
indicators available in most loss assessment studies, which meet standard
quality requirements of usefulness, reliability, feasibility, and accessibility.
In conformity with the requirements expressed since the origins of loss
assessment studies (loss prevention and reduction), it is feasible to identify a
minimum list of indicators (survey items) for the purpose of monitoring &
evaluating loss prevention and reduction programmes encompassing all
relevant categories of the system (inputs, processes, outputs, outcomes,
impacts) at both national, agro-ecological, regional and sub-regional levels, as
well as other relevant dimensions (stages of the value chain, actors, loss
causing factors, etc.). This minimum list of items is reflected in the sample
questionnaires and data items presented in this report.
B4 – STATISTICAL UNITS OF PHL STUDIES
Statistical units are the different entities for which the required data items are
collected or derived, such as food supply chain actors – farm households,
collectors, wholesalers, retailers, government warehouses, etc.
Concepts related to statistical units are sampling units and observational units.
18
A sampling unit is one of the units into which an aggregate is divided for the
purpose of sampling, each unit being regarded as individual and indivisible
when the selection is made.
The observational unit in PHL studies, is in general the container, location, or
process from which a sample will be removed to determine the loss observed in
the sample, namely the unit in which grain is held (stacks in a crop field, small
silos, granaries on a farm, woven baskets, etc.). The accuracy of the entire
survey will be determined by the accuracy with which the loss is determined on
each observational unit.
Observed gaps on the statistical units of PHL studies
Studies dealing with losses at farm level do not always provide a clear
definition of the various statistical units under investigation such as fields,
parcels, etc. Some additional potentially useful information (measure of size) is
also lacking.
Options to improve on the statistical units of PHL studies
Proper definitions of the statistical units involved should be stated keeping in
mind international and inter-regional comparability.
Ideally, for losses at farm level, key recommended statistical units are:
the agricultural holding defined as an economic unit of agricultural
production under single management comprising all livestock, and all
land used wholly or partly for agriculture for agricultural production
purposes, without regard to title, legal form, or size (World Census of
Agriculture, FAO, 2010).
The concept of household, cited by the World Programme for the
Census of Agriculture 2010 (from a definition by the United Nations in
its guidelines for population and housing censuses) is based on “the
arrangements made by persons, individually or in groups, for providing
themselves with food or other essentials for living. A household may be
either (a) a one-person household, that is to say, a person who makes
provision for his or her own food or other essentials for living without
combining with any other person to form part of a multi-person
household or (b) a multi-person household, that is to say, a group of two
or more persons living together who make common provision for food
or other essentials for living. The persons in the group may pool their
19
incomes and may, to a greater or lesser extent, have a common budget;
they may be related or unrelated persons or constitute a combination of
persons both related and unrelated. A household may be located in a
housing unit or in a set of collective living quarters such as a boarding
house, a hotel or a camp, or may comprise the administrative personnel
in an institution. The household may also be homeless”.
B5 – COVERAGE OF PHL STUDIES
This refers mainly to the extent to which the studies have covered the national
territory and its main subdivisions.
Observed gaps on the coverage of PHL studies
A somewhat ad hoc and piecemeal approach to post-harvest loss assessment
studies have prevailed so far. Studies of certain parts of the post-harvest system
have been carried out in isolation with each other mostly using the small-scale,
virtually experimental study approach which in general produces relatively
precise data, but extends only to a very small area and so cannot be logically
related to provincial, regional or national efforts. Much of the available data on
losses relates to small farms with unimproved traditional systems, particularly
of storage. Moreover, the information has been derived from studies of a small
set of crops and chain actors, and only under a limited range of climatic
conditions.
Options to improve on the coverage of PHL studies
Not undertaking a country-wide PHL assessment study is not an option.
Countries should conduct area-wide sample surveys in order to provide
reasonable estimates of loss on which sound decisions can be taken about the
scope of a loss reduction programme. Experience has shown that the point at
which by far the larger errors occur is not at the analysis stage, which the
experimental study handles quite well, but at sampling in the field. The
experimental approach adopted in the early stages of methodology
development was sound, but there is a need now to complement it with an
approach (area-wide sample survey) that provides reasonable figures for
meaningful administrative levels in a given country at minimal cost, bearing in
mind that the studies must not also become so refined that they become
prohibitively costly. There has to be a balance between cost and precision.
The area-wide sample survey will aim at filling the gaps in the present
knowledge, for example, in relation to storage and processing of pulses,
20
groundnuts and root and tubers. The area-wide survey will take into account
factors such as farm cultural practices and management techniques in addition
to physical, chemical or biological factors that impact losses. The area-wide
survey will also ensure that greater consideration is given to the post-harvest
sector as a whole.
At the initial stages of the area-wide loss assessment sample survey it is
important to ensure that the objectives are clearly defined and that the use to
which the results will be put are carefully and completely examined and
understood. They will eventually determine the scope and coverage of the
undertaking. During these stages as well, crucial issues like timing of the study
in relation to the state of the season, the time required to train staff and other
survey parameters will be carefully handled. In the ideal scenario, a pilot
survey (dummy run) prior to the main study is required. The pilot survey will
serve as a suitable check for the methodology (especially survey objectives,
scope, data collection and sample analysis); it will also allow for preliminary
training of local staff (enumerators and supervisors), which is essential.
A more practical approach is advocated that encompasses the conduct of a
large scale, area-wide survey based on sound sampling designs, with more
detailed measurements and which covers many holdings. The accuracy of this
kind of study (area-wide sample survey) might not be as precise as the small-
scale (experimental study-like approach), but should nevertheless provide
reasonable figures at reasonable cost and, contrary to experimental studies,
allow for an extrapolation of the results to the appropriate geographical or
administrative level (e.g. regional or district averages of percentage losses).
B6 – FREQUENCY OF PHL STUDIES
Observed gaps on the frequency of PHL studies
There was no mention of whether the PHL studies were undertaken within the
framework of integrated national statistical programmes; most studies seem to
have been undertaken when resources were available from donor funded loss
reduction projects taking place in certain parts of the concerned countries.
Options to improve on the frequency of PHL studies
Ideally, area-wide PHL studies should be an integral part of the ten year
statistical programme together with the other data sources of relevance to any
given country.
21
Countries are encouraged to prepare 10 year programmes of data collection
activities including primary and secondary data sources in line with FAO
recommendations. In particular countries are encouraged to conduct at least
one country-wide PHL survey in a 7 or 10-year period in order to have
sufficiently up-to-date loss parameters which can be used to discount
production and provide good measures of food availability, a key variable for
food balance sheets.
B7 – TIME REFERENCE OF PHL STUDIES
Most PHL studies did not clearly explain their time reference, usually a period
of 12 consecutive months that can be either a calendar year or an agriculture
year.
The reports of PHL studies should provide detailed information on their time
reference.
In general, the time reference of specific items is the agricultural year with its
seasons or the day of enumeration. Countries are encouraged to indicate the
reference period for specific items of their PHL studies.
B8 – ENUMERATION/SURVEY PERIOD OF PHL STUDIES
Refers to the actual period in which the enumerators are active in the field.
This information is rarely reported in the PHL studies PHL studies should
mention their enumeration period so that readers can relate it to the study
reference year (enumeration done during or after that reference year).
To the extent possible short enumeration periods should be the norm, to avoid
duplication/omission brought about by variations in collected data, especially if
enumeration work is carried out through more than one round.
B9 – GENERAL STATISTICAL METHODOLOGY OF PHL
STUDIES
Statistical methodology encompasses a sampling design (including sample
sizes), sampling frame, as well as measurement and estimation techniques.
Sampling here refers to probability sampling, a method that ensures that every
observational unit in the population of interest has a known probability to be
included in the sample; tables of random numbers are used in the process and a
22
probability sample that avoids biases associated with non-probabilistic
techniques is created. These methods are different from purposive or
judgmental sampling where the researcher chooses the sample based on who
they think would be appropriate for the study.
For agricultural surveys, a sampling frame is a set of statistical or
observational units that takes the form of a list or a map identifying the
agricultural holdings. It is needed for sample selection. Ideally the frame
should cover all holdings in the country. The register of holdings (farm
registers) established in some countries may be an ideal frame; countries that
do not have farm registers can obtain a frame from a population census that has
been well coordinated with the agricultural census; other possible frames for
surveys include a list of enumeration areas prepared for the population census,
aerial photographs, satellite imagery, and maps.
Observed gaps on the statistical methodology of PHL studies
The use of appropriate statistical methodology to carry out area-wide post-
harvest losses surveys on a large scale has not been done in a systematic
manner in developing countries; there is some literature in the subject that
relates to work done in some developed countries; in the developing regions of
the world, proper statistical techniques (selection of sampling units,
determination of sample size, etc.) have rarely been applied for the estimation
of post-harvest food grain losses. In recent years, countries like Malawi and
Indonesia (see case studies in Annex) have undertaken country-wide PHL
assessment studies. In the case of Malawi, the sampling design of the PHL
survey is mentioned only in general terms in the methodological report.
Sampling weights have not been attached to the 2009/2010 data and have not
been used in the compilation of the loss indicators.
Options to improve on the statistical methodology of PHL studies
The general statistical methodology is proposed bearing in mind that a
preliminary general survey has been conducted to provide the points at which
significant losses arise in the food supply chain. Therefore, the main PHL
survey is conducted for those specific loss points of relevance to the country;
hence there is no further need of one statistical methodology taking into
account the hierarchical structure of the whole loss chain process.
For assessing loss at any stage of the supply chain, the first necessary step is to
define the population for which it is desired to estimate the loss. The scope –
data items – should be specified as well as medium to record them (paper or
23
electronic questionnaire); appropriate sampling procedures and measurement
techniques suitable to the situation at hand are next to be derived; following the
working out of the details of field organization, data are collected, verified, and
analyzed to obtain estimates of losses in line with the sampling design. With
the ubiquitous availability of computers of all types, standard errors of the
estimates should always be computed to gauge the reliability of the estimates
thus obtained. The estimates of losses at different stages of the supply chain
(harvesting, threshing/shelling, cleaning/winnowing, drying, storage, transport,
processing, packaging, and distribution), would be considered as follows: a)
losses at the farm level, b) losses at the level of intermediaries, c) losses at the
level of public agencies (warehouses, etc.).
Although the purpose of loss assessment surveys is to achieve loss prevention
and reduction, it might not be feasible to take-up the efforts of reduction in the
whole post-harvest system. Hence the necessity to identify the most serious
grain loss points in the country's post-harvest food supply system and initially
concentrate the efforts on the high loss points only. Therefore it is advised that
first a general baseline survey as already described be conducted prior to the
main loss survey using value chain/food supply chain approaches.
The special case of value chain/food supply chains approaches in PHL studies
is highlighted in Annex II of this report.
Building a suitable sampling frame is normally the very first stage when
planning an area-wide loss assessment survey. The recommendation is to use
the same sampling frame if not the same sample as for the agricultural surveys
when investigating losses at the farm level. Outside the farm level, frames for
the intermediaries will have to be prepared. If there are no surveys, then the
country has to develop an ad hoc frame for PHL assessment.
Primary and ultimate sampling units will vary depending on the level/stage of
the post-harvest process where losses are being measured. Lists of all these
units should be prepared to allow proper selection of sampling units.
At the harvesting level of the chain, villages (or their groupings) are commonly
the primary sampling units (PSU); holders are the secondary sampling units
(SSU), and fields (or granaries) the ultimate (tertiary) sampling units.
PSUs may be clustered prior to sampling. Simple or stratified random sampling
may be used to select the holders; for each chosen holder, a simple random
sample of subplots within fields can then be selected.
24
For threshing, cleaning, drying, transportation, and processing, villages are
PSUs and holders the secondary units.
For these levels, additional sampling stages will be necessary; for instance
selecting a random sample of produce (maize cobs, etc.) to be threshed and
observed for lost or damaged grain.
For storage, villages are PSUs, holders are the SSUs, and the storage units (if
there are more than one) within the holding are the ultimate sampling units.
The villages (PSUs) may be selected via simple random sampling, stratified
random sampling, or may be sampled with probability proportional to size.
FAO and the Global Strategy have produced a number of useful guidelines and
manuals on these procedures, to be consulted by the interested readers.
Proposed statistical improvements including measurement techniques have
already been reviewed at farm (holding) level, at intermediaries’ level, and at
government warehouses level, taking into consideration the stage of the food
supply chain, in the report “Improving Methods for Estimating Post-harvest
Losses - A Review of Methods for Estimating Grain Post-Harvest Losses”.
Based on that initial review, improvements for the design of the field recording
system for the collection of the data items (paper or electronic questionnaires)
are presented.
For large-scale PHL sample surveys, data are to be collected through
interviews supplemented by observations and measurements conducted by
trained enumerators using electronic devices (computers or tablets); during the
interviews, the enumerator visits the holding and the responses of the holder
(or representative) to questions are entered in the questionnaire. In many
traditional surveys, these used to be paper questionnaire. Technological
advances have now made it possible to use hand-held computer devices, PDA,
rugged notebooks and tablet PCs to use for field enumeration instead of paper
questionnaires. This will serve the purpose of making the data collected
machine-readable from its originating collection point and will allow more
automatic consistency checks at the field level. Data collected this way (that
can also include videos, pictures of insects, and mold infested crops, etc.) can
also be sent real time (via telephone and internet networks) to a remote server
for storage and speedy processing as see fit. An additional improvement being
recommended is to provide the enumerators with hand-held global positioning
systems (GPS) to record locations of households (or any geographic feature of
interest to the study) and also perform area measurement if necessary. In some
25
cases, GPS applications may be embedded in computers, tablets or other
electronic devices used for data collection.
In addition, interviewing may be combined with questionnaire mailing or even
better using web based questionnaire via Internet, for government, large and
very large scale commercial holdings enumeration.
B10 – COST-EFFECTIVENESS ASPECTS OF SAMPLE
SURVEY PHL STUDIES
PHL sample surveys can be costly when done with objective measurements;
hence, designs that decrease the cost of the whole operation and in the
meantime maintain a reasonable degree of precision can be utilized as an
improved option along the following lines:
1. Design the post-harvest loss survey using a two-stage, two-phase
sampling with villages (primary sampling units) at the first stage and
farm households as secondary stage units. The phase I sample will
consist in the full sample selected using the two (or more) stage
approach. This sample can be reasonably big while the households or
farms selected for the phase II sample size will be made much smaller
than phase I. In fact, phase II sample will be a sub-sample of phase I. In
terms of regression estimation, values of independent variables
(harvesting period, crop variety, fertilizer usage, etc.) are collected on
both phase I and phase II samples while the values of the dependent
variable (percentage weight loss) are only collected on phase II sample.
The rationale being that complicated measurement that are very time
consuming (sampling grains, measuring humidity, using visual scales)
can be completed on time if the sample size is relatively small as in
phase II. While in phase I, only questions will be asked via properly
designed questionnaires, and virtually with no time consuming physical
measurement.
2. Use the data from phase II to directly compute ratio or regression
estimates for the dependent variables in phase I sample based on the
already tested correlations (experimental design) between the
dependent and independent variables in phase II sample. Alternatively
values of the dependent variables in phase I sample can be scored after
fitting regression models between dependent and independent variables
in phase II sample.
26
By judiciously combining the annual production surveys, and the two-phase
post-harvest loss survey together with a proper timing between surveys,
countries should be in a better position to create a consistent database of
relevant information on post-harvest losses. For instance, the post-harvest
survey as described above is conducted in year 1 and the regression parameters
relating the dependent variable to the independent ones are computed. In year 2
even without a post-harvest loss survey, the country can use the annual
production survey to collect information on the independent variables on a
relatively big sample and use the available regression coefficients to produce
relatively good estimates of the dependent variables. In addition if resources
are available, the trials experimental designs can also be conducted to provide
information.
For this system to be successful however, it is of the utmost importance that
the right institutional setup be in place; in particular, biometricians,
agronomists, sampling statisticians within government research stations,
colleges/universities crop departments, etc., be working in close collaboration
with each other.
B11 – MEASUREMENT OF HARVESTING LOSSES IN PHL
STUDIES
In the literature, methods based upon standard crop-cutting techniques have
been proposed and used to some extent in assessing harvesting losses. The
discussion here will only refer to harvesting that do not make use of heavy
machinery like harvesters. Also crop-cutting techniques in an environment of
mixed cropping, etc. have been handled in other FAO documents
(Methodology for Estimation of Crop Area and Crop Yield under Mixed and
Continuous Cropping Technical Report Series GO-21-2017. March 2017.
http://gsars.org) and will not be treated here. Different approaches have been
used but there is no single ideal method. At harvesting, grains are lost mostly
because of scattering (shattering) or when the grain remains on the plant. The
basic approach consists of estimating the potential yield of the crop under study
using crop-cutting plots. There are two common ways of achieving that:
Harvest a demarcated test area (subplot of given shape and size) within
the farmer's field, in the most precise and careful way, in order to
achieve almost zero scattering and zero grain remaining on the plant;
this data is then used to estimate potential yield.
27
Glean with the most accuracy the grain that fallen on/under the ground
and grain still attached to the plant after the subplot was preliminary
harvested by a previous team.
When using this approach, researchers, should clearly specify whether the loss
being computed is expressed as a % of potential yield or a % of actual yield;
they should also clarify if the harvesting loss consists only of grain scattered at
the time of harvest or if it also includes mature grain left on the mature plant
during the harvesting operation; moreover, they should also specify whether
other losses, for instance grain removed by birds or other vermin have been
included.
There is more; in some countries growing crops like groundnuts,
stooking/stacking, and threshing and winnowing may all take place in the field
making it very complicated to properly set up the yielding plots, dig, lift,
stack/stook, and thresh the groundnut pods.
There is need for further investigation and comparison and testing in order to
establish a standardized methodology for assessing losses at harvest time using
crop-cutting yielding plots. As known from the practice of agricultural
production surveys with crop cutting components, there are crippling
challenges associated with estimating production through crop cutting yielding
plots. In addition, assessing losses at harvesting using crop cutting techniques
might not be suitable for fruits and vegetables, and hence other techniques
should be thought out and investigated.
B12 – MEASUREMENT OF STORAGE LOSSES IN PHL
STUDIES
The methods for assessing storage losses first described in the publication
Postharvest grain loss assessment methods compiled by Harris and Lindblad
(1978), have been applied and improved upon in a number of ways during
subsequent years. The TGM (thousand grain mass method), and the improved
count and weigh method are such enhanced techniques; However, there seems
to be a lack of practical experience in the field with these techniques.
The invention of the visual scale assessments techniques was another
breakthrough that features a relatively quick and easy way to estimate the grain
weight losses that are due to insect pest attack (bio-deterioration) and also to
assess grain quality. Visual scales are easy, quick to use and provide results on
the spot; their usage is also non-destructive; after scoring, the cobs or other
28
produce can be handed back to the farmer intact. Visual scales are used in the
PHL surveys in Malawi and seem to provide good results according to the
enumerators, supervisors and researchers that are in charge of the data
collection and measurements. There are however, some problems and
difficulties with them, which prompts the need to conduct further research. The
problems include the following:
1. There are issues with how consistent are the rankings performed by
enumerators/field workers; this has everything to do with the way the
enumerators/field workers make use of the reference photos they are
given to interact with the respondents.
2. How valid are the calibrations and assumptions used in correlating
weight loss with visible damage. This is an important issue if the scales
are to be used over a wide and/or variable area, or if the calibration
made in one area (region, province, etc.) or season is to be used in
another.
3. There is also the need to gauge what happens when the number of
classes in the visual scales are modified? Most of the scales have few
classes; there might be cases for instance where it is appropriate to
increase the number of classes, with the goal to recognize small
differences more easily. Additional testing work might be needed to
gauge the way the precision and accuracy of the results are dependent
upon the number of classes in each scale.
4. It is known that visual scales were originally developed for maize cobs
(relatively large objects that can be easily sampled). Though they have
been used to some extent on loose grains for other purposes, their usage
to estimate weight losses for both loose grains and other commodities
(especially fruits and vegetables) has not been widely documented.
Therefore, there is the need to investigate such utilization. An example
of visual scale used in the 2009/2010 PHL survey in Malawi is given
below. The first two classes (out of six) are displayed.
29
Figure 1 - Malawi 2009/2010 PHL survey: visual scale for maize cobs (classes 1 and 2)
It has already been mentioned in another report that practical experience with
methodologies for assessing storage losses caused by rodents and birds is
almost nonexistent.
Figure 2. Information system framework with food loss information
Broad planning and execution guidelines for building such information systems
are provided as follows.
30
For a preliminary system design and implementation, the following steps
should lead to the contemplated outcomes:
Set up a government coordinating body with participation of relevant
stakeholders, data users and producers, for overseeing the design and
implementation of the system.
Assemble all the material available on food loss, and make a
diagnostic/assessment of the current state of the existing food system,
and distribute the material to the participants.
Organize a workshop or series of workshops at the national level to
review the material with the aim of identifying all core
variables/indicators along with their dimensions that are relevant to the
country's food loss system and the major policy concerns.
Identify the data sources, statistical data management and analytical
activities necessary to produce the indicators listed previously.
Develop a 5-year plan for PHL information that will produce the
identified statistical and analytical activities, and ensure the plan is
approved by the national coordinating body. Ideally this plan will
include all necessary human, material and financial resources – staff
and training requirements, office space, vehicles, computers and
accessories, computer services, survey equipment (moisture meters,
GPS, tablets/PDA, cell phones, etc.) and annual operating expenses.
Implement the plan and regularly monitor and evaluate its performance
including setting up a mechanism to regularly collect up to date data
from diverse and complementary sources (area-wide field surveys,
administrative records, trials/experimental designs, etc.).
Following these guidelines will ensure the system is built in a rational and
systematic manner, which will highly increase its probability of success.
Once built and operating, the information system should be monitored closely
and evaluated in order to gauge its performance, and inform upcoming plans to
improve the country's food loss information system over time. Evaluating the
system and planning for new improvement should involve all major
stakeholders (producers, users, financiers of at various levels of food loss
information and related statistics) within the country. Relevant stakeholders,
for instance might be the Ministry of Agriculture, the Bureau of Statistics, the
National Monitoring & Evaluation Office, the Prime Minister's Office,
Agricultural Research Organizations, senior agronomists, statisticians, donors,
International Organizations involved in the food sector (FAO, etc.).
31
These stakeholders should be mobilized and coordinated for an efficient
evaluation process by the national coordinating body that has been instrumental
in the design and development of the food loss information system.
Built into the heart of this system there should be proven mechanisms to
regularly collect up to date data from diverse and complementary sources
(area-wide field surveys, administrative records, trials/experimental designs,
etc.). Extracted data would be transformed (if necessary) and loaded into the
system to be combined with other relevant data in order to produce the
necessary cross-tabulations that would satisfy planners, statisticians, and
decision and policy maker’s needs.
32
C Methodological Options for
Testing As an initial first step it is proposed the following list of options to be possibly
tested:
1. Using regression estimation techniques to reduce the cost of post-
harvest loss survey (through sample size reduction), using the two-
phase approach presented previously in this report.
2. Crop cutting to estimate harvest losses.
3. Crop cutting to estimate threshing/shelling, drying, cleaning/winnowing
losses.
4. Visual scales of three (3) classes versus six (6) classes.
5. Visual scales calibration variation across seasons, areas.
6. Experimental design studies to assess the effects of storage, harvest,
processing technologies on losses.
7. GPS, CAPI, Cell phones being used in data collection.
8. Thousand Grain Mass Method and/or modified count and weigh
method (ANNEX 1).
9. GIZ, FAO-AGS, or other organization's proposed rapid techniques
Although it is not possible to test all these methodological options in the
context of this research project, the most widely used and promising ones have
been or will soon be tested, especially options 1, 2, 3, 6 and 7. Below an
additional description of each of these methods or tests (some have already
been presented previously in this report).
1. In order to help reduce costs, it should be possible to design the post-
harvest loss survey using a two-stage, two-phase sampling design with
villages (PSU) at the first stage and holdings as SSU. The phase I
sample size can be reasonably big, while phase II sample size will be
made much smaller than phase I. Complicated measurements that are
33
very time consuming (sampling grains, measuring humidity, using
visual scales) can be completed on time if the sample size is relatively
small as in phase II. In phase I, there is no physical measurement; only
qualitative questions are asked to the respondents.
Values of the dependent variables in phase I sample can be scored after fitting
regression models between dependent and independent variables in phase II
sample. These can then be used for estimating averages and percentages of
interest. For this technique to work however, good correlations (good enough
to provide a reduction in sample size) between the dependent and independent
variables are needed.
2. The crop cutting conducted during the annual agriculture production
survey is used to estimate harvesting losses. The crop cutting plot is
selected at random within the field before harvesting by the holder. The
crop inside a crop-cutting plot (usually 10 meters x 5 meters, or 5
meters x 5 meters depending on the type of crop) is harvested according
to the usual farmer practices and the yield is weighted and recorded.
After the harvested produce is removed from the plot, all grains shed or
missed are then carefully picked up for estimating harvest loss.
3. The produce harvested from the crop cutting plot is this time taken
through the processes of threshing/shelling, drying, and
cleaning/winnowing by the enumerators according to the farmers'
practices.
Losses are then estimated according to the techniques already mentioned.
4. The invention of the visual scale assessments techniques was another
breakthrough that features a relatively quick and easy way to estimate
the grain weight losses that are due to insect pest attacks (bio-
deterioration) and also to assess grain quality. Visual scales are easy,
quick to use and provide results on the spot; their usage is also non-
destructive. There might be however, some problems: how valid are the
calibrations and assumptions used in correlating weight loss with
visible damage.
Can the scales be used over a wide and/or variable area, can the calibration
made in one area (region, province, etc.) or season be used in another.
What happens when the number of classes in the visual scales is modified?
34
Visual scales were initially developed for maize cobs (relatively large objects
that can be easily sampled). Though they have been used to some extent on
loose grains for other purposes, their usage to estimate weight losses for both
loose grains and other commodities (especially fruits and vegetables) has not
been widely documented. Therefore, there is the need to investigate such
utilization.
5. Technologies used by farmers to store their produce are known to have
a profound impact on storage losses.
Experimental designs (using randomized complete block designs with between
3 to 5 replications for example) can be conducted to determine the most
relevant factors (independent variables like crop varieties, technology, timing
of operations, etc.) that have net effects on losses.
6. GPS devices can be used to record latitude/longitude data (holding
location, holding area, etc.) that can then be used to visualize important
information on maps. CAPI This will serve the purpose of making the
data collected machine-readable from its originating collection point
and will allow more automatic consistency checks at the field level.
Data collected via CAPI and cell phones (that can also include videos,
pictures of insects, and mold infested crops, etc.) can also be sent real
time (via telecommunication links) to a remote server for storage and
speedy processing as see fit. Cheap solutions for both hardware and
software are available on the market.
7. The TGM (thousand grain mass method), and the improved count and
weigh methods are enhanced techniques used to measure losses due to
insects that were mentioned in the literature.
However, there seems to be a lack of practical experience in the field with
these techniques.
8. Methodologies describing rapid techniques for the assessment of losses
may have been proposed by GIZ, FAO-AGS, member countries and
other relevant organizations.
A general survey framework (protocol) for field testing some of these options
is detailed in Annex IV.
In Annex V, we illustrate with a practical example, how this general survey
framework was used in the case of the Ghana post-harvest loss pilot study.
35
D Summary & Concluding
Remarks Designing and optimizing PHL assessments should be undertaken within the
country’s information system (with components of data design, data collection,
data storage, data processing, data analysis, and data communication and
dissemination, and information technology) for food and agriculture decision
making for reasons explained above. A medium to long term integrated
programme of data collection activities encompassing primary and secondary
data sources (that include PHL studies) should be collaboratively designed,
implemented and regularly monitored by the main key stakeholders (Ministry
of Agriculture, Central Statistics Agency, Agricultural Colleges, Agronomic
Research Institutions, Agronomic faculties of relevant universities, etc.). This
should ensure optimal sharing of resources between data sources, and hence
bring down the cost of any single data source, and provide the solid ground for
undertaking loss assessment studies along the following general guidelines that
follow (details will be worked in the final handbook).
When designing a loss assessment study, it helps to bear in mind its purpose
(effective loss prevention and reduction), the different types of studies, their
combinations/interactions, their two main dimensions (study design, study
measurements techniques) and the desired end result as depicted in Figure 3.
36
Figure 3: Desired end result of a general PHL assessment study for the case of a variety
of rice in two seasons S1 and S2.
One should remember that there are three periods of time during which food
may be lost:
a. Preharvest losses occur before the process of harvesting begins.
b. Harvest losses occur between the onset and completion of the process
of harvesting.
c. Post-harvest losses occur between the completion of the harvest
process and the moment of human consumption.
Post-harvest intermixes to various extents with portions of the maturing-
drying-processing period and hence no sharp distinction can be made. Harvest
and post-harvest losses can be combined sometimes into a one single loss given
the elements of common concern between them; the term employed in that case
is “post production losses”. Therefore, we have the following relationships
between these losses:
1. Pre-Harvest
2. Post Production (Harvest and Post-Harvest)
The flow of produce from its sources (farm field or import docks), to the
consumer is described as a food supply chain (FSC) or value chain (VC), in
which losses can happen along the entire chain – during harvesting, drying,
transport, storage, and processing. Figure 4 below is a simplified example of
such a chain.
37
Figure 4: example food supply chain
It should be clear by now that the tasks of assessing losses can be very
complicated. In effect, separate measurements will be required for the different
types of losses occurring because of mishandling, biological deterioration
caused by insects, molds, rodents, moisture or other climatic conditions
(humidity, temperature, rainfall). Hence, sampling and assessing overall losses,
will entail evaluating components in the system and calculating their overall
effects.
The next following figures illustrate for each of the main actors in the FSC, the
different types of study design and study measurement techniques by type of
operation (harvesting, stacking/stooking, threshing/shelling, drying, transport,
winnowing, cleaning, storage, and milling/processing). It should be borne in
mind however, that each type of design is not exclusive: for example, using a
linear model to estimate losses based on the correlation between objective
measurements and the determinants of losses are often based on data collected
from surveys; coefficients may also be calibrated using data from experimental
designs, etc.
Study designs include survey sampling (probability surveys), experimental
designs, linear modeling, FAO-AGS/GIZ techniques, and other techniques
including non-probability surveys.
Measurement techniques include objective crop cutting as used in yield
measurement surveys, farmer’s estimates of yield (subjective), laboratory
analysis of collected samples of grains/produce, the use of visual scales, and
VC/FSC Analysis
Farm level
Processor level
Wholesale
level Retail level
38
the weighing of produce when they enter and exit processes like transportation,
drying, or milling.
39
•Survey sampling
•Experimental design
•Linear model
•Fao-Ags/GIZ, other
Harvesting
•Survey sampling
•Experimental design
•Linear model
•Fao-Ags/GIZ, other
Stacking/Stooking
Study Design Measurement
• Crop Cutting • Crop Cutting • Crop Cutting • Subjective
• Lab analysis/Vis/Weigh - in/out
• Lab analysis/Vis/Weigh - in/out
• Aphlis • Subjective
•Survey sampling
•Experimental design
•Linear model
•Fao-Ags/GIZ, other
Threshing/Shelling
Drying, Transport
Winnowing, Cleaning
• Weigh - in/out • Weigh - in/out • Aphlis • Subjective
•Survey sampling
•Experimental design
•Linear model
•Fao-Ags/GIZ, other
Storage
• Lab analysis/Vis/Weigh - in/out
• Lab analysis/Vis/Weigh - in/out
• Aphlis • Subjective
Farm Level
40
•Survey sampling
•Experimental design
•Linear model
•Fao-Ags/GIZ, other
Milling
•Survey sampling
•Experimental design
•Linear model
•Fao-Ags/GIZ, other
Drying
Study Design Measurement
• Weigh – in/out • Weigh – in/out • Subjective
Farm Level
• Weigh – in/out • Weigh – in/out • Subjective
41
•Survey sampling
•Experimental design
•Linear model
•Fao-Ags/GIZ, other
Transportation
•Survey sampling
•Experimental design
•Linear model
•Fao-Ags/GIZ, other
Storage
Study Design Measurement
• Weigh – in/out • Weigh – in/out • Subjective
Farm Level
• Lab analysis/Vis • Lab analysis/Vis • Subjective
42
•Survey sampling
•Experimental design
•Linear model
•Fao-Ags/GIZ, other
Display
•Survey sampling
•Experimental design
•Linear model
•Fao-Ags/GIZ, other
Transportation
Study Design Measurement
• Lab analysis/Vis • Lab analysis/Vis
Farm Level
•Survey sampling
•Experimental design
•Linear model
•Fao-Ags/GIZ, other
Storage
• Lab analysis/Vis • Lab analysis/Vis
• Weigh – in/out • Weigh – in/out
43
For the purpose of loss assessment, there should be at least two moments:
1. The Planning/Research moment during which the data needed to be
collected is identified. The supply chain actors, activities, and their
levels/stages are described; points in the system where the losses are
heavy (and their magnitude) are identified; the main products, the
reasons/causes for losses as well as the main technologies used are
listed. The Food Supply Chain/ Value Chain approaches (FSC/VC) as
described by GIZ and FAO-AGS can be used to achieve the objectives
of this moment. The FSC/CV concept is explained in Annex II.
2. The Main Study will then build upon the previous moment to obtain
statistically reliable quantitative data on losses for the relevant
actors/chain level at village, provincial, regional, or national level.
Proper sampling designs, objective measurement techniques and
methods of sample analysis will be needed. These procedures are costly
and will often require enough specially trained personnel. They are very
well suited to the chain activities of harvesting, threshing, drying, and
processing, and less suitable for chain processes storage. They are the
recommended “gold standard” in terms of loss study design and
implementation, provided the measurement is done properly and as
indicated. In this perspective, the preparation of detailed enumerators
manual and the organization of trainings for the data collection teams is
key to the success of this method.
Keeping in mind cost-effectiveness issues, the main survey could be
undertaken once every 7 or ten years, preferably after a census of agriculture
(census of agriculture that can also be used to collect limited qualitative data on
losses). This main survey can also make use of two-phase sampling techniques
(with ratio and regression estimators) to reduce sample size for the variables
that are costly and time consuming to measure. The usage of modern
technologies (GPS, CAPI, etc.) can also help bring down the costs of such
undertakings; in the body of the report it has also been mentioned that a clever
integration of data sources/survey instruments within an integrated statistical
system can as well help reduce costs.
When these types of surveys cannot be undertaken, a next best option is to use
experimental designs-field trials to derive acceptable estimates of percent
weight loss as has been done in some countries. Storage losses can be
estimated using simulation trials at research stations or at farm level. Results
from these trials should however be treated with caution because they are only
44
valid for the local conditions under which they were conducted. The FSC/VC
approaches when used to estimate losses share the same limitations. They
might however produce a better objective picture if they can be replicated over
the different agro-ecological zones comprising of a country. This would
however increase the cost of these experiments and reduce their quality-to-cost
ratio with respect to traditional sample surveys and measurements.
The main phases (to be detailed in the final handbook) for a loss assessment
study using survey sampling methodology are outlined as:
1. Preparation of preliminary survey proposal with sampling design,
methodology, survey scope and coverage (including crop
commodities), survey institutional setup, and survey activities and
budget.
2. Planning/Research during which all relevant administrative and
technological research information pertinent to the survey
implementation are collated.
3. Undertake a proper stratification of the country according to agro-
ecological zones, regions, provinces, districts, etc. if that does not
already exist.
4. Map the average FSC for the given the initial scope/coverage for any
given stratum. This can be a complicated task; in effect, there can be
thousands of supply chains at any given moment in any given stratum.
Supply chains are network like structures with entry nodes,
intermediary nodes, and exit nodes. Entry nodes for instance can be
centers supplying seeds to intermediary nodes like producing farmers,
who in turn supply village wholesalers/retailers, who supply consumers.
Village wholesalers may supply city/town retailers, who in turn may
supply regional/national retailers, and so on. So, in a region where there
are three (3) centers supplying four (4) seed varieties to say 3 farmers
organizations, who sell to three (3) wholesalers, who in turn supply 20
retailers in four (4) market centers, who finally sell to 40 shop owners,
the number of chains tend to grow exponentially (3x4x3x3x20x4x40) in
this simplified scenario. Hence agricultural economists and similar
subject matter specialists must decide on an “average chain”; they will
then use social sciences techniques like focus group discussions and/or
other rapid techniques that will provide better understanding of critical
loss points within the chain to be investigated during the main survey.
45
All this pertinent information is then handed over to the sampling
statisticians who will then prepare sampling frames for relevant
farmers, wholesalers, retailers/other intermediaries, millers/processors,
and draw the samples.
5. The preliminary survey proposal is then updated based on the new
information from the FSC analysis and implementation started. Sample
size is finalized and samples are drawn.
6. Questionnaires (paper/capi) are prepared, a pilot survey conducted,
enumerators/field staff recruited/trained.
7. For a study where losses occur mainly because of bio deterioration
(storage, long term stacking/stoking), laboratory analysis and visual
scales are prepared for estimating losses, and field staff deployed to the
targets (farms households, intermediaries).
8. For losses other than bio deterioration at other links of the chain, the
other methods (weighing in and out, etc.) will be used. Field staff are
deployed (farm households, intermediaries).
9. Finally losses are stored in central databases, computed and analyzed;
reports are produced and disseminated.
46
E Annex I: Modified Count and
Weigh Method
E1. INTRODUCTION
This method was already mentioned in a previous report with the final formula
that was obtained by J.A.F Compton; however, the eight steps leading to the
formula were not reproduced. They are detailed in this part of this report.
According to J.A.F. Compton
“The count and weigh (also known as gravimetric) method is a well-known
procedure for assessing weight loss due to pest damage in stored cereals
(Harris and Lindblad, 1978; Boxall,1986). Its relative ease of use, lack of need
for specialized equipment such as moisture meters, and above all, the lack of
need for a ``baseline'' (pre-storage) sample for comparison make it a very
attractive method where resources and time are scarce.
Several refinements to the method have been published (e.g. Boxall, 1986;
Ratnadass and Fleurat-Lessard, 1991) but none has addressed the case where
grain kernels are actually lost or destroyed, as opposed to damaged, by pests. If
such missing grains are not taken into account, the count and weigh method is
likely to underestimate losses. The arrival in Africa of a new pest of stored
maize, Prostephanus truncatus Horn (Coleoptera: Bostrichidae), which reduces
many kernels to powder, highlighted this problem (Keil, 1987) and led
Pantenius (1988) to propose an alternative loss assessment method for maize
cobs damaged by this pest, which involved comparing the average weight of
total grain shelled from cob samples over time.
However, for many types of research, especially on farms, it is not practicable
to use the Pantenius method (it requires a baseline sample, a moisture meter,
47
and also large sample sizes) and the count and weigh method continues to be
widely used.”
The modification to the conventional method is described here; it is for use
with stored maize cobs, and accounts for destroyed and missing grains.
E2. THE METHODS
E2.1 THE CONVENTIONAL COUNT AND WEIGH METHOD AS
REPORTED BY (HARRIS & LINDBLAD)
As a reminder, a sample of cobs is taken from the maize store, using any of the
methods extensively described by (Harris and Lindblad, 1978; Golob, 1981;
Boxall, 1986; Nyambo, 1993). The sampled maize cobs are shelled together
and all the resulting grains are pooled together.
Next two (2) subsamples of 200±500 grains each are taken using a riffle
divider. The grains in each subsample are separated by eye into two groups,
damaged and undamaged. Separation of damaged from undamaged grains is
discussed in Boxall (1986. The damaged and undamaged grains in each group
are then counted and weighed. Percentage weight loss can then be calculated
separately for each subsample using equation (1). The average of the two
subsamples is taken as the weight loss for the sample of cobs.
The formula has been reported as follows: (Harris & Lindblad)
% weight loss = (1)
Where U = weight of undamaged grain,
Nu = number of undamaged grains,
D = weight of damaged grains,
Nd = number of damaged grains.
This formula does not require knowing the value of the mean weight of
undamaged grain.
48
E2.2 THE MODIFIED COUNT AND WEIGH METHOD BY J.A.F.
COMPTON
The modified method consists of the following eight (8) steps:
1. A sample of maize cobs is taken like in the conventional method.
According to experience, 30 cobs have been found to be a convenient
sample size with reasonably precise results.
2. The cobs are shelled one by one, and the number of destroyed and
missing grains is recorded for each cob. These are then summed over
all cobs in the sample to give the total number of destroyed and missing
grains (TND). If desired, cob-related characteristics such as husk cover
and grain type can also be recorded at this point.
For consistency purposes, the criteria used to define ``destroyed grains'' should
be clearly specified and rigorously adhered to. Destroyed grains were defined
as those which were crushed during shelling into fragments smaller than one
third of a grain, or which passed through a 3.35 mm sieve in step 3. All such
fragments must be thrown away to avoid double counting later.
3. The shelled grains from each cob are sieved through a standard sieve
set (3.35/2/0.85 mm mesh were used). If desired, the number and
species of insects on each cob can be recorded at this point.
4. The sieved grains from all the cobs are then pooled. A typical pooled
sample contains 7000±15,000 grains and weighed about 1.5±3.5 kg.
The pooled sample is weighed and the weight recorded to the nearest g.
This is the Final Weight (FW).
5. A riffle divider is used to subdivide the pooled sample several times to
obtain two subsamples containing about 400±600 grains each.
Remaining grains are discarded. The number of grains per subsample
should be increased if there is a high proportion of a damaged grain,
since it is the total number of undamaged grains which primarily
determines precision. A minimum of 50 undamaged grains per
subsample is suggested.
6. The grains in each subsample are separated into two groups, damaged
and undamaged, by eye, in the same way as in the conventional
method.
49
7. For each subsample, the groups of damaged and undamaged grains are
counted and weighed as in the conventional method.
Nd = number damaged grains in subsample,
Nu = number undamaged grains in subsample,
Wd = weight damaged grains in subsample,
Wu = weight undamaged grains in subsample.
8. Percentage weight loss =
(2)
Weight loss is calculated separately for subsamples 1 and 2, and the average of
these two values is taken as the estimated weight loss in the cob sample.
50
F Annex II: Value Chain/Food
Supply Chain Approaches In the literature about PHL studies, it is widely acknowledged that before
starting any kind of loss assessment survey, it is critical to conduct research in
order to any available information that can help plan and undertake the study.
That is similar to the research undertaken in the early stages of the agricultural
census. Next a pilot survey may be conducted with the aim to establish
preliminary estimates of losses with data on their causes. Together, the
research and the pilot survey will examine specific problem points, will help
understand the post-production system, highlight the main points at which
significant losses may arise as well as their causes. Most often than not the
research and the pilot survey are merged together into an operation called
preliminary survey.
The methodologies used by the preliminary survey could be the methodologies
of what has been described in this report as general baseline surveys that may
use sampling or non-sampling techniques.
It is suggested as a methodological option that countries conduct the
preliminary survey include the usage of the mapping techniques of Agriculture
Value Chain/Food Supply Chain for their three most important food grain
crops as defined by FAO.
Some organizations have developed rapid assessment methods for PHL case
studies. While not statistically grounded, these methods have several
advantages. They produce results in a short time span for decision making and
can be carried out even in countries that have no regular agricultural surveys or
crop-cutting surveys. While results are valid for the selected ‘chain’ only at a
given point in time, established methods ensure that a consistent approach is
followed by the organizations in carrying out the various studies.
51
Two methods in particular were developed in 2013-2015, the Food Loss
Assessments: Causes and Solutions Food by FAO’s Rural Infrastructure and
Agro-Industries Division (AGS) and the Rapid Appraisal Method by the GIZ.
The two methods analyze Food Supply Chains (FSC) and Value Chains (VC)
respectively and have several points in common. They are being implemented
in various countries as part of FLW reduction projects. The World Bank and
the United Nations Industrial Development Organization (UNIDO) have also
developed guidelines very briefly reviewed below.
FAO-AGS methodology has been field tested in a number of countries; this
methodology of food loss analysis has the following stated objective:
“The objective of this study is the identification and quantification of the main
causes of food losses in the selected food supply chains, and the analysis of the
measures to reduce food losses on their technical and economic feasibility,
social acceptability and environment-friendliness, leading to concrete
proposals to implement a food loss reduction programme”.
The essence of the FAO-AGS approach (termed 4S) encompasses four
moments described as:
1. Preliminary Screening of Food Losses (‘Screening’). Based on
secondary data, documentation and reports, and expert consultations
(by phone, e-mail, in person) without travel to the field.
2. Survey Food Loss Assessment (‘Survey’). A questionnaire exercise
differentiated for either producers, processors or handlers/sellers (i.e.
warehouse manager, distributor, wholesaler, retailer), complemented
with ample and accurate observations.
3. Load Tracking and Sampling Assessment (‘Sampling’). For
quantitative and qualitative analyses at any step in the supply chain.
4. Solution Finding (‘Synthesis’). Used to develop an intervention
programme for food losses, based on the previous assessment methods.
This supply chain food loss assessment involves the collection of data and their
analysis. Assessments are carried out using qualitative and quantitative field
methods. Subsequently, solutions to food losses will be formulated from the
results and conclusions of the assessment. In step 2 respondents are not
randomly selected, the term survey is used to indicate interview method. In
step 3, sampling means purposive selection. The methodological paper has not
been published yet but is available upon request.
52
The most interesting results are the identification of the points where most
losses occur and an estimation of the percentage volume of each food supply
chain.
GIZ (Deutsche Gesellschaft für Internationale Zusammenarbeit GmbH)
commissioned the design and pilot testing of a Rapid Loss Appraisal Tool
(RLAT) with methodology comparable to the FAO-AGS approach. According
to their (unpublished paper), “The methodology is designed to serve as a pre-
screening tool for narrowing down from the array of functions and possible
loss points along value chains (VC) to prospective leverage points. By doing
so, the tool supports the assessment of needs for more in-depth studies that can
back public and private investment decisions aimed at reducing losses at the
various value chain stages including the development of enabling policy, legal
and infrastructure framework conditions.” The Rapid Assessment Tool was
tested in Ghana in 2014. The approach encompasses the following six major
steps:
1. Desktop Study
Political, (socio)economic and agri-business frame conditions
2. Key Expert Workshop
Analysis of hot spots (critical loss points) along the VC
Validation of desktop study results
3. Stakeholder Workshop
Analysis of hot spots (critical loss points) along the VC
Validation of key expert workshop results
4. Focus Group Meetings
Appraisal on the ground (ground truthing) of workshop results
Verification of loss perceptions of VC operators
5. Key Informant Interviews
Validation/ completion of results of preceding process steps
53
6. Assessment and Presentation of Results
Plausibility check of results at the different process steps
Presentation of aggregated results
The main difference between the two approaches (FAO-AGS and GIZ) is in
the initial stakeholders’ workshop recommended by GIZ.
The World Bank has published guidelines titled: “Using value chain
approaches in agribusiness and agriculture in Sub-Saharan Africa - A
methodological guide”. The Guide includes the definition of value chains, a
description of their structure and a background on using and analyzing value
chains. It reviews existing literature on value/supply chains, including current
theories and applications and discusses individual tools used in value chain
analysis. Finally it summarizes the principles and lessons and suggests several
future steps for achieving further experience in value chain competitiveness.
The United Nations Industrial Development Organization (UNIDO) has
also defined its approach to agro-value chain analysis and development in a
guide titled: “Agro-Value Chain Analysis and Development – The UNIDO
Approach”.
For UNIDO, “A value chain describes the entire range of activities undertaken
to bring a product from the initial input-supply stage, through various phases of
processing, to its final market destination, and it includes its disposal after use.
For instance, agro-food value chains encompass activities that take place at the
farm or rural level, including input supply, and continue through handling,
processing, storage, packaging, and distribution. As products move
successively through the various stages, transactions take place between
multiple chain stakeholders, money changes hands, information is exchanged
and value is progressively added. Macroeconomic conditions, policies, laws,
standards, regulations and institutional support services (communications,
research, innovation, finance, etc.) – which form the chain environment – are
also important elements affecting the performance of value chains.”
Based on this VC/FSC approaches, a number of interesting studies have been
conducted in Sub-Saharan Africa (Ethiopia, Tanzania, Uganda, Zambia, etc.)
that have produced useful insights through the mapping of important crops.
Figure 1 below is a maize value chain mapping in Uganda, while figure 2 is a
mapping for rice.
54
These methodologies are quite easy to implement for developing countries and
should provide useful information for better planning and implementation of
the main PHL survey.
Figure 1 (Mr. Ian DALIPAGIC, Dr. Gabriel ELEPU. 2014)
Urban Retailers Export
Large Millers
Institutions Rural Retailers
Local Traders Medium Millers Wholesalers
Farmers/Farmers’ Group
Regional Level
Local Level
Maize Flour
Maize
Consumers
55
Figure 2 (Mr. Ian DALIPAGIC, Dr. Gabriel ELEPU. 2014)
Consumers
Urban Retailers
Kampala
Wholesalers
Millers Rural Retailers
Export
Local Traders
Farmers/Farmers’ Group
Schelled rice
Unschelled rice
56
G Annex III: Country Papers
on PHL Surveys
G1. INTRODUCTION
Part of this report is based on findings from country experiences. Country
representatives shared their experience on Post Harvest Loss studies with
focus on measurement techniques, information system building and use of
PHL data at the Expert Group Meeting held in Rome in April 2015. A
synthesis of the invited short papers is given in this annex. In particular, the
papers highlight good practices and successful experiences to be used as
inputs to the ongoing research work undertaken by the Global Strategy. The
papers were proposed for short presentation during the expert meeting and
were outlined as follow:
1. Main Post harvest policy concerns in the country (food security, loss
prevention/reduction, etc.) and Strategies/efforts to reduce post-harvest
losses.
2. Post-harvest management institutional set up in the country-Who is
dealing with PHL? (Ministry, Department, Unit).
3. Main players involved in the food supply chain relevant to PHL
measurement (Farmers, Collectors, Processors, etc.), size (Small-scale,
medium-scale, large-scale, commercial actors) and losses suffered at
main levels of food chain (harvest for farmers, storage for retailers,
etc.).
4. Main commodities for which PHL are measured (maize, wheat, rice,
groundnut, etc.).
5. Technologies used by the actors to mitigate post-harvest loss risks.
6. Short description of sources used for providing post-harvest loss
data/information
57
7. In case surveys or studies are conducted, provide a short description of
the last post-harvest survey/study (sample size, sampling design, main
results findings/conclusions, available reports.).
8. Lessons learned and areas of further improvement.
Papers presented by Malawi and Indonesia are summarized below.
G2. MALAWI
Background
Malawi just like many countries in Africa is a victim of post-harvest losses on
crops. Most countries in the African region estimates that the post-harvest loss
could be in the range of between 20 to 40%. Likewise, Malawi has had its
estimation ranging from about 10 to 40%.
As a quantitative and qualitative measure of agricultural production loses, Post-
Harvest Loss (PHL) has an important role in Malawi in deterring food security
levels through the establishment of the country’s food balance sheet. The PHL
levels is also important at macro-economic calculation as some levels of
discount have to be undertaken based on PHL to determine the country’s Gross
Domestic Product (GDP)
At farm level PHL, the farming communities concerns range from food
security, food safety to loss of income. As such, farmers use different
technologies/methods of reducing the risk of PHL. Some studies in Malawi
have revealed that smallholder farmers in Malawi avert PHL risk by being
conscious with the moisture content of their grain form harvest to storage.
Since most storage losses have been attributed to insects such as the large stock
borer, most small holder farmers in the country treat their grain with a number
of different insecticides before storage. Some use their indigenous knowledge
to reduce the risk of PHL.
It is evident that most of the surveys and their subsequent recommendations
have all along concentrated on grain PHL, but I believe there are other areas of
PHL which are hidden and need to be understood in Malawi. In the horticulture
subsector, for example, the PHL levels could be alarming mostly emanating
from post-harvest handling and post-harvest storage of the crops in this
subsector. The country may be experiencing huge hidden PHL in vegetables
when it comes to nutrients because of the preservation systems farmers apply
in their communities. However, this is not a subject of this paper.
58
Strategies to Reduce PHL
To mitigate the post-harvest loses the Government of Malawi in collaboration
with other players in the agriculture sector have deployed a number of
strategies including the following:
Removal of taxes on post-harvest insecticides approved by the
country’s regulatory body;
Distribution and sale of pesticides at a reduced price across the country
using the Farm Input Subsidy Programme window;
Promotion of small metallic silos (though adoption rate is currently an
issue) with some assistance from FAO and other partners;
Promotion of community concrete silos;
Continued research on post-harvest handling by the country’s
agricultural research institutions;
Promotion of warehouse receipt systems and commodity exchange
initiatives; and
Development of focused extension messages
Post-Harvest Management – Institutional Set up
Post-Harvest Loss, as indicated earlier on, is a concern for most players in the
agriculture sector. At farmer level most farmers know the experience they have
or encounter due to PHL. Most of them know crops that are susceptible to
PHL. In case of maize most farmers know that hybrid maize is more
susceptible to storage losses than local and composite varieties (MOA IWD –
FIDP concrete silos report 2014). Thus, the farming communities by default
are amongst the institutions that are dealing with PHL in Malawi. At national
level, the country’s Department of Agriculture Research, Extension and Crops
Development play a crucial role in dealing with post-harvest loss.
Commodities for which PHL are measured
Almost all the surveys that have been conducted in Malawi on post-harvest
losses have been focusing on maize grain considering its role in determining
food security for the country. The country also puts more weight on maize
when calculating inflation beside the grain’s social-economic importance in the
local communities.
59
Technologies used to mitigate Post Harvest Loses in Malawi
As indicated above most farmers in Malawi know what to do to arrest
postharvest losses especially on maize. Some specific studies on specific
projects have shown that most smallholder farmers in the project areas follow
recommended pre- and post-harvest practices to mitigate PHL. For instance
FIDP Concrete Silos Assessment Report of November 2014 revealed that
40.4% of the farmers in programme were treating their maize with storage
pesticides to arrest PHL before putting the grain in the silos. The report further
indicated that farmers use harvesting period, drying and timely application of
chemicals as means to reduce PHL. Thus, harvesting in time, In most cases,
PHL in Malawi maize grain could be attributed to grain cracking due to over
drying or use of in appropriate technologies when shelling, weight loss due to
over respiration, rodents and insects’ infestation or damage and contamination
with mycotoxins caused by moulds or bacteria. However, the most common
post- harvest losses are as a result of insect attacks. The most common insects
include: termites, Prostephanus truncatus, Sitophilus zeamais, Sitophilus
oryzae, Tribolium castaneum, Ephestia cautella.
Out of the above insects, Prostephanus truncatus (Larger grain borer) is the
most predominant and destructive insect in Malawi. In this respect, most of the
technologies used in curbing the problem target this insect. Chemicals such as:
Super Guard dust and EC, Actellic Super dust and EC, Shumba Super dust and
Wivokil dust are commonly applied to reduce PHL. The other technologies
include the use of small metallic silos and the community concrete silos.
Post-Harvest Loss Studies in Malawi
The need to understand the extent to which the economy losses its crops after
harvesting became clear towards the end of 2000. As such, the Ministry of
Agriculture, Irrigation and Water Development in Malawi decided to conduct a
post-harvest losses survey in 2009/10 season with a view to come up with
reliable and verifiable estimates of post-harvest losses (PHL) for the country.
The decision was made in the wake of conflicting estimates of post-harvest
losses that were being flashed out by various stakeholders ranging from 10% to
40% on maize grain.
Realizing that there were considerable variations between the findings of the
2009/10 post-harvest loss study and estimates people were flashing on PHL for
the country, the Ministry of Agriculture, Irrigation and Water Development in
joint collaboration with the Food and Agricultural Organization of the United
60
Nations decided to conduct a similar study in 2010/11 season to ascertain and
confirm the previous study’s findings.
Survey/Study Objective
The main objective of the survey was to provide reliable post-harvest loss
estimates for discounting crop production figures and subsequently be able to
analyze the country’s food gap and calculate its Gross Domestic Product with
confidence.
The other objective of the study was to provide reliable statistics on post-
harvest losses to inform policy formulation and decision making processes in
designing strategies to reduce post-harvest losses in Malawi.
Specific Survey/Study Objectives
Specifically the study was designed to achieve three objectives:
Provide the Government of Malawi with the post-harvest loss for maize in 2010/2011
Confirm the previous loss of 7.6 percent from 2009/2010 study
Compare between the conventional loss assessment method to the rapid loss assessment
Study Methodology
Visual Scales for cobs
Cobs were deliberately infested with weevils and cultured in a laboratory. The
cobs were weighed and high resolution pictures were taken at different levels
of infestation. In the process six classes were identified and percentage weight
loss (damage coefficient) was assigned to each scale. The damage coefficients
for the six classes were: class 1 - 0% loss, class 2 – 8.8%, class 3 – 13.3%,
class 4 – 22.6%, class 5 – 38.7% and class 6 – 55.6%. Using visual impression,
enumerators matched the farmers’ cobs with the six pictures. The percentage
weight loss assigned to the picture with corresponding appearance was entered
as the weight loss for the cob.
Standard Chart
To estimate weight loss for maize stored in grain form, a standard chart was
used. The enumerator was asked to randomly select 3 separate samples of 100
61
grain each from the farmers’ maize. The enumerator then placed the 100 grains
in a liter plate and physically counted the damaged grain. The process was
repeated for the 3 samples and an average number of damaged grain per 100
grain was established. The number of damaged grain was read off against a
predetermined regression chart to find the percentage weight loss.
Conventional loss assessment method
The study involved collection of maize samples. Five farmers were identified
for sample collection from each district. Sampling procedure involved
collection of 10 cobs of maize or 1 kg of shelled grain from each farmer. The
grain samples were collected at random points from either the granaries or bags
and put into polythene bags. Properly written labels indicating date of
sampling, farmers’ name and district were put into the bags before tying them
tightly. The samples were brought to the Crop Storage Laboratory at Bvumbwe
Agricultural Research Station. In the laboratory, the samples were subjected to
assessment. Bags were opened, insects sieved off, moisture content and the
standard Volume Weight (SVW) of the grain determined. Storage insect pests
were counted and recorded.
Household Questionnaire
To triangulate the quantitative data, a household questionnaire was
administered to the farmers. While very few questions were targeted to the
farmers when estimating losses using visual scales and standard graph forms,
the household questionnaire was used to solicit farmers’ perceptions and what
the farmer feels are the causes of post-harvest losses and how they can be
contained. However, for in-depth understanding of farmers’ perceptions and
dynamics underpinning storage losses there is need to conduct a fully-fledged
Participatory Rural Appraisal (PRA).
Main Findings
The results indicate that using the rapid loss assessment method, post-harvest
loss for maize in 2010/2011 was 8.4% and 9.7% before taking into account
consumption. Using the conventional method the loss was at 15.7%, giving a
difference of 6 to 7 percent points between the rapid method and the
conventional method. The difference may be attributed to the differences in
time of analysis among other things; in the rapid method the enumerator
instantly gives the results while in the conventional one the enumerator had to
collect the samples and these were then sent to Bvumbwe Research Station for
62
analysis, a process that would take 2 to 3 weeks during which time further loss
takes place in the samples.
Plans for the Next Post Harvest Losses Study
There is a growing demand for updated figures on the Post-Harvest Losses by
some stakeholders across the country. Some stakeholders feel that the findings
in the last assessment of 2011 cannot be relied upon considering a number of
dynamics and emerging issues in the sector. They argue that since the last
national wide survey Government conducted, there have been a number of
studies that have been done but they have mainly produced project-level
statistics on PHL, which to some extent tend to differ with the nation-wide
survey results.
In addition, stakeholders have indicated the need to also look at the PHL in
some of the major crops in the country considering that all previous studies
have focused only on one crop, maize.
Some have further urged that the surveys conducted so far in Malawi have
some methodological limitations, particularly in terms of coverage and survey
design. A proper survey with a sound methodology validated by key
stakeholders is therefore urgently needed for proper assessment of food
availability and also as a baseline to monitor the outcome of some of the
important investments the country is making in reducing PHL. Furthermore,
the country needs to conduct a proper baseline survey that may not require
conducting the PHL surveys annually, rather on a four-year interval, since
PHLs are structural variables and do not change rapidly.
With the above, the Ministry has included PHL study in its national
Agricultural Statistics Strategic Master Plan which was launched last year. In
view of this development, the Ministry in collaboration with FAO Malawi has
been working on preparations for the next Malawi Post harvest losses study.
FAO will provide technical support to the Ministry, just as they did during the
2011 study. Terms of Reference (ToRs) for the study have been drafted and a
budget line provided the Multi-Donor Trust Fund 2015/16 financial year.
Conclusion
Both government and the farmers realize the need to understand the country’s
post-harvest losses. At farm household level most farmers have had
experiences of PHL as such some surveys have shown that about 53% of the
farming households rely on storage pest chemicals to reduce post-harvest
63
losses in Malawi. The 2011 PHL survey found out that about 77% of farmers
in Malawi believe that the use of chemicals is the best method to arrest post-
harvest losses seconded by harvesting early (56%) as Large Grain Borer attack
starts right in the field.
The Government of Malawi, on the other hand, has put in place some policy
instruments to reduce PHL. Such instruments include inclusion of crop storage
chemicals in the Farm Inputs Subsidy Programme and the promotion of use of
improved storage facilities other than the traditional granaries made out of
bamboos, wood or grass which acts as breeding areas for insects.
Although the 2011 survey results indicated that only 43% of the farmers
interviewed receive extension messages on how to protect their produce from
post-harvest losses, there are some considerable efforts by government
specifically the Ministry of Agriculture, Irrigation and Water Development and
its extension agencies to put messages across to curb the post-harvest losses for
the country.
G3. INDONESIA
General
The Indonesian government has some policies for increasing its paddy
production in 2015. Some main program have been implementing such as
building new dam for irrigation, open new land of forest for paddy field, giving
tractors to the farmers, and others. The important program is to reduce paddy
harvest losses. Ministry of Agriculture is responsible to reduce the post-harvest
loss. It has given MICO (mini combine harvester) for harvesting small land
area and threshing machine to the farmers. The post-harvest loss surveys for
paddy have been carried out and the results have been used for making some
policies. The surveys have focused on actors who were the actors in rice supply
chain to post harvest loss, i. e farmers, traders, and operator of milling
The Post-Harvest Survey
BPS has collaborated with Ministry of Agriculture to conduct surveys as
shown in table below.
64
Table 1: Type of Survey
Type of survey
Number
of provinces
Number
of samples
Respondent
Year 2005-2006
1. Loss at transporting grain 8 1 680 Farmer and trader
2. Loss at grain storage 8 840 Farmer
3. Loss at transporting rice 8 1 680 Farmer and trader
4. Loss at rice storage 8 840 Trader
Year 2005-2007
1. Loss at harvest 22 4 230 Farmer
2. Loss at threshing paddy 22 4 230 Farmer
3. Loss at drying grain 22 4 230 Farmer and operator of milling
4. Loss at milling grain 22 4 230 Operator of milling
Year 2012
1. Loss in drying 30 29 351 Farmer
2. Yield milling grain 30 17 044 Operator of milling
Sampling design of the 2012 survey
Loss in Drying
Two stage sampling method was adopted. First stage, from frame (a list of
district which has potential paddy land in each province) was selected a
number of districts. In 30 provinces, 346 districts were selected. Second stage,
in every selected district was drawn a number of farmers systematically. The
selected farmers must be asked when they will harvest their paddy and where
are locating their paddy land. Some plots then will be observed.
Data Collection Method
The well trained enumerators visited the respondent to measure grain moisture
of content and weight the selected grain which minimum weight was 200
kilogram. He must do them two times. First, he must do in the selected wetland
paddy after plot (2.5 meter x 2.5 meter) harvesting. He then must do again in
drying location where farmer usually does for grain drying.
65
Calculation of weight of grain
In drying process, there will be decreasing of moisture content in grain and
losses grain physically because of eaten by poultry and losses.
Grain loss = {(100 – KA1) x BG1} – {(100-KA2) BG2} x 100 % / (100-KA1)
BG1
Reduction of moisture content = 100 – BG2 + {[(100 - KA1) x BG1 – (100-
KA2) x BG2]/ (100- KA1)} x 100% / BG1
Where,
BG1 = Grain weight before drying
BG2 = Grain weight after drying
KA1 = Moisture content in grain before drying
KA2 = Moisture content in grain after drying
Findings
1. Reduction of moisture content in drying 2012 was 10.79 percent (rse =
1.00). In 2005-2007 it was 10.71.
2. Loss grain in drying was 6.09 percent (rse = 2.90). In 2005-2007 it was
3.27 percent
Grain losses will be influenced by drying location:
Table 2: Drying Location
Drying Location Reduction of moisture content Grain loss
1.Around farmer’s house 10.76 % 6.22 %
2.Side of the road 10.65 % 4.69 %.
Yield Milling Grain
Sampling design of the 2012 survey
Two stage sampling method was adopted. First stage, from frame (a list of
district which has potential paddy land in each province) was selected a
number of districts. The selected districts for this survey is the same with the
66
loss drying survey. Second stage, in every selected district, large scale milling
was take all and for small scale milling was selected systemically.
Data Collection Method
The well trained enumerators visited the rice milling to measure grain moisture
of content and weight the selected grain. Minimum grain weight to be
measured were 1000 kilogram, 300 kilogram, and 100 kilogram for large scale,
medium scale, and small scale rice milling respectively. After milling, they
must measure moisture content in rice and weight of rice.
Calculating of the yield milling = (B2/B1) x 100 %
Finding
The yield milling in 2012 was 62.85 percent (rse = 0.52). In 2005 it was 62, 74
percent
Finding of the 2005-2007 Surveys:
1. Loss of grain at harvest was 1.20 percent
2. Loss at threshing grain was 0.18 percent
3. Loss at transporting grain was 0.91 percent
4. Loss at grain storage was 1.08 percent
5. Loss at transporting rice was 0.62 percent
6. Loss at rice storage was 0.31 percent
67
H Annex IV: General Survey
Framework (protocol) for
Field Testing This field test survey framework applies best to farms that undertake manual or
partly manual harvest and post-harvest processes. It is less adapted to measure
losses for farms mostly relying on mechanical processes. For example, in large
farms using combined harvesters-threshers, harvesting, threshing and even
cleaning are grouped into one operation. To measure these losses, specific field
experiments exist. They will not be covered in this report but will be addressed
in other publications (Guidelines, in particular). The different steps of the field-
test protocols for measuring post-harvest losses are described below:
H1 - SURVEY RELATED ISSUES
The idea behind this protocol is for any given GS target country to conduct a
reasonably good pilot PHL study based on sound statistical methodology to test
the recommended methods or measure cost-effectiveness of existing method
with the opportunity to test different methodological options as already
mentioned in previous reports. This is for countries having sound experience in
the conduct of annual agriculture production surveys with crop cutting exercise
to estimate crop yield and production.
H1.1 OBJECTIVES
The main objective of the study is to allow for testing the statistical
methodological options that are related to estimating losses at different levels
of the food supply chain (harvesting, drying, threshing/shelling,
cleaning/winnowing, transport, etc.) at farm, intermediary, and warehouse
levels and identifying the most cost-effective ones.
68
H1.2 INSTITUTIONAL SET-UP
A Technical Committee between the country's Central Statistics Agency (CSA)
and the relevant sector ministries (agriculture, etc.) and agronomic research
institutions could be established to help with the implementation of the various
tests and assist in all aspects of the survey. For example, the Survey/Field
Section of the CSA would be responsible for providing the sampling frames,
the development of a field system for the collection of data (in collaboration
with other members of the Technical Committee), training of field staff and
overseeing field operations. The Department of Planning of relevant sector
ministries would be responsible for the preparation of a final report presenting
the results of the survey; the research institutions (agronomic research,
university, agriculture colleges, etc.) would be responsible for the planning and
implementation of experimental designs/trials.
In order to conduct the pilot survey, the responsibilities of the unit in charge
would include:
development work on all aspects of the survey, including specification
of output requirements, questionnaire design, sample design and sample
selection;
development and implementation of the computer processing system;
preliminary tabulation.
In cases where a general institutional mechanism on agricultural statistics
already exists, the Technical Committee on post-harvest losses should be
directly linked to this mechanism. It could for example constitute a sub-
committee that would be convened on a regular basis to discuss issues of
relevant for PHL statistics.
H1.3 COVERAGE
Countries have different administrative make-ups encompassing regions, that
might be divided into provinces; provinces themselves might be divided into
districts, and districts subdivided into other entities such as
localities/villages/enumeration areas.
The proposed tests should cover one district or census enumeration area to
cover a number of small to medium agricultural holdings; the preliminary
Chain Analysis should be factored in to help improve on the coverage.
69
H1.4 SCOPE
Scope refers to the data/information items to be collected during the survey.
The proposed survey items refer to specific information on survey
characteristics. In this particular exercise, they are grouped into the following
questionnaires by type of the actor involved and supply chain level. It is
proposed that field tests will cover the following: On-farm PHL of small-scale
farmers (operations and storage)
Traders (storage and transportation) Millers Storage (village level, public
stocks).
Farm level: Q00 – Village Questionnaire
1. Identification particulars
Name Code
Country
State/Region/Province/District
Village/Ea/Psu
Date
2. Facilities (storage, market)
Numbers available / in use
Grain market
Public warehouse
Modern (metal/concrete) storage
structures
Intermediate (improved) storage
Traditional storage
3. Mechanical equipment in use
Numbers
Tractors
Trailers
Pump
Trucks
Harvesters
Threshers
Others (specify)
70
Farm level: Q01 Agricultural Holding Operational Questionnaire
1. Identification particulars
Name Code
Country
State/Region/Province/District
Village/Ea/Psu
Year
Holder
Size Class
2. Assets
Code
Storage facilities
Tractors
Trailers
Pumps
Harvesters
Trucks
Threshers
Others
3. Farm operations
Operation Method of operation Equipment used Code
Ploughing
Harvesting
Threshing
Transport
Others (specify)
71
4. Crops grown
Crop Field Serial
No.
Crop Area
(ha)
Threshing Harvesting
Expected
date
Method
manual/mech
Expected
date
Method
manual/mech
5. Marketing of produce (last year)
Grain
(For
example
maize)
Code
1
Grain
(For
example
millet)
Code
2
Grain
(For
example
rice)
Code
3
Grain
(For
example
sorghum)
Code
4
Quantity produced (kg)
Quantity sold (kg)
Distance to market (km)
Mode transport to market
Mode packaging for
market
Grain losses transport
market
Grain disposal in kind
(gift, etc.)
Quantity stored (kg)
Period of storage (months)
Losses in storage (kg)
Causes of losses
Measures taken against
losses
72
Farm level: Q03 Harvesting Loss Questionnaire
1. Identification particulars
Code
Country
State/Region/Province/District
Village/Ea/Psu
Season
Year
Holder
Size Class
Crop
Area under crop
Field Serial No.
2. Field particulars
Code
Soil type
Size of field
Variety
Amount of seed
Irrigation
Irrigation source
Planting date
Harvesting date
Manual harvesting
Mechanical harvesting
3. Mechanical harvesting results
Actual Per ha Moisture content %
Yield of grain of whole field
Picked grain yield plot 1
Picked grain yield plot 2
73
4. Manual harvesting results
Yield of grain (kg) Moisture content %
Crop cutting Picked
Plot 1
Plot 2
Farm level: Q04 – Stacking/Stooking Loss Questionnaire
1. Identification
Code
Country
State/Region/Province/District
Village/Ea/Psu
Year
Holder
Season
Size Class
Crop
2. Stacking/Stooking losses
Weight of sampled grain before stacking (kg) Code
Moisture content before stacking
Place where stacking was done (inside the field, outside the
field.)
Any loss due to birds, rodents, animals (Yes/No)
Stacking/stooking process duration (days)
Moisture content of sample after stacking
Weight grain after stacking process
Weight converted standard moisture content – before stacking
Weight converted standard moisture content – after stacking
Weight converted standard moisture content – difference
74
Farm level: Q05 - Threshing/shelling loss Questionnaire
1. Identification
Code
Country
State/Region/Province/District
Village/Ea/Psu
Year
Holder
Season
Size Class
Crop
2. Threshing/shelling particulars
No. of bundles sampled
Type of threshing floor
Method of threshing
Weight of grain after threshing
Weight of grain hand-stripped from straw
1.Moisture content grain normal threshed
2.Moisture content grain hand-stripped
Weight1 converted standard moisture content
Weight2 converted standard moisture content
Weight damaged grain from sample of threshed-
2kg
Farm level: Q06 - Cleaning/winnowing loss Questionnaire
1. Identification
Code
Country
State/Region/Province/District
Village/Ea/Psu
Year
Holder
Season
Size Class
Crop
75
2. Cleaning/winnowing losses
Weight of sampled grain before cleaning (kg) Code
Method of cleaning/winnowing
Weight of grain after cleaning (kg)
Weight of grain from left-over material
Farm level: Q07 - Drying Loss Questionnaire
1. Identification
Code
Country
State/Region/Province/District
Village/Ea/Psu
Year
Holder
Season
Size Class
Crop
2. Drying losses
Weight of sampled grain before drying (kg) Code
Moisture content before drying
Method of drying (mechanical/manual)
Place where drying was done (courtyard, road, etc.)
Any loss due to birds, rodents, animals (Yes/No)
Drying process duration (days)
Moisture content after drying
Weight grain after drying process
Weight converted standard moisture content – before drying
Weight converted standard moisture content – after drying
Weight converted standard moisture content – difference
76
Farm level: Q08 - Storage Loss Questionnaire
1. Identification
Code
Country
State/Region/Province/District
Village/Ea/Psu
Year
Holder
Season
Size Class
Crop
2. Storage losses
Storage system (traditional, intermediate, modern) Code
Storage capacity
Storage mode (bulk/bags or loose)
Storage duration
Weight grain stored at beginning w1
Weight grain stored at end w2
Difference between w1 and w2
Interval of periodical observations (monthly, weekly, etc.)
Periodical observations – weight grain in storage
Periodical observations – weight grain taken out (on-farm fortnightly visits)
Periodical observations – weight grain remaining (on-farm fortnightly visits)
Moisture content of grain (on-farm fortnightly visits)
Relative humidity (on-farm fortnightly visits)
Temperature (on-farm fortnightly visits)
Loss due to rodents, if any (on-farm fortnightly visits)
Control measures adopted (on-farm fortnightly visits)
Observations on 1000 grains taken by sampling probe – weight of 1000 grains (lab)
Observations on 1000 grains taken by sampling probe – no. & weight of damaged grains (lab)
Observations on 1000 grains taken by sampling probe – no. & weight of damaged grains by
mites, insects (lab)
Observations on 1000 grains taken by sampling probe – no. & weight of damaged grains by
micro-organisms (lab)
No. & weight of excreta (lab)
No. & weight of dead bodies of insects, etc. (lab)
Acidity level (lab)
Toxicity level (lab)
77
Farm level: Q09 - Transport Loss Questionnaire
1. Identification
Code
Country
State/Region/Province/District
Village/Ea/Psu
Year
Holder
Season
Size Class
2. Transport loss
Wheat/Rice/Sorghum/Maize/Millet Code
Weight of grain sold
Mode of transport (head of person, etc.) - field to
threshing floor
Mode of transport (head of person, etc.) - threshing
floor to drying place
Mode of transport (head of person, etc.) - threshing
floor to cleaning place
Mode of transport (head of person, etc.) - farm to
storage (case manual harvesting, threshing, drying,
etc.)
Mode of transport (head of person, etc.) - farm to
storage (case mechanical operations)
Mode of transport (head of person, etc.) - others
Before/After/Difference
Weight sampled produced transported – field to
threshing floor (bundles)
Weight sampled produced transported – threshing
floor to drying place (grain)
Weight sampled produced transported – threshing
floor to cleaning place (grain)
Weight sampled produced transported – cleaning
place to storage (grain)
Weight sampled produced transported – farm to
storage (grain) (in case of mechanical operations)
Weight sampled produced transported – others
Before/After/Difference
78
Wheat/Rice/Sorghum/Maize/Millet Code
Moisture content transported grain (in lab) –
threshing floor to drying floor
Moisture content transported grain (in lab) –
threshing floor to cleaning place
Moisture content transported grain (in lab) –
cleaning place to storage
Moisture content transported grain (in lab) – farm
to storage (mechanical operations)
Moisture content transported grain (in lab) – other
if any
Total loss due to transport at farm level
Farm level: Q10 - Processing Loss Questionnaire
1. Identification
Code
Country
State/Region/Province/District
Village/Ea/Psu
Year
Holder
Crop and Season
Size Class
2. Processing loss (crop-wise)
Code
Weight of sampled grain (wheat, sorghum, maize, millet) - before
Weight of sampled grain (wheat, sorghum, maize, millet) - after
Weight of sampled grain (wheat, sorghum, maize, millet) - difference
Weight of sampled paddy
Moisture content
Mode of processing (manual/mechanical)
Weight of products – head grain
Weight of products – broken grain
Weight of products – chalky grain
Weight of products – brown grain
Weight of products – other
Loss due to processing
79
Farm level: Q11 – Packaging/handling Loss Questionnaire
1. Identification
Code
Country
State/Region/Province/District
Village/Ea/Psu
Year
Holder
Crop and Season
Size Class
2. Packaging/Handling loss
Rice/wheat/sorghum/millet/maize Code
Modes of packaging (gunny bags/plastic bags, loose,
etc.)
Weight of grain handled
Loss of grains due to packaging and handling
Intermediaries level: Q12 – Market Particulars Questionnaire
1. Identification
Code
Country
State/Region/Province/District
Year
Name of market
Address of market
Size of market
2. Market general information
Code
Whether connected by rail, road, motorway, others
Quantity of grain handled
Rice/wheat/sorghum/millet/maize
Total intake/handling of grains
80
Intermediaries level: Q13 – Traders Particulars Questionnaire
1. Identification
Code
Country
State/Region/Province/District
Year
Name of market
Name of trader
Address of trader
Is wholesaler or retailer
Size of market
2. Business details
Code
Annual turnover in weight (for all crops)
Rice/wheat/sorghum/millet/maize
Quantity procured from different sources - Producer
Quantity procured from different sources – Govt.
warehouse
Quantity procured from different sources - Market
Annual turnover
Transport – Mode of transport
Transport – Loss during transport
Rice/wheat/sorghum/millet/maize
Storage – mode of storage
Storage – quantity stored
Storage – loss due to rodents
Storage – loss due to dampness
Storage observations (periodically) on sample taken
with sampling probe (1000 grains) - weight
Rice/wheat/sorghum/millet/maize
Storage observations (periodically) on sample taken
with sampling probe (1000 grains) – moisture
content
81
Code
Storage observations (periodically) on sample taken
with sampling probe (1000 grains) – relative
humidity
Storage observations (periodically) on sample taken
with sampling probe (1000 grains) – no. & weight of
damaged grains by insects
Storage observations (periodically) on sample taken
with sampling probe (1000 grains) - no. & weight of
damaged grains by mites
Storage observations (periodically) on sample taken
with sampling probe (1000 grains) - no. & weight of
damaged grains by micro-organisms
Storage observations (periodically) on sample taken
with sampling probe (1000 grains) - no. & weight of
excreta
Storage observations (periodically) on sample taken
with sampling probe (1000 grains) - no. & weight of
dead bodies of insects, etc.
Acidity (periodically)
Fat (periodically)
Toxicity (periodically)
Control measures taken
Total loss in storage
Processing – loss due to processing if done by trader
Packaging/Handling – mode of packaging – gunny
bags
Packaging/Handling – mode of packaging – plastic
bags
Packaging/Handling – mode of packaging – loose
Packaging/Handling – mode of packaging – other
Loss due to packaging & handling
Distribution Rice/wheat/sorghum/millet/maize
Mode of distribution – in bags
Mode of distribution – loose
Mode of distribution – others
Loss due to damage to container
Loss due to moisture content
Loss due to pilferage
82
Code
Loss due to handling
Loss due to rain
Loss due to others (specify)
Total loss due to distribution
Transactions information (status at each fortnightly
visit)
Interval (week, fortnight, month, etc.)
Stock (in kg) at the beginning
Purchase (in kg)
Sale (in kg)
Stock (in kg) at the end
Loss (in kg)
Total loss (transport, storage, packaging and
handling, distribution)
Intermediaries level: Q14 – Mill Particulars Questionnaire
1. Identification
Code
Country
State/Region/Province/District
Year
Name of mill
Address of mill
Size class of mill
2. Particulars of mill
Rice/wheat/sorghum/millet/maize Code
Annual turnover
Point of purchase - farm
Point of purchase - market
Point of purchase – Govt. depot
Procedure of weighment (specify)
Sale outlet
Transportation
Transportation – Loss during transport due to
spillage
83
Rice/wheat/sorghum/millet/maize Code
Transportation – Loss during transport due to
pilferage
Transportation – Loss during transport due to
handling
Transportation – Loss during transport due to
moisture content
Transportation – Loss during transport due to
relative humidity
Transportation – Loss during transport due to rain
Transportation – Loss during transport due to other
Total loss due to transportation
Storage
Storage system – weight of grain stored (kg) -
traditional
Storage system – weight of grain stored (kg) -
intermediate
Storage system – weight of grain stored (kg) -
modern
Storage system – weight of grain stored (kg) - total
Storage capacity - traditional
Storage capacity - intermediate
Storage capacity - modern
Storage capacity - total
Weight of grains kept in - bags
Weight of grains kept in - bulk
Weight of grains kept in - loose
Weight of grains kept in - total
Duration of storage system (days/months) weight of
grain stored (in kg)
Loss due to rodents
Loss due to dampness
Storage observations (periodically) on sample taken
with sampling probe (1000 grains)
Rice/wheat/sorghum/millet/maize
Interval of periodical observations (weekly,
fortnightly, monthly, etc.)
Storage observations (periodically) on sample taken
with sampling probe (1000 grains) - weight
84
Rice/wheat/sorghum/millet/maize Code
Storage observations (periodically) on sample taken
with sampling probe (1000 grains) – moisture
content
Storage observations (periodically) on sample taken
with sampling probe (1000 grains) – relative
humidity
Storage observations (periodically) on sample taken
with sampling probe (1000 grains) – no. & weight of
damaged grains by insects
Storage observations (periodically) on sample taken
with sampling probe (1000 grains) - no. & weight of
damaged grains by mites
Storage observations (periodically) on sample taken
with sampling probe (1000 grains) - no. & weight of
damaged grains by micro-organisms
Storage observations (periodically) on sample taken
with sampling probe (1000 grains) - no. & weight of
excreta
Storage observations (periodically) on sample taken
with sampling probe (1000 grains) - no. & weight of
dead bodies of insects, etc.
Acidity (periodically)
Fat (periodically)
Toxicity (periodically)
Control measures taken
Total loss in storage
Processing
Processing method
Total weight of product before processing
Total weight of grain after processing
Loss due to processing
Other observations rice processing – moisture
content
Other observations rice processing – weight of head
grain
Other observations rice processing – weight of
broken grain
Other observations rice processing – weight of
brown grain
Other observations rice processing – weight of
chalky grain
Other observations rice processing – other
85
Rice/wheat/sorghum/millet/maize Code
Packaging/Handling
Packaging/Handling – mode of packaging – gunny
bags
Packaging/Handling – mode of packaging – plastic
bags
Packaging/Handling – mode of packaging – loose
Packaging/Handling – mode of packaging – other
Loss due to packaging & handling
Distribution Rice/wheat/sorghum/millet/maize
Mode of distribution – in bags
Mode of distribution – loose
Mode of distribution – others
Loss due to damage to container
Loss due to moisture content
Loss due to pilferage
Loss due to handling
Loss due to weighment
Loss due to rain
Loss due to others (specify)
Total loss due to distribution
Transactions information (status at each fortnightly
visit)
Interval (week, fortnight, month, etc.)
Stock (in kg) at the beginning
Purchase (in kg)
Sale (in kg)
Stock (in kg) at the end
Loss (in kg)
Total loss (transport, storage, packaging and
handling, distribution)
86
Intermediaries level: Q15 – Mill Particulars Questionnaire
1. Identification
Code
Country
State/Region/Province/District
Year
Name of warehouse
Address of warehouse
Size class of warehouse
2. Particulars of warehouse
Rice/wheat/sorghum/millet/maize Code
Annual turnover
Point of purchase - farm
Point of purchase - market
Point of purchase – Govt. depot
Sale outlet
Transportation
Weight of grain transported to warehouse
Weight of grain received at warehouse
Transportation – Loss during transport due to
spillage
Transportation – Loss during transport due to
pilferage
Transportation – Loss during transport due to
handling
Transportation – Loss during transport due to
moisture content
Transportation – Loss during transport due to
relative humidity
Transportation – Loss during transport due to rain
Transportation – Loss during transport due to other
Total loss due to transportation
Storage Rice/wheat/sorghum/millet/maize
Type of storage (specify)
Storage capacity
Weight of grain stored
Storage system – weight of grain stored (kg) - bags
87
Rice/wheat/sorghum/millet/maize Code
Storage system – weight of grain stored (kg) - bulk
Storage system – weight of grain stored (kg) - loose
Storage system – weight of grain stored (kg) - total
Storage duration (days/months, etc.)
Loss due to rodents
Loss due to dampness
Storage observations (periodically) on sample taken
with sampling probe (1000 grains)
Rice/wheat/sorghum/millet/maize
Interval of periodical observations (weekly,
fortnightly, monthly, etc.)
Storage observations (periodically) on sample taken
with sampling probe (1000 grains) - weight
Storage observations (periodically) on sample taken
with sampling probe (1000 grains) – moisture
content
Storage observations (periodically) on sample taken
with sampling probe (1000 grains) – relative
humidity
Storage observations (periodically) on sample taken
with sampling probe (1000 grains) – no. & weight of
damaged grains by insects
Storage observations (periodically) on sample taken
with sampling probe (1000 grains) - no. & weight of
damaged grains by mites
Storage observations (periodically) on sample taken
with sampling probe (1000 grains) - no. & weight of
damaged grains by micro-organisms
Storage observations (periodically) on sample taken
with sampling probe (1000 grains) - no. & weight of
excreta
Storage observations (periodically) on sample taken
with sampling probe (1000 grains) - no. & weight of
dead bodies of insects, etc.
Acidity (periodically)
Fat (periodically)
Toxicity (periodically)
Control measures taken
Total loss due to storage
Processing Rice/wheat/sorghum/millet/maize
Loss due to processing if the processing done at
warehouse
88
Rice/wheat/sorghum/millet/maize Code
Packaging/Handling
Packaging/Handling – mode of packaging – gunny
bags
Packaging/Handling – mode of packaging – plastic
bags
Packaging/Handling – mode of packaging – loose
Packaging/Handling – mode of packaging – other
Loss due to packaging & handling
Distribution Rice/wheat/sorghum/millet/maize
Mode of distribution – in bags
Mode of distribution – loose
Mode of distribution – others
Loss due to damage to container
Loss due to moisture content
Loss due to pilferage
Loss due to handling
Loss due to weighment
Loss due to rain
Loss due to others (specify)
Total loss due to distribution
Transactions information (status at each fortnightly
visit)
Interval (week, fortnight, month, etc.)
Stock (in kg) at the beginning
Grain received (in kg)
Grain dispatched (in kg)
Stock (in kg) at the end
Loss (in kg)
Total loss (transport, storage, packaging and
handling, distribution)
H1.5 SAMPLING METHOD
A stratified two-stage random sampling design could be used. The estimation
formulas that are presented at the end of this annex apply to this type of design.
A sample of Primary Sampling Units (PSUs) should be selected at the first
stage from the Frame to be provided by the country's Central Statistics Agency
or relevant Ministry. A sample of agricultural holdings should then be selected
89
within each of the sampled PSUs. For each selected holding, fields would then
be listed to allow selection for crop cutting experiment.
For countries that are interested into using a two-stage two-phase sampling
with ratio or regression estimation techniques, there are several research
documents published by the Global Strategy that elaborate on that and that
provides the formula on how to use them. They will not be repeated here. The
reader is invited to refer to “Measuring Crop Area and Yield under Pure Stand,
Mixed and Continuous Cropping: Findings from the Field Tests in three
countries” (Global Strategy, December 2016).
H1.6 DATA COLLECTION METHOD
For the purpose of testing, the survey should use a face to face interview
method. But it should move away from the traditional pencil and paper process
for a newer approach of Computer Assisted Personal Interview (CAPI)
method. Each interviewing team of enumerators should be given a laptop
computer (or, better, tablets) with the software application (CSPRO from the
US Census Bureau, Survey Solutions from the World Bank or any other
relevant application) installed which could load the questionnaire. The tablets
might be cheaper, and more portable; in addition, they have built-in GPS with
map applications suitable for area sampling of agriculture fields. The
interviewer then asks the questions as in the questionnaire from the respondent
and then records the responses using drop down menus in the application. The
skip patterns and most consistency checks should already programmed into the
application and those in turn, should guide the interviewing process.
Agricultural holdings and the agricultural holder (as defined) should be
identified and data relevant to the holding collected from the holder. One of the
important aspect of the data collection exercise during the survey is the
measurement of land area. Another one is the estimation of the production
based on the yields. Both of this required information would be collected by
subjective and objective methods. The subjective method is to ask the holder to
estimate the area, the production and losses. The objective method is to
measure the area objectively using appropriate instruments and also to carry
out the crop cutting exercise to measure the yields under the important crops
for the purpose of estimation of loss and production.
90
H1.7 SURVEY PERIOD
The field work of the pilot survey should be launched during the main cropping
season, depending on the country crop calendar. Within this specified time, all
required data for the given agricultural year will be collected in two visits –
visit I, planting period and visit II, harvesting period.
H1.8 TIME REFERENCE
Different time reference is used depending on various characteristics. The time
reference for area and production should refer to the day of the enumeration.
However, the time reference for agricultural practices should be the
agricultural season. The reference period for other data items mentioned in the
questionnaires should be the last twelve months.
H1.9 FIELD ENUMERATION
Enumerators should operate in teams of at least two (2), and there should be at
least two visits (as mentioned already) to the selected holdings:
Visit I should take place early during the cropping when the plants are
few centimetres tall to allow easy setting up of the yielding plot.
Visit II should take place when the crops are ready to be harvested for
the enumerators to conduct the crop cutting, drying, threshing/shelling,
cleaning/winnowing, and weighing.
There will be at least 2 phases of field operations.
1) Enumerators will visit selected PSUs to compile a full and accurate list of all
agricultural holdings in the PSU, using the listing forms prepared by the unit in
charge. These listings will be done in close cooperation with the village
headmen.
2) Using the sampling specifications provided on the household listing form for
example, staff of the unit in charge of the survey, will proceed to select the
sample of holdings. Enumerators will then copy the identification information
(to be provided by the unit) for selected holdings onto blank data collection
forms.
91
Two or more sub samples of holdings will be selected: one to be used for the
objective area measurement and crop cutting to estimate harvesting losses, the
others to be used for crop cutting to estimate losses for other operations
(threshing/shelling, cleaning/winnowing, etc.).
H1.10 MEASUREMENT PROCEDURES
The area to be measured is limited to that one within the selected PSU/EA and
includes the area of parcels/fields and plots under various crops. Area should
be measured with a Global Positioning System (GPS) device (whether
standalone or incorporated into a tablet) and recorded as hectares with two
decimal places. Experience shows that the GPS device may not provide
accurate readings when it comes to very small plots (less than 0.25 hectare). In
case like that, perform GPS area measurement as follows: (a) measure the area
the first time going clockwise (say from point X to point Y) and record the
reading; (b) then, measure the same area for the second time but now going
anti clockwise (point Y back to Point X) and record both readings. Take an
average of the two readings.
In order to carry out the area measurement exercise, the enumerators should
have the following tools available: (i) GPS device; (ii) notebook/pen; (iii) area
measurement form.
In practice, the enumerator would then enlist the help of the respondent to
survey the parcel to be measured by going around, recognize the boundaries as
well as the number of fields and plots which constitute it, and then assign each
parcel, field and each plot a unique serial numbers. The starting point of the
field to measure is marked temporarily by fixing a peg on the ground. The GPS
device is turned on next and properly set up depending on the model at hand.
Finally, the enumerator walks around the perimeter of the area and selects the
option/button “Calculate” once returning to the starting point. The GPS device
will also capture other location information such as the location of the holders'
dwellings.
H1.10B – CROP CUTTING
In order to estimate the production of a crop, two components are necessary
namely the area under the crop and the yield (production per unit area) of the
crop. The area measurement of plots has been done and the results of the
measurements properly recorded.
92
Crop-cutting allows the estimation of the yield of a crop by harvesting,
threshing and weighing the produce of randomly located subplots in selected
plots.
In order to conduct the crop cutting exercise, assuming 5x5m subplots, the
following equipment and material might be needed:
1. A list of selected plots for Crop Cutting
2. Tables of random numbers.
3. One measuring tape
4. One 7.07 meters plastic rope a ring attached at each end
5. Two 5 meters long ropes with a ring attached at each end
6. Three (3) poles/pegs
7. Bags to store the harvested crops for drying.
8. Plastic bags for weighing the threshed and cleaned samples
9. Weighing scale
As soon as the crop is ripe, and some time in advance before the holder will
start harvesting the plot, the subplots for crop-cutting must be harvested. The
procedure to select fields/plots for crop-cutting is given below. There should be
a form where all plots eligible for crop cutting will be copied with their
corresponding IDs in different columns so that selection of crop cutting
fields/plots can be done. Hence, the crop fields/plots for the crop-cutting
exercise with their corresponding holder ID, parcel number, field number, plot
number and crop name and crop code will be recorded in that form. After that,
selection of crop plots for crop-cutting will be performed. Using a table of
random numbers, crop fields from each crop type will be selected for crop-
cutting experiment. In one specific section of the form, all the crop-cutting
results obtained from the 5mx5m field plots on the selected crop fields will be
recorded. The way crop-cut or harvest is done on the plot should be done in
similar way as the holder is harvesting. Also prior to start the crop-cut on the
field, the enumerator should get the permission from the holder to do the crop-
cut.
The crop-cutting experiment procedures should go along the following lines: in
each selected field for crop-cutting one randomly selected plot will be taken
and the crop on the plot should be cut, threshed and weighed. The steps for
selecting a random plot are as follows:
The South-Western corner of the field will always be selected as a
starting point, say SW. For convenience, a pole is fixed at the starting
point, SW.
93
Beginning from the starting point, the length and the breadth of the
field is measured by pacing, fractions of the paces being rounded off to
the next whole number
Two random numbers, one each for the length and breadth and not
exceeding the respective numbers are selected by using random number
table.
A random point in the field corresponding to the selected paired number is
chosen by walking from the starting point of field, SW, along its length and
stop at a distance of a selected number corresponding to the length, then walk
into the field along the breadth and stop at a distance of the other selected
number. A pole is fixed at this point, call it point A. if the random point does
not fall within the field, the pair of random numbers is rejected both for the
length and the breadth and another random point is selected using a fresh pair
of random numbers.
Plot allocation: to set up a plot, a framed rope is used which has three parts:
first rope is 7.07 meters long, and on each end of the rope is one ring; second
and third ropes which are 5 meters long each, with one ring at the end of each
rope.
A line which is 7.07 meters is set up in the direction of the enumerator’s
shadow at the random point A. The rope is pulled straight and fixed to poles A
and B by rings as shown on the diagram above. The two sides of the rope are
pulled with C to the left side of the line AB until both sides of rope AC and BC
are stretched and the ring C is fixed by a pole. This is a right–angle triangle,
one half of the required square plot for one crop-cutting. Then, all the stalks
lying on the boundaries under the rope AC and BC of the triangle inside the
triangle will be included in the yield of sample plot. After cutting all the stalks
94
of the triangle ABC, then the ring C will be pulled to the opposite side to form
the second right-angle triangle ABC, which is the remaining one-half of the
plot. All stalks that are clearly located inside the triangle ABC are cut as
before. However, the stalks on the boundaries AC and BC of the second
triangle should not be included in the yield of the sample plot.
Freshly cut samples from the selected plot are then immediately threshed and
put into bags with identification information (holder ID, parcel number, field
number, plot number and name of crop. The results of the weighment should
be recorded in the space provided. The crop that is cut and threshed contains
some moisture, therefore, it is important to dry it for 10-15 days and then re-
weigh the dried crop and the result should be recorded in the space provided. In
case moisture meters are available, moisture content should be recorded.
H1.11 ESTIMATION PROCESS
In essence there are only two basic formulas, one for losses at harvesting and
the other one for losses at the other levels (threshing/shelling,
cleaning/winnowing, etc.).
1. Losses in harvesting
A stratified two-stage random sampling design has been used with village as
primary sampling unit and holder as secondary sampling unit; estimates of
production and loss per hectare can be produced as follows (assessment and
collection of data on post-harvest food grain losses, FAO Rome 1980):
is production per hectare for holder j in village i
95
is the area under the crop for holder j in village i
is the area under the crop for village i
is loss per hectare for holder j in village i
is the number of sampled holders
is the number of sampled villages for the given stratum
Pi = is the estimate of production for village i
P = is the estimate of production per hectare for the given stratum
Li = is the estimate of loss for village i
L = is the estimate of loss per hectare for the given stratum
Hence the percentage loss in this harvesting stage is given by
PCL = X 100
Estimates of variances for P and L can easily be derived as well as for their
ratio PCL.
The variance of PCL can be estimated by the following formula:
Variance of PCL = (covariance term is ignored)
96
Variance of Y =
Where Y is either L or P and is the estimate of average Y for village I and
the finite sampling corrections are ignored.
2. Losses at the other levels
Estimates of average losses for the given stratum can be computed as follows
(Assessment and collection of data on post-harvest food grain losses, FAO,
Rome, 1980):
Pi is the estimated grain production for village i
is grain production for holder j in village i
is the percentage loss of any kind for holder j in village i
is the number of sampled holders
is the number of sampled villages for the given stratum
Xi = is the estimate of percentage loss of grain for village i
X = is the estimate of percentage loss of grain for the given stratum
Variances and hence standard errors can also be worked for X.
Variance of X =
is an estimate of average percentage in village i
97
I Annex V: Ghana 2016-2017
Field Testing We used the general field test protocol described above as a guideline to plan
and implement the Ghana 2016-2017 post-harvest loss pilot survey. Only on-
farm operations have been tested and are described in this report. Field testing
for off-farm operations may be conducted in 2017-2018.
I-1: OBJECTIVES OF THE STUDY
The initial main objective of the study was to allow for the testing of some of
the statistical methodological options that are related to estimating losses at
different levels of the food supply chain (harvesting, drying, threshing/shelling,
cleaning/winnowing, transport, etc.) at farm level.
Additionally, in keeping with the needs of the FAO AGRIS program, a second
objective of estimating the difference between farmer’s own declarations of
crop production and measured production was incorporated. Table 1 below
provides more details on the piloted farm operations.
In keeping with the main aims underlined above, crop cutting was done to
estimate post-harvest losses for the main farm level operations (harvesting,
threshing/shelling, drying, and cleaning/winnowing).
Moreover, for estimating storage losses on farm, the count and weigh method.
Data was collected through enquiry (respondent’s estimates) and objective
observation/measurement in order to allow the use of regression estimation or
ratio techniques or other method to improve loss estimations.
98
Table 1: farm operations
Farm operation Description of the operation Experiment
1. Harvesting Cutting of standing crop
Declared vs. measured
production
Post-harvest losses
2. Collection
Stacking, bundling and
transportation up to the
threshing floor
Declared vs. measured
production
Post-harvest losses
3. Threshing
Separation of grain from crop
manually or using thresher
and collection of straw and
grain
Declared vs. measured
production
Post-harvest losses
4. Sorting/grading Selection of product
according to quality criteria
Not covered because is
mainly for fruits and
vegetables
5. Winnowing/cleaning
Collection of threshed
material, winnowing to
remove chaff, dust, etc.
Declared vs. measured
production
Post-harvest losses
6. Drying
Collection of material after
cleaning, spreading for
drying, heaping after drying
Not covered because of
timing constraints.
Results will be adjusted
to a standard humidity
level.
7. Packaging
Collection after the above
operations and filling in
bags/baskets/other packaging
material
Not covered because of
timing constraints.
Results will be adjusted
to a standard humidity
level.
8. Transportation
Loading of packed material in
threshing yard, transportation
to store of farmer, unloading
for storage, transportation
from threshing yard to market
yard, unloading at market
yard
Not covered because of
timing constraints.
Results will be adjusted
to a standard humidity
level.
9. Storage at farm level
During storage,
cleaning/grading, before
sending to market for sale or
own consumption
Post-harvest losses
99
I-2: INSTITUTIONAL ARRANGEMENTS
In Ghana, data gathering activities for the agricultural and rural sectors are
being undertaken by the Ghana Statistical Service (GSS) and the Statistics,
Research, and Information Directorate (SRID) of the Ministry of Food and
Agriculture (MOFA). These two institutions are the backbone of the Ghana
Agricultural Statistics System. The most relevant data collection activities in
relation to agriculture undertaken by them are i) the Agricultural Sample
Census (ASC); the multi-round annual crop and livestock survey (MRACLS)
with its improved version known as the Ghana Agricultural Production Survey
(GAPS); and household surveys with an agricultural component (Living
Standards Measurement Conditions Survey – Integrated Surveys on
Agriculture).
This experiment was coordinated by the Global Office of the Global Strategy
(FAO, Rome), with support of FAO-Ghana.
A Technical Committee between the country's central statistics agency GSS
(Ghana Statistical Services), the Ghana’s Ministry of Food and Agriculture
and the Savannah Agricultural Research Institute (SARI), was established to
help with the implementation of the various tests and assist in all aspects of the
survey.
The fieldwork was organized and undertaken by MOFA, under the supervision
and with the technical support of the Global Office. GSS provided all sampling
material and experts and also provided support for data entry, editing, data
processing and database creation. SARI provided the grain laboratory analysis
services.
For this pilot test in Ghana, the adopted methodology did not explicitly address
the following issues that are all known to cause post-harvest losses:
Extreme weather conditions/aberrations in isolated areas.
Market gluts of food commodities.
Paucity of storage facilities.
Lack of proper handling (like when bags of commodities are carelessly
thrown away, etc.).
Dearth of transport.
These factors lead to exceedingly variable losses in time and space, requiring
special efforts for taking them into consideration.
100
I-3: COVERAGE
Countries have different administrative make-ups encompassing agro-
ecological zones, and regions that might be divided into provinces; provinces
themselves might be divided into districts, and districts subdivided into other
entities such as localities/villages/enumeration areas.
In Ghana, there are clearly defined agro-ecological zones i) the North with one
rainy season (April to October) ii) the Transitional zone, and iii) the South with
two rainy seasons (April to July and September to November).
For the conduct of statistical surveys in Ghana, the country comprises of
Regions subdivided into Districts; Districts are subdivided into Enumeration
Areas (EAs); EAs in turn contain dwellings with households. The proposed
tests covered Sawla district in the North agro-ecological zone, and Kintampo
district in the Transitional agro-ecological zone.
I-4: SCOPE
Scope also refers to the data/information items to be collected during the
survey, for the chosen crops/commodities. In the Ghana pilot study, the
following cereal crops were covered: maize, rice, sorghum and millet.
The survey items refer to specific information on survey characteristics. In this
particular exercise, they are grouped into the following questionnaires by type
of the actor involved and supply chain operation. For the purpose of the test,
on-farm post-harvest losses of small to medium-scale farmers (operations and
storage) were initially investigated.
These questionnaires, once again were derived from the questionnaire bank
provided in the general protocol for undertaking post-harvest losses field tests.
The survey consists of a set of seven questionnaires and the instructions for
completing these questionnaires will be discussed in the sections that follow.
The purpose of these farm household questionnaires was to provide
information on general characteristics of the farm and the losses they suffer as
they relate to crop production. The questionnaires used are illustrated on the
following pages.
101
QUESTIONNAIRE FH0 - FARM HOUSEHOLD LISTING
102
QUESTIONNAIRE FH1 - SAMPLED FARM HOUSEHOLDS
103
QUESTIONNAIRE FH2 - FARM HOUSEHOLD DEMOGRAPHY
104
QUESTIONNAIRE FH3 - FARM HOUSEHOLD CHARACTERISTICS
105
QUESTIONNAIRE FH4- FARM HOUSEHOLD LIST OF FIELDS
106
QUESTIONNAIRE FH5 - FARM HOUSEHOLD FIELD AREA
107
QUESTIONNAIRE FH6 - FARM HOUSEHOLD CROP CUTTING
108
QUESTIONNAIRE FH7 - FARM HOUSEHOLD STORAGE
LOSSES BY OBSERVATION
109
QUESTIONNAIRE FH8 - FARM HOUSEHOLD STORAGE LOSSES
GRAIN SAMPLE
110
I-5: SAMPLING DESIGN AND SELECTION OF
DOMAINS/STRATA For selecting the respondents from which to collect the data for assessing post-harvest losses a stratified multistage random sampling method is used most of the time. Agro-ecological/climatic zones are used as main strata. Depending on the administrative division of any given country (region, province, district, etc.), the highest level administrative division (for instance a province containing districts, etc.), chosen as Primary Sampling Units (PSU) (districts, etc.) are selected at the first stage, then Secondary Sampling Units (SSU) (enumeration areas, villages, etc.) at the second stage, and farming households Tertiary Sampling Units (TSU) at the third stage. Additional or less levels might need to be dealt with depending on the specificities of any given country. For the Ghana PHL pilot survey, however, two agro-ecological zones were purposively selected; in each zone, one district was also purposively selected based on the availability of existing sampling frames (from the 2015 agricultural survey) and given the time and resource constraint. The Secondary Sampling Units (SSU) or Enumeration Areas (EAs) were then randomly selected within each district, with a probability of selection proportional to the size of each EA measured in number of households. Finally from each selected EA, a fixed number of farming households, the tertiary Sampling Units (TSU) was randomly selected. Hence, the results of the pilot survey could be extrapolated at any given district level.
I-6: SAMPLE SIZE AND SAMPLING PROCEDURE
I-6-1: SAMPLE SIZE
The sample size for data collection was decided on the basis of time and
resources available. Sample sizes and sampling procedures for each operation
are described below.
Farm level post-harvest operations:
Purposive selection of two agro-ecological zones: North and Transitional.
Purposive selection of one (1) district (PSU) from each zone: Sawla in North
and Kintampo in Transitional.
Random selection of twenty (20) Enumeration Areas (SSUs) from each chosen
district.
Random selection of fourteen (14) farming households for data collection via
enquiry; for data collection by observation/measurement, six (6) farming
111
households from the list of already selected fourteen farming households were
selected randomly.
Storage at farm level:
The same sample of farmers (as drawn for data collection at farm level) was taken for data collection on storage losses at farm level by inquiry and observations.
I-6-2: SAMPLING PROCEDURES
As already stated, the selection of sampling units was done on the basis of
simple random sampling techniques without replacement for each
crop/commodity. Procedures for each stage are described below.
Selection of PSUs:
The two districts in the two agro-ecological zones were purposively selected.
Selection of SSUs:
A list of all PSUs was provided by GSS from their most recent Ghana
Agriculture Production Survey (GAPS). Then twenty PSUs selected randomly
using probability proportional to a measure of size – with the measure of size
being the number of agricultural holdings within the EA.
Selection of TSUs:
A fixed number – fourteen – of SSUs were then chosen from each EA using
random systematic selection.
Selection of fields, plots, and storage structures:
This selection was done in order to record the losses during farming operations
by measurement and for storage of the grains in granaries or other storage
structures.
For the selected field crops (maize, rice, millet, sorghum), plot selection for
each crop was done by preparing a list of all the fields of the selected farmers
for each crop grown; one field for a particular crop was then selected
randomly. After field selection, a crop cutting plot of 5mx5m (3mx3m for rice)
was placed to assess the losses by actual observation/measurement.
112
The table below provides additional details on the final sample and its
breakdown by district
Number
of EAs
Number of holdings per EA Total number of holdings
Total Enquiry Measurement
& Enquiry Total Enquiry
Measurement
& Enquiry
SAWLA 20 14 8 6 280 160 120
KINTAMPO 20 14 8 6 280 160 120
Total 40 560 320 240
I-7: DATA COLLECTION METHOD
For the purpose of testing, the survey used a face-to-face interview method
using a paper questionnaire (PAPI). Computers or tablets (CAPI/TAPI) were
not used for this experiment because of the lack of experience of MoFA
enumerators in using these devices for surveys. The project team did not want
to add an additional layer of complexity to an already complex experiment and
given that the prime objective was to assess losses and not to compare CAPI
and PAPI interview methods.
Agricultural holdings and the agricultural holder (as defined) were identified
and data relevant to the holding collected from the holder. One of the important
aspect of the data collection exercise during the survey was the measurement of
land area. Another one was the estimation of the production based on the
yields. Both of these required information were collected by subjective and
objective methods. The subjective method consisted in asking the holder to
estimate the area as well as the production. The objective method consisted in
measuring the area objectively using appropriate instruments (GPS) and also in
carrying out the crop cutting exercise to measure the yields under the important
crops for the purpose of estimation of loss and production. The physical
collection of the required information from the respondents was accomplished
through the proper use of the nine (9) questionnaires that were initially tested
on a small-scale and after thorough training of the enumerators.
I-7-1: COLLECTING DATA VIA ENQUIRIES
The questionnaires that were developed for the purpose of collecting the data
by the specially trained enumerators and via face to face interviews have
already been presented in the previous paragraphs. They will be succinctly
described again here.
113
Farming households listing:
Each of the selected villages/enumeration areas was already completely
enumerated during the previous GAPS (2015) and provided the following units
of information at the beginning of the survey:
Identification particulars of the village/enumeration area
Details of farmers with respect to operations of the holding,
crops/commodities grown or expected to be grown in year of survey,
and areas under crops.
The preliminary selection of the farmers was done on the basis of this
information.
Crop production and losses during farm operations:
The data collection concerning the harvest and other operations prior to storage
should take place at harvesting time or within a week after harvesting. Farmers
have been interviewed for their own assessment of the production quantities
and quantitative losses for each of the operations. The farmers have also been
asked to provide the main cause of the loss and details on the method and
equipment used in the farm operations.
A proper timing is really key to a proper measurement of production and losses
and therefore it is key for data collection teams to prepare a detailed planning
of their field visits and inform the farmers ahead of time. In this respect, it is
insightful to note that a significant share of the farmers that had been selected
for crop-cutting in one of the two districts had already harvested most of their
plots when the enumerators arrived on the farm. This meant that the data
collection teams had to look for additional fields with standing crops and to
perform all the preliminary steps (random selection, placement of yielding plot,
plot measurement, etc.) before the actual measurement. This has certainly
created a bias in the results and delayed the activities.
Storage losses at farm level:
Estimates of storage losses by farmers were collected monthly over a period of
3 months. The enumerators asked the farmer estimates of:
Its previous balance/stock of produce
Additions and withdrawals since the last enquiry/visit
Total quantity stored at the time of the enquiry
114
The farmers were also asked about the type of storage facility
(modern/traditional) and the main causes for losses.
The choice of the study period length of 3 months reflected more the time and
resource constraints attached to the pilot study than actual farming practices. A
study over a wider time-period, from harvest to harvest for example, would
provide results that are more complete.
I-7-2: COLLECTING DATA VIA OBSERVATIONS/MEASUREMENTS
Harvesting losses at farm level for cereals:
The data collection protocols for losses during harvesting, threshing, and
cleaning/winnowing of rice, maize, millet and sorghum in the Ghana survey
were similar. Details of the selected field, variety, soil type/condition, planting
date, harvesting date, method of harvesting, equipment used (all variables
having an impact on losses) have been recorded.. For the case of traditional
harvesting, a plot of 5mx5m (except for rice, for which 3mx3m plots were
used) was set up and harvested in line with farmers’ methods. The harvested
crop was collected separately; then fallen grains inside the plots were next
collected and weighed.
In order to estimate losses during threshing/shelling, the same harvested crop
from the 5mx5m (or 3mx3m) plot was also threshed following farmers
practices. Produce and straw were weighted separately. A sample of few
hundred grams of straw was collected and the grains within were duly
counted/weighted.
The estimation of losses for cleaning/winnowing operations a sample of around
one kg of unclean grain-straw mixture was drawn and cleaned/winnowed using
the method followed by the farmers. Grain and straw were next collected
separately afterwards from the operation. A sample of few hundred grams of
grains recovered were counted/weighed and the measurements recorded.
Storage losses at farm level for cereals:
The loss estimation techniques used in these cases dictated making use of the
collection of samples of 1-1.5 kg of commodity to be taken every month
(conditional to availability). Consumption, addition to stock, sales, or
processed stock during the previous month as well as remaining stock were
recorded during the survey period. The samples were packed appropriately and
given identity slips and sent to the laboratories for further analysis immediately
115
after they had been collected. They were analyzed for moisture content, 1000
grains weight, number of undamaged grains, and infested/damaged grains, and
their total weight duly recorded. The results of these analyses will be combined
with the information on the quantity of grain stored obtained through the
enquiry to compute % and quantity loss estimates.
I-8: TIME REFERENCE, SURVEY PERIOD AND FIELD
ENUMERATION
Different time reference are normally used in many surveys depending on
various characteristics. The time reference for area and production used during
the pilot survey in Ghana referred to the day of the enumeration. However, the
time reference for agricultural practices were the agricultural season. As
Kintampo district benefits from two seasons for maize (one major and one
minor), the farmers were asked whether their answers referred to the last
season (major) or to the current one (minor). This aspect will need to be taken
into account in the compilation of the results and in the analysis of the results,
to avoid comparing and/or merging production and losses estimates referring to
different seasons. The reference period for other data items mentioned in the
questionnaires, were for the last twelve months.
The field work of the pilot survey was done from November 2016 to February
2017, which corresponds to harvesting period for the four crops under study
(either the main or minor season, depending on the crop and district). Within
that specified time, all required data for the given agricultural year were
collected in two visits – visit I, during the planting period and visit II, during
the harvesting period. Prior to that:
1. Enumerators visited selected PSUs to update the list of agricultural
holdings in the PSU, using frame material provided by GSS and the
listing forms prepared by GSS and MOFA. This activity was done in
close cooperation with the farmers and village headmen.
2. Using the sampling specifications provided on the household listing
form for example, staff of the unit in charge of the survey proceeded to
select the sample of holdings. Enumerators then copied the
identification information (provided by GSS and MOFA) for selected
holdings onto blank data collection forms.
That facilitated the selection of sub samples of holdings: one to be used for the
objective area measurement and crop cutting to estimate harvesting losses, the
116
others to be used for crop cutting to estimate losses for other operations
(threshing/shelling, cleaning/winnowing, etc.).
Enumerators and their supervisors operated in teams of two and there were a
number of visits to the selected holdings:
Visit I took place during the cropping season (when the plants were
already quite tall) to set up the yielding plot.
Visit II took place when the crops were ready to be harvested for the
enumerators to conduct the crop cutting, drying, threshing/shelling,
cleaning/winnowing and weighing.
I-9: FARM CROP CUTTING TECHNIQUE
In the Ghana survey a crop cutting technique different from the one in the
proposed general field test protocol was used. Hence we are reproducing it for
the convenience of the reader. This part is directly taken from the enumerator’s
manual that has been prepared for the training of the field teams.
Overview
In order to estimate the production of a crop, two components are necessary,
namely the area under the crop and the yield (production per unit area) of the
crop. The area measurement of fields has been done and the results of the
measurements recorded on QUESTIONNAIRE FH5.
Crop cutting will allow us to estimate the yield of a crop and the losses during
harvesting, threshing/shelling, and cleaning/winnowing. This will be done
through processing the produce of randomly located yielding plots in selected
fields.
For crop-cutting to be done properly two teams of two supervisors each should
execute the job.
The first team will do the crop cutting manually according to the techniques
used by the farmer.
After the manual harvesting has been done, the second team of supervisors will
enter the field and collect all fallen ears/cobs, grains and weigh and record the
information.
117
Equipment and material that you will need for crop-cutting;
1. QUESTIONNAIRE FH5 FARM HOUSEHOLD FIELD AREA
2. QUESTIONNAIRE FH6 FARM HOUSEHOLD CROP CUTTING
3. Tables of random numbers
4. One measuring tape
5. One solid demarcating tool
6. Two shears.
7. Two knives.
8. Bags to store the harvested crops for drying.
9. Labels for the samples.
10. Plastic bags for weighing the threshed and cleaned samples.
11. Rolls of string.
12. One sheets made of jute and measuring not less than 1 X 1 square
meter.
13. Moisture meters
As soon as the crop is ripe, the yielding plots for crop cutting must be
harvested.
You will need one QUESTIONNAIRE FH6 for each field under study. If there
are several crops on the same field, record the information for each crop on a
separate line (Part B).
This questionnaire will be filled in partly during the first visit (placement of the
yielding plot) and entirely during the second visit (harvesting of the yielding
plot).
Using Questionnaire FH1, you have selected 8 farming households out of the
fourteen in the given EA. These 8-farming households constitute a sub-sample
of farmers for the crop cutting. For each of these 8 farmers, select one field of a
given crop (maize, millet, sorghum, rice) at random. It is on those selected
fields that the yielding plot for crop cutting will be set up.
As soon as the crop is ripe, and ideally some time in advance before the holder
will start harvesting the field, the yielding plots for crop-cutting must be
harvested. The procedure to select fields for crop-cutting is given below. There
should be a form where all fields eligible for crop cutting will be copied with
their corresponding IDs in different columns. Hence, the fields for the crop-
cutting exercise with their corresponding holder ID, parcel number, field
118
number and crop name and crop code will be recorded in that form. After that,
selection of fields for crop-cutting will be performed. Using a table of random
numbers, at least one field from each crop type will be selected for crop-cutting
experiment.
In Parts B to D of Questionnaire FH6, all the crop-cutting results obtained from
the 6mx6m or 3mx3m yielding plots on the selected fields will be recorded.
The way crop-cut or harvest is done on the field should be done in similar way
as the holder is harvesting. Also prior to start the crop-cut on the field, the
enumerator should get the permission from the holder to do the crop-cut.
The crop-cutting experiment procedures should go along the following lines: in
each selected field for crop-cutting one randomly selected field will be taken
and the crop on the field should be cut, threshed/shelled, cleaned and
winnowed and weights recorded for each stage.
The steps for randomly selecting a yielding plot within a field are as follows:
PLACING THE YIELDING PLOT WITHIN A FIELD
1. First of all find the field and identify the starting point which was used
for area measurement (Point A, side A-B) by checking the lengths (in
meters or paces) of some sides from QUESTIONNAIRE FH5.
Compute half-perimeter of the field (in meters or paces), by dividing
the perimeter by 2.
2. Use the table of random numbers to select a random number between 1
and the number of vertices or sides for the yielding plot. Remember that
from QUESTIONNAIRE FH5 all sides/vertices are numbered. This
first random number will determine the vertex (point) and side from
which you will go inside the field. Look up the length in meters or
paces of that selected side and select a random number between 1 and
the length of that side. The random number will give you the distance
from the vertex (point) from which you will enter the field. You will
then select a third random number less than half the perimeter of the
field that will take you inside the field. Instructions on how to use a
table of random numbers have been provided. Write down the second
random number in columns [27-29] and the third random number in
columns [31-34].
119
For the yielding plot, the second random number (Second) will determine the
point on the perimeter from where you go into the field. The third random
number (Third) will determine how far you go inside the field.
The next activities to be carried out can best be illustrated through the use of an
example.
Let us assume that the first random number selected is 5 (vertex E on side E-F).
Look up the corresponding length of side E-F from the table in
QUESTIONNAIRE FH5. Let us assume it is 13 m or 13 paces. Then select a
random number between 1 and 13, for instance 7. Measure paces (or 7 m)
meters from point E on side E-F. This will give you point P1 from which to go
inside the field in a perpendicular direction. Mark this point with P1 on the
sketch in QUESTIONNAIRE FH5. See figure below
For this example the second random number for the yielding plot is 058. You
should measure 58 meters or paces into the field from the point “P1” at a right
angle against the side E-F.
Mark the point arrived at with a peg as the first vertex on the diagonal of the
yielding plot. Keep on setting up the yielding plot with the three ropes
(8.485m, 6m, 6m) for the 6mx6m yielding plots (maize, millet, sorghum) and
(4.242m, 3m, 3m) for 3mx3m rice yielding plot.
If the distance into the field (3rd random number) goes beyond the boundary
by a certain amount, move back by the same amount. For instance if the
selected second random number is 145, but the length across the field from
“P1” is only 89 meters, the boundary is reached before you have measured the
full distance. In that case you should go back straight in the opposite direction
145 - 89 = 56 meters.
120
If any of the vertex points falls into an area that has been deducted from the
measured area, e.g. not planted area within the field, you have to select a new
point. Cross both the previously selected numbers for that yielding plot and
select a new pair of random numbers.
I-10: ADDITIONAL SAMPLING CONSIDERATIONS
These considerations were also not provided in the proposed general field tests
protocol. Therefore, they are being described here for the benefit of the reader.
They are also taken directly from the enumerator’s manual.
I-10-1: SPEAR SAMPLING
In post-harvest loss studies, the ultimate sampling unit from which a sample
will be collected will differ according to the stage of the post-harvest system
being studied. It refers to the smallest unit in which grain is contained. For
estimating yield at harvesting period, it may be individual yielding fields in a
farmer's field; for estimating losses at threshing, it may be the stacks of un-
threshed panicles; and for estimating storage losses, it may be a storage bin or
similar modern or traditional structure or even an individual bag within a store.
For making the collection of grain samples easy to perform, the usage of grain
spears of certain type is recommended.
Spears are cheap, simple to use and are a quick way to obtain sample grain
from bagged commodities. By pushing the tip of the spear into the bag,
maintaining the body with the open side face down inserted for the required
distance; grain is then collected by twisting the spear in such a way that the
open side is turned upwards; withdrawing the spear from the bag, will tip out
the grain into a container; figure 2 illustrates the process.
121
Figure 1. Area of produce sampled when a spear is introduced into the sack.
Spear sampling though of widespread use, can be quite inaccurate to the extent
it may nullify most of the results obtained upon analyzing samples collected by
this method. Figure 3 tells the story.
Hence, spear sampling should be avoided to the extent possible. When taking a
probe sample, a compartmented probe (figure 4) available in various sizes,
should be used instead.
Figure 2. Compartmented bag probe – 54 cm length. (Harris and Lindblad, 1978)
I-10-2: THE SAMPLING OBSERVATIONAL UNITS ON FARM
The observational unit is the container, location, or process from which a
sample will be removed to determine the loss evident in the sample.
This is the smallest division or unit in which grain is held. It might be stacks in
the field, bags, small silos or granaries on a farm, or woven baskets. It would
be a single basket or rather than all of a farmer’s storage baskets; it would be
individual bags rather than the whole warehouse. Accuracy of the entire survey
will depend on the accuracy with which the loss is determined on each
observational unit.
122
To facilitate sampling, the observational unit should be as small as possible.
This makes it easier to get a representative sample since it will be possible to
mix all the grain thoroughly and reduce the sample taken by quartering or
using a sample divider. This may be feasible where the grain is in baskets or in
stacks in the field. In silos or granaries, it may not be possible and, unless the
sampling is done with skill, the sample may contain systematic error which
cannot be removed by any later calculation or analysis.
When any container is sampled as a unit, it is assumed that the defect,
contamination or other characteristic to be determined is uniformly or at least
randomly distributed within the unit. In practice, such is usually not the case.
Obtain samples (minimum 2) from one observational unit (stack, basket, crib,
etc.) on each farm. Choose the unit with random numbers after seeing how
many units there are on the farm.
All samples must be labeled and retain their identity as to date collected , exact
location of source, how the sample was obtained, grain type, variety (if
known), time in storage, and type of storage.
Procedures for Sampling
Standing Grain in the Field
Do crop-cutting as explained.
In the Field in Stacks (If each stack contains more than 2 kg of shelled grain)
Give each stack a number starting with 1 and going as high as
necessary.
Choose as many random numbers from the table furnished as there are
samples to be taken (2 minimum).
Shell each stack whose number was chosen.
Reduce the grain by coning and quartering to a sample of 1.5 kg.
Package the sample for transmission to the laboratory.
Note: If each stack contains less than 2 kg of shelled grain, choose twice as
many random numbers as there are samples to be taken. Combine the grain
from two stacks into a single sample for transmission to the laboratory.
123
Figure 3. Coning and Quartering
When the Shelled Grain is Stored in Baskets
Give each basket a number starting with 1 and going as high as necessary.
Choose as many random numbers as there are samples to be taken.
Reduce by coning and quartering each basket whose number is drawn to a
sample of 1 to 1.5 kg.
Package the sample from each basket for transmission to the laboratory.
When the Unshelled Grain is Stored in Small Units (Such as Baskets and
Bags).
If the grain is stored in small units on the cob, head, or panicle, shell the
contents of the whole unit before coning and quartering to yield a 1- to 1.5-kg
sample.
124
When the Unshelled Grain is Stored in Large Cribs, Silos, or Granaries.
To sample grain stored unshelled in cribs, silos, or granaries, unload and shell
the entire lot. Then cone and quarter to obtain a sample of I to 1 .5 kg. Or
unload the grain equally into baskets and then use the method for unshelled
small units (choosing baskets by stratified random sample).
Note: In storage, ears of cob maize or panicles of sorghum/ millet and maize
can be labeled randomly as the crib is filled. The farmer can then be asked to
set these ears aside as he encounters them during emptying. Determining an
adequate sample of ears or heads from a crib can be a problem, however. This
procedure should be used only after careful study of its applicability to the
local situation.
Large Bulk Storage Units, Shelled.
Obtaining a representative sample from a large bulk container is difficult.
Ideally the grain would be transferred into another container in such a way that
samples could be obtained from the grain as it falls into the new container. A
container small enough to be handled easily should catch the entire falling
grain stream until it is full or passed through the entire stream and the caught
grain placed into a larger sample container. This procedure would be repeated
at frequent, regular times throughout the transfer.
When all the grain has been transferred, the sample that has been collected
may be reduced by coning and quartering to 1 to 1.5 kg for transmission
to the laboratory .
If it is not possible to sample the grain during a transfer, then a probe may be
used. It is recognized from research results that a probe sample is not
representative. When probe sampling is used a note should be made of that fact
in the final report. In using the probe, an effort should be made to reach every
part of the storage container. Several times as much grain as is necessary for
the final sample should be taken and then reduced by coning and quartering.
Samples should be taken with the probe in at least the positions shown in Fig.
4, using a compartmented probe that samples at all levels.
125
Figure 4. Probing locations in rectangular and round bins
Mass Storage in Bags.
Obtaining a representative sample of a large mass of grain stored in bags can
only be done if every bag is accessible. To sample such a store requires that
one chose enough random numbers and then move the grain one bag at a time
to a new location diverting bags for sampling corresponding to the random
numbers. The diverted bags should be sampled, preferably by coning and
quartering the whole bag to obtain l to 1.5 kg of sample for the laboratory. The
remainder can be returned to the bag and to the store.
A less satisfactory alternative is to obtain a sample from each randomly chosen
bag by probing. A probe long enough to reach diagonally from corner to
corner of the bag should be used and the bag should be probed on both
diagonals and in enough other locations to obtain 1 to 1.5 kg of grain from
each bag.
It should be noted if every bag is not available to be sampled so the result will
refer only to those bags that were accessible. The bags sampled should be
chosen by assigning numbers to those that are available and using a table of
random numbers to choose the bags.
Sampling procedures should always be reported, especially when the sampling
is suspected to be non-representative as in the case of stacked bags, unshelled
grain heads and cobs, and when there are visually observed concentrations of
insects or mold, or both.
126
J Annex VI: Summary of the
Desk Study of the Malawi
2009-2010 PHL Survey Data Maize postharvest losses (PHL) in Malawi have been analyzed using data
provided by the Ministry of Agriculture, Irrigation and Water Development of
Malawi. The Ministry conducted a post-harvest losses survey in 2009/10
season in order to come up with estimates of maize post-harvest losses arising
from harvest to storage.
The study has to be considered an explorative starting point for a more
comprehensive analysis on postharvest losses. The following is a summary of
the main results and challenges encountered.
The main objectives of this study are the following:
1. Evaluate the incidence of PHL in Malawi
2. Assess the relevance of losses in terms of weight (quantity losses)
3. Investigate the factors related to the probability of experiencing PHL
4. Highlight the variables associated to a higher amount of losses
5. Explore the possibility of building a statistical model to predict PHL,
using relevant and easy-to-collect variables.
Data used
The sample of 850 households was randomly selected in each district across
the country (apart from Likoma), sampling one Extension Planning Area
[EPA] per district and interviewing 30 households per EPA. In Lilongwe and
Mzimba districts 60 households were chosen per district. The survey consists
of three waves: August 2009, January and March 2010. However, due to
problems during the selection and reporting phase, there is no unique identifier
127
for farmers that could allow to construct a panel. Therefore, only the first wave
was considered.
The two most relevant variables for the study are PHL experience and the total
weight loss of maize. The former is measured by asking the farmers whether
they suffered post-harvest losses in the period considered or not. The latter is
collected by asking the farmers to provide an estimate of the weight loss they
suffered.
The dataset includes also total weight loss of maize due to LGB (Large Grain
Borer) attack, variety of cultivated maize, holders’ total area planted with
maize, number of household members, farm geographical location and
demographic variables.
Other important variables for the analyses are those on loss prevention
techniques. Information about practices were collected by asking farmers
which practices they consider as useful and which they actually implemented
to prevent/arrest the losses. It was decided to focus only on the strategies
actually put in place to cope with PHL, even if this choice may lead to
endogeneity problems (the existence of PHL reduction measures on the farm
may be correlated with higher PHL losses).
Loss prevention techniques practiced by the farmers:
Early harvesting Re-drying Improved storage hygiene
Stooking when harvesting Use of chemicals Use of protected granaries
Processing with better care Use of ashes Use of other techniques
Source: PHL database, Malawi
Data quality
Insufficient data quality and lack of clear documentation constitute the main
limitations of the study. The data present numerous missing values for the
variables of interest which called for extensive data cleaning before the
analysis. Another major problem affecting the data is the lack of coherence
between the variable on PHL experience and the total weight of post-harvest
losses. Total weight of losses was set to 0 when the farmers declared that they
did not experience any loss. Despite this, 88 observations are still reported as
missing for the total weight of losses. Recoding was undertaken also for the
total losses caused by LGB.
128
Table 1
In spite of the recoding work undertaken, the data still presents some
inconsistencies. In particular, continuous variables such as total weight of
losses and size of cultivated land include some outliers (Figure 1). Standard
deviations from the mean of total weight of losses and relative weight of losses
were used to detect outliers. Three observations were recorded as outliers for
both the variables and it was decided to simply delete these records.
Figure 1: Boxplots
129
Lack of documentation on sampling procedures, together with missing
sampling weights did not allow to compute reliable regional or national
estimates of losses and the summary statistics were calculated only considering
the sample. Moreover, due to the lack of clear information on the unit of
measure, it was assumed that the total weight of losses was measured in
kilograms and the total area under cultivation in hectares.
Modelling
After a descriptive analysis of the incidence and relevance of post-harvest
losses in the sample, an evaluation of the relationships between losses and
some relevant factors has been carried out. The probability of experiencing
losses and the amount of losses are considered. In both cases the aim of the
analysis is to find some variables related to PHL, which could be used to
simply describe or even to predict them.
The main explanatory factors of PHL experience and weight losses included in
the modelling are: the size of cultivated land, number of household members,
sex of the household head, region, variety of seeds, practices put in place to
reduce PHL and the eventual information received by the extension services.
The first step is the analysis of the probability of suffering PHL, which was
carried out by performing a logistic regression using the experience of PHL
(dichotomous variable) as a dependent variable and the abovementioned
controls as independent variables. The relevance of the losses was instead
examined using OLS regressions with the total weight of losses, its natural
logarithm and the total weight of losses relative to cultivated land as dependent
variables. Given the high number of practices put in place to reduce PHL, it
was decided to take into account all the interactions between them in the
regressions using a Factor Analysis. The two resulting factors were then
included as controls in the regressions described above.
Results
Given the data quality issues underlined above and the absence of key variables
known to be correlated with losses, such as production quantities or yields, the
results of the models are to be taken with the utmost care and are reported here
only for illustration purposes (detailed results can be obtained directly from the
authors).
130
Logistic model:
Sex of household head, number of household members, area under
cultivation and region are not significant for PHL experience;
Farmers cultivating composite (i.e. recycled) and hybrid seeds’ varieties
presented a higher probability of experiencing PHL with respect to
local varieties (significant and positive coefficient).
The factor including practices related to harvest operations (early
harvesting, re-drying, improved storage hygiene, stooking, use of
protected granaries and processing with care) has a negative
relationship with postharvest losses experience (significant and negative
coefficient).
Furthermore, it is interesting to note that information supplied by
extension services does not display a significant coefficient when
probability of experiencing losses is considered.
OLS models:
Structural variables do not display significant coefficients, with the
exception of household size, when the relative measure of losses is
considered (losses by hectare).
Composite seeds are characterized by a negative and significant
relationship with the total weight of losses, when compared to local
variety.
Factor related to practices are not generally significant.
Finally, extension officers’ information present a negative relationship
with all the three dependent variables considered: total weight of losses,
relative weight of losses and the logarithm of total weight of losses.
Possible improvements
The analysis presented, although preliminary, gives some insights on which
factors are associated with the probability of experiencing PHL and with the
importance (in quantity terms) of these losses. However, it is subject to many
limitations. It is presented here more for illustrative purposes, to show what
type of analysis or modelling can be done using survey data on PHL and, more
131
importantly, to highlight the data gaps that affect the quality and relevance of
such models.
Among the data gaps, the limited availability of metadata was a major
constraint because it implied the use of some assumptions to be able to proceed
with the analysis. The lack of key information related to production, storage
practices/facilities and climatic conditions during the post-harvest period
considerably limited the modelling possibilities and explained the poor results.
Moreover for many of the variables considered, especially those related to
practices, endogeneity could be a major issue, which could lead to biased
estimates if it is not accounted for in the modelling.
As a result of these limitations, the models fail to explain a relevant proportion
of the variance in the sample. Other models could be explored for better
prediction but more accurate data is needed first. This did not allow to build a
reliable statistical model to predict PHL using a reduced set of key variables. In
addition, the two-phase approach explained several times in this document
could not be tested because measurements were not available at the farm-level,
and could therefore not be used in a model regressing measured losses on a set
of explanatory variables (PHL reduction practices, farm size, etc.).
132
k Annex VII: Selected
Bibliography Kenton L. Harris, Carl J. Lindblad. 1976. “Postharvest Grain Loss
Assessment Methods. A Manual of Methods for the Evaluation of Postharvest
losses”. The league for International Food Education The Tropical Products
Institute (England) Food and Agriculture Organization of the United Nations
Group for Assistance on Systems Relating to Grain After-Harvest.
FAO. 1980. “Assessment and Collection of data on post-harvest foodgrain
losses”. Statistics Division Economic and Social Policy Department Rome.
J.A.F. Compton, Sherington 1998. “Rapid assessment methods for stored
maize cobs:weight losses due to insect pests” . Natural Resources Institute.
J. A. F. Compton, S. Floyd, Ofosu and B. Agbo. 1998. “The Modified Count
and Weigh Method: An Improved Procedure for Assessing Weight Loss in
Stored Maize Cobs”. Natural Resources Institute.
Ministry of Agriculture and Food Security, Malawi. FAO. 2011. “Maize
Post Harvest Loss Assessment Survey in Malawi”.
Boxall, R.A. 1986. “A critical review of the methodology for assessing farm-
level grain losses after harvest”. Natural Resources Institute.
R. A. Boxall, P.S. Tyler, P. F. Prevett. “Loss Assessment Methodology –The
Current Situation”. Tropical Products Institute. Slough, Berks, England.
133
Prepared for the World Bank by J.E. Austin Associates, Inc. “Using value
chain approaches in agribusiness and agriculture in Sub-Saharan Africa - A
methodological guide”.
UNIDO, 2009.“Agro-Value Chain Analysis and Development – The UNIDO
Approach”.
Jon Hellin and Madelon Meijer. 2006. “Guidelines for value chain analysis”.
Mr. Ian DALIPAGIC, Dr. Gabriel ELEPU. 2014. "Agricultural value chain
analysis in Northern Uganda: Maize, Rice, Groundnuts, Sunflower and
sesame". ACF Intern Food Security and Livelihood. Department of
Agribusiness & Natural Resource Economics Makerere University, P.O. Box
7062, Kampala.
Joel Johnson Mmasa, Elibariki Emmanuel Msuya."Mapping of the Sweet
Potato Value Chain Linkages between Actors, Processes and Activities in the
Value Chain: A Case of “Michembe” and “Matobolwa” Products". 2011.
School of Economics and Business studies Department of Economics of
University of Dodoma PO box 259, Dodoma, Tanzania. Department of
Agricultural Economics and Agribusiness Sokoine University of Agriculture
PO box 3007, Chuo Kikuu, Morogoro, Tanzania.
Rhoda Mofya-Mukuka and Arthur M. Shipekesa. 2013. "Value Chain
Analysis of the Groundnuts Sector in the Eastern Province of Zambia". The
Indaba Agricultural Policy Research Institute (IAPRI).