Farmers' Post-Harvest Grain Management Choices Under Liquidity Constraints

and Impending Risks: implications for achieving food security objectives in Ethiopia§

Abebe H. Gabriel and Bekele Hundie

Department of Economics
Ethiopian Civil Service College

Summary

Based on an empirical observation and analysis the paper set out to identify the different post-
harvest management processes and techniques farmers have been employing, the underpinning
reasons for their choice of certain (or a set of) techniques over the others, whether and in what
ways would liquidity constraints and/or perceptions about post-harvest risk of loss would
influence their post-harvest grain management techniques including their marketing behavior,
which aspect of risk (physical crop damage, or market risk) would be more important in
influencing their marketing practices, whether inter-crop differences in post-harvest risk of loss
are significant determinants of the relative importance of crop-share marketed, and which options
could be harnessed to achieve the objective of food security through improvements in post-
harvest grain management practices and capacities both at micro and macro levels.

Results indicate that farmers perceived post-harvest grain loss as an imminent risk. Both farmers’
expectations as well as realized post-harvest losses were found to be quite high especially for
crops such as maize. Application of insecticides, aeration and traditional methods of treatment
were the most widely techniques used by farmers. Other stock management techniques (lending
to others, etc) exist but they are less important. Marketable surplus was estimated to be above
40% of total grain output but more than one-half of it was actually sold within the first three
months following harvest. Econometric results suggest that instant sales of grains immediately
after harvest are triggered by temporary but immediate liquidity preferences to meet purchased
items for the family members in absence of or limited sources of cash other than crops sales, and
by an impending risk of post-harvest grain loss and the limited capacity to prevent it. Hence,
family size, female headedness, and tax dues and loan repayment schedules were found to be
significant variables positively associated with instant crops sales. On the other hand, livestock
ownership and the capacity to apply chemicals (both indicators of wealth) were also found to be
significant variables but negatively associated with instant crops sales. Cropping patterns suggest
that those farmers (and regions) who cultivate more susceptible crops to pest-attacks (e.g., maize)

§
The research was conducted as part of the International Food Policy Research Institute's (IFPRI)
Competitive Research Grants Program of the 2020 Vision Network for East Africa. Comments are
welcome!
are most likely to dispose off their crops immediately after harvest at cheaper prices than those
who cultivate less susceptible ones (e.g., teff).

The results underscore the need for taking on board post-harvest grain management issues as a
matter of strategic policy concern, not just from the perspective of reducing losses but also from
the view point of considering it as a viable and dynamic economic activity in terms of generation
of employment, value addition and income linkages.

The paper highlights the importance of relaxing the seasonal liquidity constraints through the
operation of financial credit markets and/or through rescheduling of financial obligations that are
due to government until later seasons as crop prices get better (e.g., payment of land taxes, input
loans repayment). The significance of chemical treatment suggests for linking input markets with
future product markets. Diversification of cash sources and in general integration of the
production process with markets so that farmers would be more market informed and oriented in
their decisions than subsistence driven is necessary for relaxing the liquidity constraints. It is also
important that disadvantaged sectors such as female headed households and the poor need special
support since they are more susceptible to shocks than the relatively better-endowed ones and that
interventions take into account and build on farmers’ resources and knowledge including
indigenous techniques.

Relaxing of the farmers’ liquidity constraints must be complemented with an intervention to

enhance the capacity of farmers to prevent post-harvest losses, which could be organized at
household, community and national levels. Otherwise, there is an imminent danger that farmers
would dispose or lose their grains any way (in fear of risk of loss) and might find it difficult to
repay their loans, which makes the matter even worse. There are also legitimate reasons to
suspect that increased production and availability of grains on farm might lead to increased
consumption by the farmers themselves instead of being marketed.

Since markets often do not perform so efficiently as to achieve optimal solutions, there is a need
for introducing and strengthening appropriate institutions that would enable markets work better.
One viable option could be to introduce grain warehouse receipt system so that farmers would
deposit their marketable surplus to be sold when prices get higher. Some of the reasons as to why
warehouse receipt system of grain management and forward grain markets make a lot of sense, in
addition to preventing post-harvest loss, are because they help facilitate monetization of the
production process, reduce availability of grains for direct consumption by the producers, and
facilitate processing. There are lessons to be drawn from other countries’ experiences that there
are a number of advantages including easing access to finance at all levels in the marketing chain,
moderating seasonal price variability, maintaining quality standards and promoting instruments to
mitigate price risks. They also help reduce the need for government intervention in grain markets
as well as the costs of such interventions. Above all, the introduction of well managed
warehouses in rural villages will reduce post-harvest grain losses which ultimately support the
country’s effort to ensure national food security.

More in-depth further studies are necessary to inform policy on credit and saving options
(including options for introducing grain warehouse receipt systems), traditional methods of grain
treatment (effectiveness, economy, health issues, etc.), and farm-nonfarm linkages and the scope
for the development of agro-processing industries, including those small-scale farmer-managed
grain processing technologies.
1. Background

1.1. The Significance of Marketed Surplus of Food Grains

Food grains (cereals, pulses and oil crops) constitute the major sources of food intake in Ethiopia
accounting for 82% of total calorie intake1 and 70% of total food expenditure (CSA, 1988;
Abebe, 2000:260). Cereals alone would provide about 70% of the average Ethiopian’s calorie
intake (Howard, et al, 1995). Other studies report quite high figures2; for example, Shiferaw
(1986) reports that in the eastern part of the country (Hararghe) 98% of food requirements come
from grains only. Similarly, Tesfaye (1989) estimates that grains meet 95% of the food
requirements in the central highlands. Perhaps a reflection of this, the bulk of agricultural
production is dominated by grain production; of the 16.5 million hectares of land that is under
cultivation, 14.6 million (88.5%) is under annual crop production (Tesfaye, et al, 2001). Grain
production in Ethiopia is virtually a smallholding farmer’s activity and yield levels are one of the
lowest in the world. Studies indicate that of the total food grains produced by farmers, only about
a quarter is marketed (Eleni, 2001; Abebe, 2000; Gebremeskel, Jayne, and Shaffer, 1998) with
the bulk of production being retained for on-farm consumption3.

Table 1. Area, Output and Yield levels of Cereals (1974-1999 average)

Area Output Yield (Quintals/ha)

Crop (000 ha) (000 qts) Ethiopia World Best World Average
Wheat 665.57 7,787.27 11.71 57 (Holland) 28
Barley 838.97 9,164.25 11.06
Teff 1,535.83 13,097.21 8.55
Maize 1,033.11 16,063.96 15.48 74 (New Zealand) 30
Sorghum 837.86 10,134.58 11.92 48 (Spain) 13
Cereals (Total) 5,115.00 57,869.41 11.28
Source: (1)Ethiopian Economic Association, Annual Report on Ethiopian Economy, 2002 (for Ethiopia)
(2) Beets, 1990 (for World best and World Average figures).

The concern over marketed surplus of food grains has always been at the center of policy
formulation and implementation in Ethiopia. The previous regime attempted to get hold of
marketed surplus of food grains through captive production processes; that is, by setting up state
farms – the major reason being the difficulty of controlling marketed surplus of food through
markets only. That practice lasted for more than a decade, i.e., between late 1970s and early
1990s. The contribution of state farms was insignificant both in terms of share in total area
cultivated as well as output produced as it constituted only 2.8% of total area cultivated and 3.6%
of total grain production (Abebe, 1990; PMGSE, 1984)4; but it permitted direct access and control
of output in a sense that whatever was produced could be captured directly. There was also a
heavy hand of government in marketing of grains when the state-owned Agricultural Marketing

1
The most important grains in terms of consumption are teff, wheat, and maize, which together constitute
roughly two-thirds of caloric intake in Ethiopia (Alemayehu, 1993).
2
A high concentration of cereals (over 75%) in food intake is a symptom of unbalanced diet. Diets high in
cereals and tubers are low in micronutrients. (see FAO, 1998)
3
Of the total grain production, some 72 percent is retained for on-farm uses (Gebremeskel, Jayne, and
Shaffer 1998).
4
Provincial Military Government of Socialist Ethiopia, 1984. Ten Year Perspective Plan.

1
Corporation (AMC) was set-up in 1976 with unprecedented powers and duties including, among
other things, of fixing grain prices at below market levels and setting of compulsory procurement
quota (see Abebe, 2000: 80-83). It used to run about 2,200 warehouses throughout the country
with a total capacity of over 1 million tons of grains (Alemayehu, 1992). With the change in
government in early 1990s the importance of AMC dwindled as it was downsized into an
Ethiopian Grain Trade Enterprise with primary role to play as market stabilizer through managing
buffer stocks. With pro-market reforms in place, the single most important source of marketed
surplus of food grains is expected to come from the smallholding farmers who also retain a
significant proportion of their output for household consumption.

1.2. Post-Harvest Grain Management as a Missing Link

The crucial importance of ensuring sustained levels of marketed surplus of food both in terms of
sufficient quantities as well as fair prices cannot be overemphasized if development is to take
place for poor countries such as Ethiopia. However, government policy seems to have
concentrated more on aspects of production and marketing proper and less on what happens in
between these two processes. For example, there has been a lot of emphasis and support given for
increased grain production through the package of agricultural technologies and inputs. Indeed,
partly due to these interventions, output for some grains (maize) has increased to the extent that it
caused maize prices to fall dramatically. Following that, there has been an argument whether
market stabilization mechanisms should be introduced to absorb the shock so that depressed
market prices would not have disincentive impacts on the part of producers. This may serve a
short-term purpose, but it also shows that attention revolves around the orbit of production proper
and marketing only.

There seems to exist an assumption that the bulk of food production would be marketed instantly;
this assumption tends to fail to appreciate the fact that farmers produce food grains first of all to
meet their household consumption requirements and then consider marketing. In many cases, the
farming system that integrates crop-livestock husbandry tends to facilitate liquidity preferences
through rearing and selling small ruminants rather than food grains; i.e., with small ruminants
playing as major sources of cash. Hence, the fact that farmers produce food grains not primarily
for sale (which is also evidenced by the low proportion of marketed grains) implies that much of
the production is retained and therefore would not reach marketplace. It also implies that food
grains are retained for longer period of time and that marketing is staggered over several seasons
even though much of the marketing is concentrated in the few months immediately following
harvest; about 79 percent of the annual sales of farmers occur in the period between January and
March5 (Gebremeskel, Jayne, and Shaffer 1998). Taking the 75% retention rate by farmers and
the 79% concentration rate of marketing season between January and March, the proportion of
grains that are marketed immediately after harvest season would be about 20% of the total
production – which also implies that as large as 80% of total production is either retained for
household consumption or marketed in a seasonally staggered fashion.

This situation makes a strong case for sound post-harvest grain management system if food loss is
to be minimized, which should be justified especially for Ethiopia where a great majority of
people is food insecure. It is important to recognize that post-harvest grain management practices
and capacities (not just production and marketing) are consequential for many reasons including
attaining high level of food security objective. Some estimates would suggest the magnitude of

5
This implies that farmers bear the full price risk in marketing their output, without informal mechanisms
to lock in prices and without the financial means to withhold sales to mitigate the fall in prices at
harvest (no forward contracts or credits) (see, Eleni, 2001)
post-harvest loss in Ethiopia to be tremendous; for example depending on the type of post-harvest
handling method, losses could range between 5 and 19% for maize, 6-26% for millet, 6-23% for
wheat, and 5-20% for teff (see, Dereje, 2000). Clearly a better post-harvest grain management
capacity and practice would minimize the extent of loss. This is in addition to the potential
employment and income linkage effects and gains that would result from post-harvest grain
management practices as economic activities. Despite its significance however, there has been
very little effort in terms of studying post-harvest management practices and capacities in
Ethiopia. Those studies that remotely touched upon post-harvest aspects emphasized only on
marketing and understandably their policy recommendations would not go beyond improvement
of transportation, storage and information infrastructure and/or regulatory frameworks (see, for
example, Alemayehu, 1993; Wolday, 1994, 1998; Bekele and Mulat, 1995), with little mention of
processing as an important post-harvest grain management system. Other limited studies (e.g.,
Jonsson, 1972; Dereje, 2000) focus on the engineering and design aspects of storage
infrastructure without any reference to wider perspectives such as food security. Empirical studies
of the sort proposed here, i.e., investigation of the relationships between post-harvest grain
management practices and food security at different levels of analysis, are fundamental for
designing sound food policies. This is what this study aims to achieve.

2. The Problem and the Purpose

It is ironic that the immediate victims of a state of food insecurity, or even worse famine, have
traditionally been farmers who are the very producers of food. Each year hundreds of thousands
of rural households are food insecure and therefore literally depend on food-aid for their survival;
and this is despite weather conditions.

The fact that farm households lay direct claims on their own produces means that they do depend
less on markets for meeting the bulk of their consumption requirements. Production turn over is
mostly only once a year, and at most twice, in which about 95% of production comes from the
main season only. Moreover, long gestation period is involved between planting and harvesting
of crops, which in addition to low productivity levels has constrained grain availability from own
production. Consequently many rural households run the risk of food insecurity for several
seasons of a year. In other words, seasonality of food security follows food production cycles;
that is, a relatively higher level of household food security during the season immediately after
harvest is followed by longer season of food insecurity extending up to next harvest season. The
extent of food insecurity is most severe during seedbed preparation and sowing season, whereas
seasons following harvest are those in which one could find the largest volume of available grain
at household level but also at market levels, since the major marketing seasons are also these ones
as far as peasant households are concerned. Some studies indicate that 79 percent of the annual
sales of farmers occur in the period between January and March (Gebremeskel, Jayne, and
Shaffer 1998). Some of the factors that would attribute to such an instant disposal of food grains
by farmers may include cash needs to effect payments of various dues (to government such as
taxes, fees, loans, etc), social obligations (loans, festivities, etc), or fear of the risk of post-harvest
grain loss.

Such seasonal fluctuations of grain availability both at household as well as market levels and
thus of the level of household food security could be related with farmers’ post-harvest grain
management systems and capacities. In fact, smoothing out of food consumption seasonally, as
well as ensuring stability in food availability in markets, thus in food prices, could be determined
by the extent to which there exists an efficient post-harvest grain management system not only at
a household level but also at macroeconomic level. Unfortunately, this crucial area has not
received the attention it deserves, the reason probably being the often easily held assumption that
what matters after all is production, and that if success could be achieved at the level of
production, then there will be more availability of grains both at the household and market levels,
etc. (see also Goletti and Wolff, 1999). It is interesting to note that ‘the strategy of decreasing
post-harvest losses is more economical because it requires smaller inputs per unit of the final
product than a strategy of increasing production extensively, especially in the short-run’ (Toma,
et al 1990).

If there is one lesson that should be drawn from a recent past grain production and price trends in
Ethiopia, it must be that of the failure of post-harvest grain management systems to sufficiently
respond to increases in production levels so as to stabilize grain markets. Often boom cropping
years are followed by depressed prices. Evidences show that cereal production has increased
between 1999/2000 and 2000/2001 crop year in the order of 19%. Prices on the other hand
declined up to 40% (see MoA, 2001). There were mounting field reports that farmers were
finding it difficult to sell their products since the prices offered were extremely low. This may
partly be due to the fact that agricultural products are not processed before they are sold to the
consumer; they are marketed as harvested without undergoing any transformation process. Such
processes have made storing of agricultural products difficult for farmers since most of the
products are perishable while farmers generally do lack the necessary infrastructures to keep them
for longer time without being damaged. Therefore, what had been witnessed is that, at the macro
level, the poor post-harvest grain management capacities could not live up to the production
expectations, which have led to depressed market prices with the pernicious disincentive effects
on the producers. Similarly, at a more micro-level, the situation depicts one of poor capacity of
the post-harvest grain management systems by farm households in the sense that they tend to
dispose of most of the grains immediately after bumpy harvest at lowest prices of any season in
the year. Such lack of post-harvest grain management capacities that has surfaced both at macro
(national) and micro (farm household) levels are inter-related; but they mirror the extent of what
post-harvest loss might be in terms of physical crop damage, quality deterioration and value
depreciation. Obviously, its implication on food security at national as well as household level is
of paramount significance for a country such as Ethiopia in which food insecurity has become a
structural problem. It is important to examine the post-harvest grain management practices and
capacities of farmers and its impacts on levels of food security.

This paper aims to achieve the following objectives:

a) identify and document the different post-harvest grain management processes and techniques
practiced by farmers and investigate the factors that determine the choice of a given (or a set
of) post-harvest grain management technique over other techniques;
b) examine the extent to which farmers’ perception of risk about post-harvest grain loss
influences their marketing behavior, that is, to explain as to why farmers dispose of most of
their output immediately after harvest and at cheaper prices and explore in what ways would
such marketing behaviors relate with levels of household food security;
c) investigate the relative importance of the potential risk associated with physical post-harvest
grain losses and that of loss in income due to inter-seasonal price variations in explaining
farmers’ post-harvest grain management practices and its impacts on food security.
d) explore the extent to which inter-crop differences in post-harvest risk of loss are important
factors in determining the share of marketed surplus of food grains;
e) highlight on crucial policy issues that relate to attainment of food security objectives through
improvements in capacities and practices of post-harvest grain management both at a micro
as well as macro level of organization.
3. A Conceptual Framework

Post-harvest grain loss is the loss of grains (quality and/or quantity) between the moment of
harvest and consumption of the crop. Reduction in food losses are sometimes considered as the
‘third dimension’ to the world food supply equation, i.e., in addition to increase in food
production and increases in population (Toma, Fansler and Knipe, 1990). Crop losses could occur
at all levels of the post-harvest system, during pre-processing, transportation, storage, processing
and packaging and marketing. Post-harvest grain losses could be classified into two categories,
namely quantifiable (weight loss, economic loss) and non-quantifiable (nutritional loss, energy
loss, and quality loss) (Toma, et al, ibid).

The commonest forms of post-harvest grain management practices include storing, treatment,
processing, and through stock management (loans, sales, purchases, etc). In Ethiopia, grain
storage at farm level is carried out by means of grain pits, underground holes, and sacks. Grain
losses due to poor storage by farmers are reported to range between 11-19% (Kassahun, 2000).
Grain traders also store grains in warehouses with small but varying capacities; moreover, these
are characterized by poor ventilation and dirt floors (Dadi, Negassa, and Franzel 1992). Hence,
poor storage naturally leads to high grain damages from pests and moisture whether at farm level
or anywhere along the marketing channel for that matter. Although storage capacity has increased
since the enactment of market reforms, roughly two-thirds of traders indicate that their storage
facilities are still inadequate in terms of availability, capacity, and location, with 19 percent of
traders reporting in 1998 that they were unable to obtain rented storage space (Gebremeskel,
Jayne, and Shaffer 1998). This inadequacy of storage, combined with the vulnerability of crops to
damage, makes traders unwilling to store stocks beyond the minimum turnover period. Similarly,
pre-damage treatment (aeration, application of pesticides, etc) is seldom practiced by most
farmers, and post-damage treatment is limited to aeration. This may be because of the designs of
the storage facilities not allowing for an easy ventilation or poor capacity of farmers to purchase
pesticides or even non-availability of treatment facilities.

Often what happens between production of a given crop until it reaches consumers’ table is a
black-box. In Ethiopia, consumers buy unprocessed crops and much is actually unchanged since
processing is not undertaken either by producers or middlemen; it is essentially undertaken as and
when grains are prepared for consumption rather than as an important economic activity. As a
matter of fact, it is generally the ultimate consumers who undertake the processing of grains
(milling, etc.) which in a way reflects the addition of no value of substance in space and time
between points of sale by farmers and purchase by ultimate consumers – the sole marketing
function being transportation and temporary storage, that is without subjecting grains to further
transformation. Hence, in view of fast perishability of food grains, the risk of loss involved could
be of quite a higher order of magnitude.

Absence of mechanisms for proper handling, treatment and generally post-harvest management of
grains may be an important factor that may induce farm households to dispose of most of their
grains immediately after harvest thinking that they would buy when needed; however, cash is
often quickly dissipated than grain holdings through other ways6. In fact, this has also intra-
household distributional dimensions; i.e., grain consumption is distributed more fairly among
members of households than cash obtained from grains sales. In many rural areas marketing
(particularly that of goods and services for household consumption) is primarily the domain of

6
For example, there are reports that husbands who usually got direct claims on such cash would squander it
on drinks.
female members (housewives) of households by which they take some small amount of grains,
chicken, some eggs, butter, etc. to the market and purchase in exchange household consumption
items (oil, detergents, salt, spices, etc). Usually such practice takes place throughout a year, in
some places as frequent as once or twice a week. For expenditure involving relatively larger
sums, households resort to small ruminants (sheep, goats), heifers, or grains of larger volume,
which is entirely undertaken by the head of the household who are predominantly men.

In some sense, storing of grains could also be considered as farmers’ choices of saving in kind.
Farmers’ cash holdings may be small, but since saving institutions are non-existent in rural
Ethiopia it may be considered as a close substitute for the later. In addition to prevalence of
missing markets, the convenience of constantly drawing on grain stock for consumption purposes
year round could also be an important factor explaining farmers’ behaviors to save food grains in
kind. Farmers may practice some kind of stock management whereby they may lend/sell out
grains at different seasons of the year to other farmers and get repaid either in kind or cash later.
Hence, in absence of well functioning formal financial institutions, stock management may play a
very important role as post-harvest grain management practice.

The extent of post-harvest losses widely differs from crop to crop; but it could be quite enormous
in the case of some perishable vegetables reaching as high as 100% (Toma, Fansler and Knipe,
1990). It could also be quite significant for food grains and horticultural crops; for example,
Coursey and Proctor (n.d) report that about 25% could be lost between harvest and consumption7.
Cereals and pulses are vulnerable to insect attack at all stages and therefore require continuous
protection. Inadequate storage methods immediately after harvest and before processing add to
the problem, and infestation continues to increase during transportation and long-term seasonal
storage before processing, causing an estimated overall loss of over 30 per cent.8 For particular
crops and countries earlier reports indicate wheat loss by weight in a range of 8 - 52 per cent for
India, 6 - 19 per cent for the Sudan, and 15 - 20 per cent for Brazil (ibid). Experts in this kind of
analysis cite minimum post-harvest losses of 10 per cent for durable crops and 20 per cent for
non-grain staples, perishables, and fish.

Post-harvest loss studies are generally rare in Ethiopia; and even those few available are bent
more on aspects of engineering structures and are not linked with aspects of food security.
However, some preliminary estimates show that post-harvest grain losses could range between 5
and 26% of production, which is enormous given the state of food insecurity in the country.

The relationships between post-harvest grain management practices and food security could be
conceptualized as in Fig. 1. At the household level, food security could be defined by both
availability and access to enough food at a point in time. It is determined by the amount of output
produced (1 in fig. 1), which is a function of resources endowment (4) - which is often identified -
as well as by the amount of post-harvest grain losses (3) and food prices (2) - which is often
neglected. Grain losses could either be in terms of quantity damaged or in quality deteriorated,
but it is related with post-harvest grain management practices (5), i.e., storage, handling,
processing and stock management, which in turn is a function of endowment and access to

7
For horticultural crops, for e.g., in the tropical world, some conservative estimates would suggest that out
of the more than 200 million tons of horticultural crops (including perishable staple foods) produced
annually, a proportion of the order of 25 per cent is lost between harvest and consumption (Coursey and
Proctor (n.d.). The average minimum losses reported for roots and tubers and fruits and vegetables were
16 per cent and 21 per cent respectively; many more 'qualitative' references, not included here, indicate
estimates of 40-50 per cent and above (see http::/www.unu.edu/unupress/food/8f012e/8f012E01.htm.)
8
see http::/www.unu.edu/unupress/food/8f012e/8f012E0b.htm
 resources including credit (6). Output level may also influence the choice of grain management
practices (7), which would determine the magnitude of post-harvest grain losses (8) as well as
farmer’s stock management behavior. The magnitude of post-harvest grain losses, whether actual
or expected, would influence farmers’ behaviors in marketplace (8); that is, in absence of a
working insurance markets, the fear of running high risk of grain loss may induce farmers to
dispose of most of their produces immediately after harvest time and probably repurchase at some
future date. Obviously, this would entail inter-seasonal price discrepancies (9), which would
affect the level of farmers’ income and therefore access to food (2). Hence, the level of household
food security may be affected directly by the magnitude of post-harvest physical grain loss or
indirectly by loss of grain prices due to quality deterioration and inter-seasonal price variations.

Production Process Food security:

Endowment & 4 Output • Availability
access to
level 1 • Access
Resources (e.g.
credit)

3
7
6 2
Post Harvest
Grain Management
Practices Grain Loss
• quantity
• quality
PHGMP 5
9
• Storage+treatment
• Processing 8 Price variations:
• Stock management 8
(Sales-repurchase)
9
• Instant grain sales
• Grain Repurchase

Fig. 1. A Conceptual Framework: Post-Harvest Grain Management as determinant factor for household
food security levels.

Given resource endowments and output level, poor post-harvest grain management practices
would result in low level of food security through: (a) output effect, that is, reducing grain
availability due to physical losses and involuntary sales; and (b) income effect, that is, reducing
access to food due to low prices received when grains are sold immediately after harvest and/or
higher prices when grains are repurchased, or quality of grains deteriorates so that they fetch
lower prices. It could be hypothesized that:.

a) Farmers’ perception of risk of post-harvest grain loss and seasonal liquidity constraints
influence their marketing behavior, that is, it explains as to why farmers dispose off most of
their output immediately after harvest and at cheaper prices, eventually producing sub-
optimal outcomes and resulting in low level of household food security.
b) The potential risk associated with physical post-harvest grain losses are more important than
the potential risk of loss in income due to inter-seasonal price variations in explaining
farmers’ post-harvest grain management practices and its impacts on food security.
c) Inter-crop differences in post-harvest risk of loss are important factors determining share of
marketed surplus of food grains; that is, the crops which are mainly marketed instantly are
those for which risk of post-harvest loss are the highest.

4. The Data and Methods of Analysis

The study basically employed a household level analysis. Relevant secondary information were
collected from the archives of various organizations (e.g., CSA). Primary data were collected by
conducting field surveys. The most important set of information collected include resource
endowment patterns, socio-economic characteristics, production, post-harvest infrastructure,
inventory of storage, processing, stock management practices, food availability, sales, purchases,
prices, functions and importance of rural institutions (formal or non-formal) such as credit,
saving, cooperatives, etc.

Three areas that are dominantly producing maize (Bako, west of Addis), wheat (Chilalo, south of
Addis), teff (Ada, east of Addis) were included in the study for the obvious reason that these
crops are the major food items (together they would constitute for 2/3 of total calorie intake).
Although these three crops have been considered for selecting target sites, data wee collected for
all other crops grown in the area and analysis was done accordingly. Two peasant associations
from each region were purposively considered - one with a superior potential and another with
moderate potential, making a total of 6 peasant associations. The list of peasant households that
was furnished by peasant association served as a sample frame for stratification purposes and
ultimate selection of sample households. Peasant households were randomly selected after
excluding the food-deficit (or poorer) ones. This is because post-harvest grain management
assumes the existence of surplus production over and above immediate consumption, which the
poorer households cannot afford. 50 peasant households from each peasant association were
selected; hence, a total of 300 households were included in the survey.

Primary data were collected using questionnaire, interview guides, participatory discussions with
focus groups and informants.

Both descriptive (both qualitative and quantitative) and econometrics methods9 of data analysis
have been employed. Farmer’s post-harvest grain management systems, as well as production
levels, past post-harvest grain losses and their causes, traditional loss minimizing strategies, etc.
are described in detail.

We wanted to know what factors would influence farmers’ grain marketing behavior; in
particular their decision to dispose of most of their grain produces immediately after harvest, and
if the risk of post-harvest loss is significant in explaining such a behavior. Also we wanted to

9
For specification of the econometric estimation model, see Annex.
know if and in what ways these are inter-related with household food security issues. Hence, our
dependent variable is farmer’s instant disposal behavior (Di); while the explanatory variables
would be risk of post-harvest grain loss (z) characteristics of the post-harvest grain management
practices (m), grain prices (p), household characteristics (w), and farmers’ resource endowments
(x).

6. Empirical Results

7.1. Description of Survey Findings

7.1.1. Farm Resources and Grain Production

As explained above, grain production in Ethiopia is almost entirely based on rain-fed agriculture
and is characterized by a dominant harvest (meher) around November and December and a
secondary harvest (belg) around May and June. Of course, this is a general description and there
exist variations among different agro-ecological zones in relation to production cycles. The study
areas depend on meher production of cereal crops.

Land is the major farm resource in the study areas, and it could be accessed through two ways:
land allocations by peasant associations (PAs) and/or through informal land markets. About 98
per cent of the sample households acquired land from their peasant associations. In addition, 51
per cent of the households acquired some of their farm plots through either sharecropping or
rental arrangements. On the other hand, few sample households (5.4%) shared/rented out their
farm plots during the year under consideration. The low proportion of farmers who shared out/
rented out their land compared with those who shared in/ rented in land is because our sampling
procedure was biased towards the relatively better-off farmers who usually seek for additional
lands. As discussed in the methodology section, the selection of better-off farmers was deliberate
because the basic research problem, post harvest grain loss, assumes ‘surplus’ production and
therefore the selection of surplus producing farmers.

On average, cultivated area per household was about 2.7 hectares. This is on the higher side,
which might perhaps reflect the importance of informal land markets; i.e., since the farmers
included in this study are the relatively better-off ones, there is a tendency for them to have leased
in land from other resource-poor ones. Significant variations exist among the three sites vis-à-vis
the average area allocated to production of different crops. Table 1 shows the cultivated area
under major crops included in the study during the 2002/03 main season. Apparently, wheat, teff
and maize respectively take the lion’s share of total cultivated land in Hetosa (65%), Ada (47%),
and Bako (75%).

Table 2. Average Cultivated Land Area per Household of by Major Crops (hectares)

Hetosa Ada Bako

Crops Area % of total Area (ha) % of total Area (ha) % of total

Teff 0.052 1.79 1.17 46.54 0.20 7.69

Wheat 1.88 64.56 0.80 31.82 0.00 0.00
Maize 0.27 9.27 0.024 0.95 1.95 75.00
Sorghum 0.14 4.81 0.0 0.00 0.23 8.85
Other crops 0.57 19.57 0.52 20.68 0.22 8.46
 Total 2.91 100.00 2.51 100.00 2.60 100.00

Livestock are the other important assets owned by the sample households. Almost all sample
households own some livestock. In fact, as mentioned above the study areas are characterized by
mixed farming in which farmers rear some animals while producing crops. Various types of
animals are raised in the study areas including large and small ruminants, pack animals (equines),
and poultry. An average livestock holding, excluding poultry, was found to be 11 animals per
household, which is equivalent to 8.5 tropical livestock units10. Almost every household owns at
least two milking cows and three oxen, again a reflection of a relatively better resource
endowment position of sample households.

Livestock are important income sources for the sample households. More than one-half of the
households sold either their livestock or livestock products to generate cash revenue. On average,
a household generated about 460 birr from the livestock sub-sector in the year under
consideration11. Households enumerated various reasons for selling livestock and their products;
the main ones included repayment of input loans (28.1% of households), to replace old animals
(19% of households), to purchase house equipment and furniture (14.3% of households), and to
cover expenses related to education and health (9% of households).

7.1.2. Post Harvest Grain Management Practices and Losses

(i) Farmers’ Grain Management Practices

Unlike what is observed in mechanized farms where mowing and threshing could be undertaken
simultaneously, grain harvesting by Ethiopian farmers takes place manually and involves hand
mowing of crops using sickles and later threshing by letting a group of animals trampling upon it
on a leveled ground until the grains and the shaft are separated. Harvesting of maize is unique in a
sense that it is undertaken by removing the ears by hand. Several days could elapse before mowed
crops are threshed; i.e., crops stay piled for some time either around homestead or in situ before
threshing. In most cases losses during harvesting process could be significant (due to attack by
rodents, moisture, livestock, etc.).The length of time in which crops are kept piled varies among
crops. Usually crops that are susceptible to loss (such as pulses) do not stay long while some
other crops (such as teff) can be piled for longer time. It also varies depending up on the status of
the household vis-à-vis food and financial requirements; i.e., those with immediate need would
thresh their crops instantly.

Farmers use various methods and types of facilities to store their crops. Maize could be stored
either shelled or unshelled. In the latter case, maize is shelled manually in a staggered manner
when required for consumption or sale. Shelling of maize is also done by using hired machines
especially when production is large. The traditional grain stores identified in the study areas
include gotera (grain pits), bags (made of polyethylene, sisal or goat skin12), earthen pots and
some others. More than 70% of the respondents used polyethylene bags and sacks made of sisal
while about two-third of them use gotera (Table 2). The latter is usually preferred to store crops
for longer period of time and also in a larger quantity. Some farmers also use pots and other small

10
TLU is a standardized measure of livestock numbers. (see Abebe, 2000).
11
This amount doesn’t include the revenue generated indirectly from livestock use, for instance, by renting
out draft animals to others.
12
This traditional container is commonly known as silicha and it is estimated to hold 50 to 60 kilograms of
grain. To produce silicha, a special skill and care is required when the goat is skinned.
 granary entirely made of mud. Slight variation has been observed among the three sites vis-à-vis
the proportion of farmers using each type of storage; gotera is widely used in Bako whereas bags
(or sacks) are the common grain stores in the other two sites, perhaps suggesting that farmers
store grains only temporarily in Ada and Hetosa.

Table 3: Types of Stores Used (by number & percentage of respondents)

Hetosa Ada Bako Total

Type of Storage No. % No. % No. % No. %

Gags/ sacks 93 94.9 84 84.0 33 33.0 210 70.5

Gotera 73 74.5 49 49.0 80 80.0 202 67.8
Pots 14 14.3 9 9.0 5 5.0 28 9.4
Underground pits 1 1.0 - - - - 1 0.3
Others 4 4.1 32 32.0 18 18.0 54 18.1

Farmers acquire their grain stores in different ways. The majority of gotera users construct it by
their own, which is a reflection of weak dependence on markets. In most cases, goteras are
constructed from flexible materials (such as bamboo trees) and are mud-walled and thatched-roof.
Although not common, farmers could purchase these traditional granaries from local markets.
The average owner-estimated monetary value of gotera is about 62 birr with the figures ranging
from 10 birr to 500 birr. Virtually, the value of a store depends upon its capacity and materials
used for its construction. Polyethylene bags and sacks (made of sisal) are commonly purchased
from local markets. In fact, farmers use fertilizer bags to store grains when the bags are empty of
the fertilizer input. The average estimated monetary value of these containers is 3.25 birr. We
have also observed some higher figures (up to 35 birr) probably indicating the value of silicha
which is relatively durable. The other small traditional granaries are mostly self-made. These are
usually used to store seeds or grains produced in smaller quantity and their estimated values range
from 4 birr to 30 birr.

Table 3 shows the number of grain stores and their storage capacities owned by peasant
households. As could be observed, the average storage capacity of the sample households was
3.27 metric tons, which can only hold food grains for one or possibly two production seasons. If
all kinds of storage mechanisms (temporary and fixed) are considered,13 few households could
store more than 5 metric tons of food grains. If, on the other hand, bags and sacks are excluded
from the analysis, the average storage capacity per household could be reduced to 2.52 metric
tons. This is quite low as compared to the average volume of total grain produced (4.39 metric
tons) indicating the existence of storage constraints in the survey areas. To put it in a different
way, about 79 per cent of the farm households didn’t have adequate fixed stores for their produce.
A disaggregated analysis by site may show a more detailed picture (Table 4). While fixed stores
could absorb more than three-quarters of total output in Hetosa and Bako, these are not widely
used in Ada possibly arising from the fact that the major crop in Ada (teff) is relatively less
susceptible to storage pests (particularly weevils), hence could be readily stored in bags, or even
sold instantly. That this wasn’t the case is clearly apparent, with respect to Hetosa and Bako
where the bulk of grain production respectively is wheat and maize which are more susceptible to
storage pests.

13
Strictly speaking, bags and sacks are not permanent grain stores. Rather they are used to store grains which
would be disposed to the market instantly.
 Table 4. Number and Capacity of Grain Stores Owned by Households
Number Capacity (MT)
Storage type Mean Maximum Mean Maximum

Gotera 1.6 8 2.105 26.0

Bags/sacks 12.9 200 0.739 11.0
Pots 0.5 7 0.025 2.5
Others 2.7 96 0.385 20.0
Total 3.27 37.0

Table 5. Capacity of Stores as Compared to Total Grain Production

Total Capacity (MT) Total output
Study Sites All Stores Percenta Fixed Stores Percenta (MT)
Hetosa 4.74 102.6 3.52 76.2 4.62
Ada 1.69 39.4 0.78 18.2 4.29
Bako 3.41 80.2 3.35 78.8 4.25
All Sites 3.27 74.5 2.54 57.9 4.39
a
Figures represent storage capacity as percentage of total grain produce

Grain losses could arise either from poor post-harvest handling or from production over and
above the capacity of the available stores, or both. In case production exceeds total storage
capacity, two options are possible to solve the problem: instant disposal of the excess produce
(e.g., sales, loans, etc.) or increase storage capacity. If grain markets are demand constrained, or
supply price elastic as it is true in the case of grains, instant sale of grains immediately after
harvest will have a price reducing effect. Hence normally this option seems to be less preferable
for producers. In the study sites, however, the majority of the sample farmers (about two-third)
resorting to selling their grains whenever production is in excess of their storage capacity.

Grains are stored both outside as well as inside residential houses. The majority of the
respondents (71.6 per cent) reported that they locate their stores inside their homes; the reasons
for doing so include the need for protecting grains from theft and moisture as a result of rain
during post harvest seasons. Similarly, grains are stored outside but very close to home because
of the advantage of outdoor environment that provide better aeration to keep crops for longer
period of time without damage due to insect pests. Locating grain stores outside home may also
have another advantage; the living rooms would be free which would otherwise be crowded of
grains especially after harvest14. It is common to locate gotera outside residential houses as it
occupies larger spaces as compared to other stores.

(ii) Post Harvest Grain Losses

The majority of the farmers (93.3%) perceived an imminent risk of grain lose due attack by
storage pests and/or other factors if they stored their crops for longer period of time. The
minimum per household expected average amount of loss is about 150 kilograms (for Ada site,

14
In most cases peasants live in traditional huts having single room in which the whole family dine and sleep;
the same room may also be used to prepare food.
where teff – the less susceptible crop - is the dominant crop) for all grain crops whereas the
maximum per household expected average amount of loss is more than 800 kilograms (for Bako
site, where maize – the most susceptible crop – is the dominant crop). Also as could be observed
from Fig. 3, there are large dispersions about the median post-harvest grain loss in Hetosa and
Bako as compared to Ada, again depicting the inter-crop differences in post-harvest grain losses.
In some cases the magnitude of loss could even go as high as 2000 kilograms.

The respondent farmers reported to have actually lost large quantities of their produce due to
various factors. The per household average actual loss (as opposed to expected loss) was about
500 kilograms of total grain output during the previous year, which is equivalent to 12 per cent of
the average total grain produce of the sample households. Farmers in Bako and Hetosa areas
reportedly lost about 700 and 620 kilograms of crops respectively due to post harvest damages,
while the figure for Ada was much less. Of course, these variations reflect differences in
cropping patterns among the three sites and therefore the high degree of susceptibility of some
crops such as maize and wheat than the others such as teff. The reported amount of loss is quite
substantial and definitely suggests that what matters is not only getting success in raising yield
levels but also ensuring the entire produce reaches the consumers’ table without loss. Hence,
yield enhancing interventions cannot be considered in isolation from minimizing post-harvest
grain losses.

Fig. 3 Box-plots Distribution of Expected Loss if Grains were to be Stored Until the 4th
Quarter.
4000

3500

3000
Expected loss (Kg)

2500

2000

1500

1000

500

0
N= 98 100 99

Hetossa Adda Bako

Sites
Table 6. Estimated Grain Loss During 12 Months Preceding the Survey Period (kgs by site)
Study Areas Mean SD CoV
Hetosa 618 438.6 0.71
Ada 237 308.9 1.30
Bako 707 1118.6 1.58
All Sites 520 744.0
 As could be seen, maize is the most affected crop i.e. on average a maize producing household
lost 500 kilograms of maize during 12 months preceding the survey time. The average loss by
wheat producers could be around 250 kilograms. On the other hand, as expected, teff was reported
to be less susceptible to post harvest loss enhancing factors as compared to wheat and maize i.e.
on average, the sample households lost about 106 kilograms of teff after harvest for the year
under consideration. Taking the nominal average producer price15 that prevailed during the year
2002 for the three crops, the amount of loss in birr could be estimated to be 244, 185, and 360 for
wheat, teff, and maize producers respectively. As could be seen the variations have narrowed
down owning to price differentials; teff and wheat fetched higher prices than maize.

It could also be noticed that estimated grain losses for the three crops are less than the estimates
for the regions (all crops); perhaps such variations would suggest that post-harvest losses of other
crops such as pulses might have been larger.

Defining shelf-life of grain crops as the number of months in which they could be stored without
facing loss in quantity and quality, it may vary depending up on the type of storage infrastructure,
type of crop, weather condition, treatments made to reduce loss and so on. In the study sites, grain
crops can reportedly be stored on average for about 7 months without loss. Of course, the shelf-
life may increase to about 10 months if grains are treated with chemicals. It was reported by the
sample farmers that grains, particularly maize, would turn virtually into powder just in few
months due to weevils’ attack if not treated with insecticides. The largest average shelf-life was
reported for tef (with or without chemical treatment) and the smallest was reported for maize
(without chemical treatment) (Table 6). It is also apparent from the figures that maize can only be
stored for a relatively shorter period of time (half year) even if it is treated with chemicals.

Table 7. Estimated Shelf-life of Selected Crops With and Without Chemical Treatment
Estimated Shelf-life (Months)
Crop With chemical treatment Without chemical treatment
Teff Not applicable 19.0
Wheat 11.5 6.4
Maize 6.9 2.8

There are several causes for the observed post-harvest losses experienced by peasant households.
Attacks from weevils and rodents, and moisture or growth of molds are among the factors and
agents contributing to the losses. Among these, weevils’ attack was reported as the most
important and commonest cause (Table 7). Rodents and moisture and/or growth of molds are also
important in areas such as Hetosa and Bako.

Table 8. Respondents’ Identification of the Causes for Grain Losses

Hetosa Ada Bako All Sites
Causes (N=98) (N= 100) (N=100) (N=298)

Rodents 28.2a 18.9 59.1 30.8

Weevils 90.4 82.2 95.3 87.7

15
Producer Prices were computed from CSA various reports; average producer prices for 2002 for wheat, teff
and maize were 100.2, 174.6, and 71.9 per 100 kilograms.
 Moisture/molds 22.6 11.4 45.7 22.5
Others 1.1 1.9 9.4 3.3
a
figures indicate percent of respondents who reported each cause of grain losses.
b
number of respondents

The most important techniques used by farmers were chemicals and aeration. The majority of the
sample farmers (83.2%) used chemicals to reduce grain losses that would result from pest attack
because storage losses without chemical treatment can be substantial. Chemicals are usually used
to treat maize and wheat since these crops are among the most susceptible to weevils’ attack as
indicated above. In addition, a quarter of farmers reported that they use aeration. Grain
management practices such as lending to others, selling and repurchasing latter are not widely
used. Some farmers also reported that they practiced various indigenous methods to reduce post
harvest grain losses due to pest attack. For instance, in Bako and Hetosa areas grains (especially
maize and wheat) are mixed with chilly pepper to prevent storage pests. Farmers hold the opinion
that the pepper has a repelling effect on pests hence that they would constitute effective measures
to prevent pest attack. Similar observations are reported elsewhere (see, Blum and Abate, 2002).
However, the effectiveness of such practices is yet to be confirmed by scientific research.

Table 9. Techniques Used by Farmers to Reduce Grain Losses

Percent of households
Techniques Used
(N = 298)
Aeration 25.1
Use of Chemicals 83.2
Sell and repurchase latter 0.8
Lend to others 1.1
Others (including indigenous techniques) 8.5

Moreover, farmers were asked if they knew techniques other than those they have been practicing
so far. The objective was to explore whether farmers have some options by which they could
reduce grain losses. While the majority of them (71.4%) replied that they didn’t know any other
technique, the remaining nearly 30% reported that they in fact were aware of the existence of
various post-harvest grain management techniques. However, they couldn’t apply them due to
various reasons. For example, more than fifty percent of the interviewed farmers didn’t utilize
those ‘other’ techniques because they perceived that the existing techniques they have been
practicing were reportedly found to be more effective in reducing grain losses than other
techniques.

7.1.3. Grain Marketing Practices and Patterns

Almost all farmers (99%) reported that they sold some crops under normal conditions. Some
farmers also did purchase grain crops; about 40% of the respondents purchased crops by making
claims on their previous savings. Usually, crops are sold in a staggered pattern across the quarters
of a year. Table 9 presents the proportion of grains marketed to total output produced. As could
be observed the proportion of marketed surplus of grains to total grain output for Hetosa, Ada and
Bako were respectively 40%, 63% and 48%. With the exception of Ada, more than one-half of
total grain production does not reach market place; it is consumed within the farm household. The
fact that farmers do produce large assortment of crops and sold some proportion (not most of it)
mirrors the extent of production orientation towards household consumption rather than for sale.
On the other hand, the proportion of total production that is marketed, i.e., the extent to which a
given crop resembles a “cash crop,” varies among crops as well as among sites. A considerably
higher proportion of total productions of the major crops, i.e., wheat, teff and maize, were
marketed compared to other crops in Hetosa, Ada and Bako sites respectively. Obviously these
three crops are dominant in terms of area cultivated as well as output produced in the respective
areas. For example, 65% of the total cultivated area in Hetosa was allocated for wheat production;
and about 42% of wheat output was marketed. Similarly, of the total cultivated land in Ada, 47%
was allocated to teff, while 32% was allocated to wheat production; the corresponding figures for
proportion of marketed output of these crops were 67% and 57.4% respectively. Similar pattern
could also be observed in Bako where 75% of land cultivated was allocated to maize production,
and 44% of maize output was marketed.

Table 10. Percentage of Crops Marketed in Total Output Produced

Crops Hetosa Ada Bako
Tef 0.0 67.0 3.9
Barley 0.0 14.1 ---
Wheat 41.9 57.4 ---
Maize 5.2 0.0 44.0
Sorghum 1.5 --- 12.6
Horse bean 1.8 14.6 0.0
Field Pea 1.8 11.2 ---
Chickpea 0.0 47.4 ---
Others 0.0 31.9 29.1
All Crops 39.7 63.2 48.3

On aggregate levels, proportion of actually marketed output during the first quarter of 2002 was
about 51% of total marketable surplus for the entire year. The average percentage of crops sold
declines as one moves from the first quarter through to the forth quarter (Fig. 4a). On the
contrary, producers’ grain prices tend to rise as one moves from the first quarter through to the
third and in some cases even during the fourth quarter (Fig. 4b).

Fig. 4a. Movement of Crops Sales by Quarter (Percent of total sales)

 70

60
Percentage Crops Sold
50
Hetosa
40
Ada
30
Bako
20

10

0
1st 2nd 3rd 4th
Quarters
Fig. 4b. Movement of Producers’ Prices of Selected Crops (2002)

West Shoa, Bako Area East Shoa, Ada Area

2 2.5
1.8
1.6 2
P ro d u c e rs ' p ric e

P ro d u c e rs ' P ric e
1.4
1.2 Tef 1.5 Tef
1 Wheat Wheat
0.8 Maize 1 Maize
0.6
0.4 0.5
0.2
0 0
Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4
Quarter Quarter

Arssi, Hetosa Area

1.8
1.6
1.4
P ro d u c e rs ' p ric e

1.2
Tef
1
Wheat
0.8
Maize
0.6
0.4
0.2
0
Q1 Q2 Q3 Q4
Quarter

Also a considerable variation exists among the three sites with respect to the proportion of crops
sold during the first quarter. In Hetosa and Bako sites, for instance, the average figures are
apparently higher as compared to the Ada site indicating that farmers in the former two sites sold
the bulk of their crops immediately after harvest (Table 10). It is worthwhile to note that farmers
are unorganized in their sales decisions; hence they lack bargaining power while dealing with
traders who have better resources, information and coordination to influence market outcomes.
Table 11. Crops Sold as Percentage of the Total Crop Sales in a Year (by Quarter)
Percentage Crops Sold
Quarter
Hetosa Ada Bako All Sites
1st Quarter (January-March)
57.6 35.8 59.5 50.9

2nd Quarter (April-June)

15.9 35.4 24.9 25.5

3rd Quarter (July-September) 12.3 11.4 13.9 12.5

4th Quarter (October-December) 14.6 17.6 1.8 11.1

Such a general pattern of seasonal price fluctuation is apparently known by farmers. The farmers
reported that they were in fact aware of the fact that prices would increase after the first quarter,
especially during the lean season. Then, it would be interesting to ask the question that, if farmers
actually expect that grain prices would rise during the lean months, why would they want to sell
large proportion of their marketable surplus just immediately after harvest when grain prices are
the lowest in the year?

One important reason could be the high demand for cash during the first quarter. That particular
season is the time when farmers need money to settle various types of dues including social
obligations (such as marriage, religious festivals and ceremonies), input loan repayment,
payments for land tax etc. For instance, the interviewed farmers reported that about 33 per cent of
their marketed grains during the first quarter went to finance repayment of input loans and/or pay
land taxes.

The other important reason might be fear of risk associated with post harvest grain losses. As
indicated earlier, the majority of the interviewed farmers perceived that post-harvest problems are
matters of grave concern.

The commonly held view, which apparently happens to be true most of the times, is that crop
prices would be the lowest during the first few months following harvest and start to rise
monotonically until they reach maximum during the months of July-August after which they start
to decline. The decline might be associated with expectation of harvest in few months time. But,
what happens to the pattern of prices during seasons following a bumper harvest? It
monotonically declines including during the months in which it normally peaks and gets even
depressed further with the expectation of a better harvest that is forthcoming. This is what
actually happened in the year 2001. (See Fig. 5).

The year 1999 was followed by an expectation of good harvest in 2000. Prices in 1999 were
consistently increasing from January up to September for all crops after which they started to fall.
Maize price dropped earlier than that of wheat and teff because of early harvest season for maize.
But the decline of prices after October in all crops is associated with expectation of next harvest
in 2000.

Now compare this with what actually happened to prices in 2001. As could be observed it is the
reverse of the previous story. Instead of rising, prices continuously declined and in fact started to
pick-up only after October with expectation of bad harvest the following season. As could be
noted, that year was a bad year in terms of harvest which brought famine in much of 2002/03.
This illustrates the extent of market failures. The crisis is both when harvest is good as well as
when it is bad. In times of good harvest, especially when this is matched with expectation of good
harvest next season, grain prices tend to be depressed; hence storing of grains might not be an
advisable option since prices might not get better even during lean seasons. On the other hand,
during ‘normal years’, storing of grains until the third quarter might pay if the risk of grain loss is
minimized below market risks.

When grain prices are substantially reduced, it discourages farmers to apply chemicals to prevent
pest attacks, thus leading at times to substantial crop decays. This is because the marginal benefit
as a result of chemical treatment could be less than the marginal cost of the treatment. For
example, during 2001/02 crop season, maize prices were reduced by up to 80% as a result of an
increase in output by about 40%. Under such circumstances crop prices might not be attractive
enough for farmers to pay even transport charges to take their grains to the nearest market place
for sale let alone realizing profits.
Fig. 5a. Monthly Average Wholesale Prices (birr/100kilograms. 1999 and 2001)

Teff Wheat
teff 2001 teff 1999 wheat 1999 wheat 2001

260 350 250 140

250 300 120

200

240 250 100

150
230 200 80

220 150 60
100

210 100 40

50
200 50 20

190 0 0 0
ri l y e y st l y
a ry a ry a rc
h
ap ma ju n ju l gu be be
r
be be
r a ry a ry a rc
h p ri ma
y
ju n
e ju l u st be er e er
ja n
u ru m au tem to m
cem ja n
u ru m a
au
g
tem tob mb b
fe b oc ve fe b oc ve cem
se p no de sep no de

Maize
maize 1999 maize 2001

180 90

160 80

140 70

120 60

100 50

80 40

60 30
au g u st

novem ber
a p ril
ja n u a ry

fe b ru a ry

ju ly
ju n e

s e p te m b e r

decem ber
m a rc h

m ay

o c to b e r

40 20

20 10

0 0

Source: Central Statistical Authority, various issues.

7.2. Regression Results

Table 12 presents the tobit regression results16. Non-bag storage capacities hold the right sign but
it was found to be not significant. Education also was not found to be significant in explaining
variations in the dependent variable. Similarly, farmers’ expectation of grain loss, should grains
were to be stored until the 3rd quarter, carries the right sign but found to be not statistically
significant.

Table 13 Regression Results (Tobit estimates – dependent variable is proportion of sales

during the 1st quarter to total sales during the year)

Parameter Coefficient T-ratio

Constant 0.7694* 6.296
FSIZE 0.0135** 2.119
SEX -0.1407*** -1.926
EDUC 0.0014 0.303
CHEM -0.0762** -2.440
STCAP -0.0011 -0.183
TLU -0.0084** -2.547
PRICDIF -0.3026** -2.429
EXPLOSS 0.0984 1.081
TAXLOAN 0.0001* 3.150
Sigma 0.2598
Number of samples 292
Log likelihood function -30.9562
*, **, *** significant at 1%, 5% and 10% levels respectively.

As could be seen a total of six variables have turned out to be significant explaining the variations
in proportion of crops sales during the 1st Quarter to total sales by households. Moreover, with the
exception of the SEX variable, the signs of all of these variables are also according to
expectations, hence leading to the following conclusions.

(i) The pressure on households triggered by larger family size to meet non-grain purchased
consumption needs tends to force farmers to instant crop sales;
(ii) Female household heads are more susceptible to instant crop sales during 1st quarter than
male-headed ones. This may be ascribed to fewer options available to female headed
households in terms of resorting to other sources of cash incomes (e.g., livestock sales);
(iii) Application of chemicals (insecticides) reduces the risk of pest attack hence that of the
tendency to instant sales;
(iv) Livestock ownership reduces the pressure on cash needs as households tend to resort to
livestock sales and of its products rather than crop sales;
(v) inter-crop differences are important explanatory variables for households’ behavior in
sales decisions; i.e., farmers who mainly grow maize (more susceptible crop) and wheat
tend towards instant sales than those who mainly grow teff (less susceptible crop); and
(vi) Households’ immediate cash needs to meet various obligations explain the reason as to
why they dispose of their grain crops instantly at cheaper prices.

16
for details see footnote
8. Conclusions

8.1. Summary of major points

Ethiopia is a structurally food-deficit country. It has also been a food-aid dependent country.
Production of food crops in sufficient magnitude is a strategic objective for the country’s food
policy (e.g., for achieving food security both at national and household levels, industrial
competitiveness, etc). Post-harvest loss is an impending problem – it reduces the already
produced output both physically and in terms of quality deterioration and thus leading to value
deprecation. In a sense it is a wasted effort and wasted resource. Fearing such an impending risk,
farmers tend to sell larger proportion of their marketable surplus during the season that
immediately follows harvest when prices are actually depressed. Despite its strategic significance,
however, not much attention was given to post-harvest grain management practices. Studies are
not generally available.

Efficient and effective grain management practices minimize post-harvest losses both at
household, community and national levels. Their additional advantage is that is that apart from
reducing loss, they generate additional employment opportunities, add significant value on
products, maintain quality, enhance competitiveness in market place (of producers, the country at
large), maintains market stability, etc.

The purpose of the study was to identify such post-harvest grain management techniques as
practiced by farmers and determine the extent to which farmers’ perception of risk (market and
post-harvest grain loss) influences their grain marketing patterns. Three areas that are known for
the production of the major food crops in Ethiopia were selected for the study; these include Ada
for teff, Bako for maize, and Arssi for wheat. A survey on a total of 300 sample farm households
drawn from six peasant associations was conducted to collect the relevant data. Both descriptive
and econometric (tobit regression) analysis of the data was conducted.

Apparently crop losses are not limited to just post-harvest; in fact substantial grain losses also
occur before and during mowing and threshing of crops. However, the scope of the present study
is limited to only what happens after grain crops have been harvested. The findings of the study
are summarized as follows.

Storage capacities are limited – fixed structures have capacities of about 58% of average volume
of total output; about 80% of households reported that they didn’t have adequate fixed storage
structures, and use temporary packing materials such as sacks as storage. As a result, as large as
two-thirds of the households resorted to instant sales when production exceeds storage capacity.

There is a widespread perception among farmers that post-harvest grain loss is an imminent risk.
Expected post-harvest grain losses could vary substantially from a minimum of 150 kilograms to
over 2000 kilograms depending on crops and seasons. Actual reported post-harvest grain losses
during the year preceding the study period was reported to have varied from about 240 kilograms
in Ada area (where teff is the dominant crop) to about 700 kilograms in Bako area (where maize
prevails). On average about 12% of grain production was reportedly lost after harvest and while
in stores – assuming that this portrays what is obtaining at a national level, this amount of loss if
prevented could substantially narrow down the deficit in food balance.

Some crops such as maize cannot stay longer (about 3 months only) before they are attacked by
storage pests (notably weevil). In absence of processing, treatment with chemicals and other
indigenous techniques are practiced to reduce grain losses due to pest-attack. Aeration and
treatment with chilly pepper are the most common indigenous techniques used by farmers. Other
stock management techniques (lending to others, etc) exist but they are less important.

The proportion of marketed surplus of grains to total grain output for Hetosa, Ada and Bako were
found to be respectively 40%, 63% and 48%. A considerably higher proportion of total
productions of the major crops, i.e., wheat, teff and maize, were marketed compared to other
crops in Hetosa, Ada and Bako sites respectively. On aggregate levels, proportion of actually
marketed output during the first quarter of 2002 was about 51% of total marketable surplus for
the entire year17. The average percentage of crops sold declines as one moves from the first
quarter through to the forth quarter. On the contrary, grain prices tend to rise as one moves from
the first quarter through to the third and in some cases even during the fourth quarter.

Tobit regression was estimated to identify the factors that explain why farmers would resort to
instant sales and at cheaper prices. That is, proportion of actual sales during the first quarter
(following harvest) was regressed on a number of explanatory variables including family size, sex
of the household head, education level achieved, chemical usage to reduce pest attack, storage
capacity, livestock ownership, expectation of loss, inter-seasonal price variations, and financial
dues to the government including taxes and repayment of input loans.

Results indicate that instant sales of grains immediately after harvest are triggered by temporary
but immediate liquidity preferences to meet purchased items for the family members in absence
of or limited sources of cash other than crops sales, and by an impending risk of post-harvest
grain loss and the limited capacity to prevent it. Hence, family size, female headedness, and tax
and loan repayment schedules that coincide with the season, were found to be significant and
positively associated with instant crops sales. On the other hand, livestock ownership and the
capacity to apply chemicals (insecticides), both indicators of wealth, were also found to be
significant but negatively associated with instant crops sales. Cropping patterns suggest that those
who cultivate more susceptible crops to pest-attacks (e.g., maize) are most likely to dispose off
their crops immediately after harvest at cheaper prices than those who cultivate less susceptible
ones (e.g., teff).

8.2. Policy Implications

As indicated above, traditionally there has been a policy focus on production and marketing of
food grains almost to a complete neglect of post-harvest grain management practices. This is true
both at macro and micro levels of organizations. Not surprisingly, to date there is very little
capacity of an efficient and effective post-harvest management system in place both at national as
well as household level. As a result, grain markets are characterized by high inter-seasonal as
well as inter-temporal volume and price fluctuations. On the one hand, there is a high incidence
of food insecurity in the country. On the other hand, an important opportunity seems to have been
lost by way of reducing post-harvest losses.

Hence, it is important to consider post-harvest grain management as strategic policy concern. In

particular, it is significant that policy makers raise their awareness of the state of affairs and
appreciate the importance of post-harvest grain management practices not just from the

17
The figures for marketable surplus of grains are significantly higher than the national average reported
elsewhere (i.e., 26%), whereas the percentage of actually marketed grain crops during the 1st quarter
immediately after harvest is lower than the national average reported elsewhere (79%). This reflects the
upward bias in the selection of regions and sample households which was on purpose.
perspective of reducing losses, but also equally important, from the view point of considering it
as an economic activity with employment, value and income linkages.

Liquidity constraints and impending risk of post harvest grain losses are the two most important
factors explaining farmers’ tendencies to instant sales of their grain crops immediately after
harvest when grain prices are at their lowest levels. On the other hand, practices of selling and re-
purchasing of grains as a post-harvest grain management practices was found to be less
significant. As it stands now, both grain credit and output markets are not to be relied upon and
market risks are substantial. Lending grains to others was also found to be less important as a
stock management strategy in the context of an imminent risk of post-harvest grain loss. This
perhaps has to do with low demand for grain credit during seasons immediately after harvest.
Apparently the demand is expected to come from other rural households who would be deficient
in their grain production or those unable to buy from markets due to cash constraints. However,
demand for grain credit is the least during the 1st quarter and the highest during the 3rd quarter
after harvest; a pattern that strictly follows grain price levels.

Liquidity constraints could be relaxed through the operation of credit markets and/or through
rescheduling of financial obligations that are due to government until later seasons as crop prices
get better (e.g., payment of land taxes, input loans repayment). Mechanisms could be sought that
link credit markets with grain markets – for example, farmers could take loans in cash to cover
various obligations during this particular season whereas repayments (including those that are due
to government such as fertilizer credits, land and income taxes, etc.) could be scheduled for latter
seasons when prices pick up. Diversification of cash sources and in general integration of the
production process with markets so that farmers would be more market informed and oriented in
their decisions than subsistence driven is necessary for relaxing the liquidity constraints. It is also
important that disadvantaged sectors such as female headed households and the poor need special
support since they are more susceptible to shocks than the relatively better-endowed ones.

But this must be complemented with an intervention to enhance the capacity of farmers to prevent
post-harvest losses, which could be organized at household, community and national levels.
Otherwise, there is an imminent danger that farmers would dispose or lose their grains any way
and might be unable to repay the loans, which makes the matter even worse. The significance of
chemical treatment suggests for linking input markets with future product markets. Interventions
need to take into account and build on farmers’ resources and knowledge including indigenous
techniques.

The problem is that quite often markets do not perform well; hence there is clearly a need for
introducing and strengthening appropriate institutions so as to enable markets work better. One
viable option could be to introduce grain warehouse receipt system so that farmers would deposit
their marketable surplus to be sold when prices get higher. There are legitimate reasons to suspect
that increased production and availability of grains could lead to increased consumption instead
of marketed surplus. Farmers produce primarily for own consumption, not for sale. One of the
reasons as to why warehouse receipt system of grain management and forward grain markets
make a lot of sense, in addition to preventing post-harvest loss, is because they help reduce
availability of grains for direct consumption by the producers; they facilitate monetization of the
production process; and facilitate processing. There are lessons to be drawn from other countries’
experiences that there are a number of advantages including easing access to finance at all levels
in the marketing chain, moderating seasonal price variability, maintaining quality standards and
promoting instruments to mitigate price risks. They also help reduce the need for government
intervention in grain markets as well as the costs of such interventions18 (see, Coulter, J and G.
Onumah, 2002). Above all, the introduction of well managed warehouses in rural villages will
reduce post-harvest grain losses which ultimately support the country’s effort to ensure national
food security.

The importance of attitudinal changes, although not reported as a variable as such in this study, in
respect of consumption vis-à-vis saving and accumulation cannot be overemphasized if efforts
should lead to better livelihoods among the farming community and if agriculture as a whole is to
play a productive role for the country’s socio-economic transformation in general. There is a
concern with regard to the extent of expenditure (grain or cash) during the season immediately
following harvest to be quite high by the farmers’ standards. Wedding and other social
ceremonies, etc., prey on much of the farmers’ produce.

Finally, the recommendations are forward looking in a sense that more in-depth further studies
are necessary to inform policy on credit and saving options (including options for introducing
grain warehouse receipt systems), traditional methods of grain treatment (effectiveness, economy,
health issues, etc.), and farm-nonfarm linkages and the scope for the development of agro-
processing industries, including those small-scale farmer-managed grain processing technologies.

18
Based on warehouse receipt system (WRS) farmers store their marketable surplus (and/or part of their
produce not required for immediate consumption) in a modern warehouse located in their villages. To
certify their deposition they receive a receipt from warehouse managers that indicate the type and
amount of produce they stored. WRS will facilitate the development of efficient and accessible rural
financial systems; i.e. it will facilitate the introduction of inventory credit system. Since WRS improve
production database, farmers can easily get access to formal credit by pledging their grain deposits. This
can allow peasants to progress from being “price takers” to become “price negotiators” and operate
successfully in the local market economy. At present, due to high seasonal grain price fluctuation
inventory credit may be profitable as the inter-seasonal price differences can adequately cover storage
costs. Nevertheless, in the long-run, as agricultural markets become more efficient, grain price
fluctuations will move down and eventually even eliminate, thereby making inventory credit infeasible.
Virtually, in the long-run peasants will be commercial farmers who may adopt a different way of running
farm businesses. Therefore, it is important to note that the introduction of inventory credit is a means to
an end rather than an end in itself. The introduction of improved farmer-managed warehouses in rural
areas may also improve agricultural marketing i.e., it may help farmers to hold commodities for later sale
when prices are higher. It will also enable peasants to supply raw materials year-round to small-scale
food processors uniformly which paves a way for the development of rural small-scale industries.
Furthermore, the WRS will facilitate efficient and effective grain marketing system since marketing
functions such as grading and standardization can be more easily applicable in rural areas. (see, Coulter
and Onumah, 2002). However, sufficient care needs to be exercised if grain warehouse systems achieve
intended results. For example, Berg and Kent (1991) state that cereal banks were bound to fail because of
basic misconceptions in their design and implementation. Gunther and Muck (1995) put some conditions
if cereal banks were to operate successfully. These include they (a) consider and facilitate the needs of
the users, (b) provide social cohesion in the target village, (c) provide human, financial, and technical
resources and with management support, and (d) be profitable.
References
Abebe H. Gabriel (2000). Development Strategies and the Ethiopian Peasantry: Supply Response and Rural
Differentiation. Shaker Publishing: Maastricht.
Abebe H. Gabriel (2001) Towards Explaining Some Distortions in Agriculture-Industry Inter-sectoral Linkages:
Generation Capacities and Patterns of Utilization of Marketed Surplus of Food. In Mulat Demeke and
Tassew Woldehanna (eds) Explaining Economic Growth and Development in Ethiopia. Proceedings of the
Tenth Annual Conference on the Ethiopian Economy. Nazareth, Ethiopia.
Abebe H. Gabriel (2002a) Cereals Production and Marketing in Ethiopia: Opportunities, Constraints and Entailing
Impacts for achieving Food Security Objectives in the Greater Horn of Africa. In Mwalya S. et al (eds)
Exploiting Existing Potentials in Cross-Border Trade to Achieve Food Security in the Greater Horn of
Africa. Technoserve. Nairobi, Kenya.
Abebe H. Gabriel (2002b). The Challenges of Achieving Agricultural Development in Ethiopia: A Macro-policy
Perspective. In Workineh Negatu, Legesse Dadhi and Abebe H. Gabriel (eds) Agricultural Policy in
Ethiopia's Economic Development: Scope, Issues and Propsects. Proceedings of the 6th Annual Conference
of Agricultural Economics Society of Ethiopia. Addis Ababa, Ethiopia.
Alemayehu Lirenso, (1987) Grain Marketing and Pricing in Ethiopia: A Study of the Impact of Grain Quota and
Fixed Grain Prices on Grain Producers. Institute of Development Research Report No. 28, Addis Ababa
University.
Alemayehu Lirenso, (1993) Grain Marketing Reforms in Ethiopia: A Study of the Impact of Deregulation on the
Structure and Performance of Grain Markets. PhD Dissertation, University of East Anglia, Norwich.
Amemiya, T. (1985) Advanced Econometrics. Oxford : Blackwell
Bekele Sinke and Mulat Demeke (1995) Market Integration after the 1990 Reform: the Case of Food Markets in the
Arssi Catchment. Ethiopian Journal of Development Research, Vol. 17, No. 2.
Berg, E., and L. Kent (1991). The Economics of Cereal Banks in the Sahel. Paper prepared for the USAID,
Bethesda, USA.
Blaum, A. and Abate Bekele (2002). Storing Grains as a Survival Strategy of Small Farmers in Ethiopia. Journal of
International Agricultural and Extension Education. Vol. 9, No. 1.
Central Statistical Authority. 1988. Rural Household Income, Consumption and Expenditure Survey. Statistical
Bulletin 61. Addis Ababa, Ethiopia.
Coulter, J. and G. Onumah, (2002). The Role of Warehouse Receipt Systems in Enhanced Commodity Marketing
and Rural Livelihoods in Africa. Food Policy, 27, 319-337.
Coursey, D. G. 1983. Post-harvest Losses in Perishable Foods of the Developing World. NATO Advance Study
Institute Series A46:485
Coursey, D.G.and F.J. Proctor (n.d.) Towards the Quantification of Post-Harvest Loss in Horticultural Produce.
ISHS Acta Horticulturae 49: III Africa Symposium on Horticultural Crops.
http://www.actahort.org/books/49/index.htm
Dereje Ashagari (2000) The Utilization and Conditions of Postharvest Concept in Ethiopia. In EARO. Postharvest
Food Preparation and Storage, and Research and Extension on Agricultural by-products Utilization
Technologies. Amharic Version.
Eleni Gabre-Madhin. Z. (2001). Market Institutions, Transaction Costs, and Social Capital in the Ethiopian Grain
Market. Research Report 124. IFPRI. Washington, D.C.
Ethiopian Agricultural Research Organization (EARO) (2000). Postharvest Food Preparation and Storage, and
Research and Extension on Agricultural by-products Utilization Technologies. Amharic Version.
Proceedings of the Workshop and Exhibition on Postharvest Food Preparation and Research and Extension
on Agricultural by-products Utilization Technologies. Addis Ababa, Ethiopia.
Food and Agricultural Organization of the United Nations, (1998). Map Shows the Food Supply Gap between Rich
and Poor Countries. News and Highlights. http://www.fao.org./NEWS/1998/981204-e.htm
Food and Agricultural Organization of the United Nations/United Nations Environmental Programme. 1981. Food
Loss Prevention in Perishable Crops. 1980 Report of the FAO/UNEP Expert Consultation on the Reduction
of Food Losses in Perishables. Rome.
Gebre-Maskel Dessalegn, T. S. Jayne and J.D. Shaffer (1998) Market Structure, Conduct and Performance:
Constraints and Performance of Grain Markets. Working Paper No. 8. Grain Marketing Research Project,
MEDaC.
Goletti, F. and C. Wolf (1999). The Impact of Post-harvest Research. MSS Discussion Paper No. 29. International
Food Policy Research Institute. Washington, D.C.
Gunther, D. and O. Muck (1995). Les bankques de cereals ont-elles fair banqueroute? GTZ Escchborn.
Howard, J., Ali Said, Daniel Molla, P. Diskin, and S. Bogale (1995) ‘Toward Increased Domestic Cereal Production
in Ethiopia.’ Working Paper No. 3. Addis Ababa. Grain Market Research Project, Ministry of Economic
Development and Cooperation.
Jonsson, L.O. (1972) Possibilities for Improvements of Threshing Methods in Ethiopian Highlands. Agricultural
Engineering Department. Swedish University of Agricultural Sciences.
Kassahun Bedhadha, (2000) The Importance of Postharvest Technologies to Reduce Stored Grains with special
reference to farmers. EARO. Postharvest Food Preparation and Storage, and Research and Extension on
Agricultural by-products Utilization Technologies. Amharic Version.
Kwansoo Kim and Jean-Paul Chavas. 2003. Technological change and risk management: an application to the
Economics of Corn Production. Agricultural Economics. 29 (2) pp. 125-142. October.
Kwarteng, J. (ed) 1999. Enhancing Postharvest Technology Generation and Dissemination in Africa. Proceedings of
the Workshop on Enhancing Postharvest Technology Generation and Dissemination in Africa. Sasakawa
Africa Association, Mexico City.
Legesse Dadi, Gemechu Gedeno, Tesfaye Kumsa and Getahun Degu (1987). Bako Mixed Farming Zone Diagnostic
Survey Report. Department of Agricultural Economics and Farming Systems Research, Report No. 1.
Maddala, G. S. (1988). Introduction to Econometrics. New York, NY: Printice-Hall, Inc.
Mandefro, N., D. Tanner and S. Twumasi-Afriyie (eds) (2001) Enhancing the Contribution of Maize to Food
Security in Ethiopia. Proceedings of the Second National Maize Workshop of Ethiopia. EARO/CIMMIT,
Addis Ababa, Ethiopia.
Ministry of Agriculture, (2001). Crop Production and Marketing in Ethiopia: Paper Presented at Panel Discussion
Organized by the Press and Information Department of the Ministry of Information and Culture. August,
Addis Ababa. Amharic Version.
National Academy of Sciences. 1978. Post-harvest food losses in developing countries. National Academy of
Sciences, Washington, DC.
Pariser, E. R., E. S. Avensu, and M. C. Bourne, et al. 1978. Post-harvest Food Losses in Developing Countries.
National Academy of Sciences, Washington DC.
Shiferaw Gurmu (1986) Baseline Survey of Production, Marketing and Proceedings of Sorghum and Other cereals
in Alemaya area. Research Report No. 25 Institute of Development Research, Addis Ababa: Addis Ababa
University.
Shimelis Admassu (2001) Post- harvest Sector Challenges and Opportunities in Ethiopia. Ethiopian Agricultural
Research Organization, Addis Ababa.
Tesfaye Assefa (1989) An Economic Analysis of Cooperative Farming in the Highland of Ethiopia with special
Emphasis on the Mixed farming Systems. Germany: Wissenschaftsverlag Vauk, Kiel.
Tesfaye T., Girma, T., Douglas, T., Hugo, V. Aklilu, A. Wilfred, M. (2001). Adoption of Improved Bread Wheat
Varieties and Inorganic Fertilizer by Small Scale Farmers in Yalmana Densa and Farta Districts of
Northwestern Ethiopia. Ethiopian Agricultural Research Organization, Addis Ababa.
Toma, R. B., L. T. Fansler, and M. T. Knipe. 1990. World Food Shortage: The Third Dimension. Science of Food
and Agriculture, July. http://www.arec.umd.edu/arec365/castpharv.htm
Wall, P.C. (ed) (2000). Wheat and Weeds: Food and Feed. Proceedings of the Two Stakeholder Workshops. EARO
and CIMMIT.
Wolday Amha (1994) Food Grain Marketing Development in Ethiopia after the market Reform of 1990: A case
study of Alaba Siraro District, Koster Publisher, Berlin.
Wolday Amha (1999) Maize Marketing In Ethiopia: Liberalization and Price Integration issues. Ethiopian Journal of
Development Research. Vol. 21, No. 1.
Wos, A. 1983. Interaction Plan Origin. FAO, between pre-harvest and post-harvest systems and their implications
for socio-economic development. Food and Nutrition Bulletin 7(2):8.
Annexes

Annex I. The Econometric Estimation Model

Taking extremes, a farmer may decide either to dispose of all of his/her produces during the first quarter
immediately after harvest season (when prices are at their lowest levels) or waits until later in which case
sales during the first quarter would be zero. Hence, the utility associated with each alternative is a function
of the independent variables. Denoting Ui1 and Ui0 as the ith farmer’s indirect utilities associated with instant
disposal and late disposal respectively, and assuming a linear function, (for ease of presentation, taking
only two sets of explanatory variables Z and W) we have

(1) Uio=α0+β0Zi0+γ0Wi0+єi0
and
(2) Ui1=α1+β1Zi1+γ1Wi1+єi1

The ith farmer disposes of all of his/her grains immediately after harvest if Ui1 > Ui0 and delays sales to a
later season (in this case beyond the first quarter after harvest) if Ui1 < Ui01. Hence, the probability that the
ith farmer disposes of all of his/her produces at the first quarter immediately after harvest can be given by:

P(Di=1) = P(Ui1 > Ui0)

= P[(єi0-єi1< α1-α0 + β(Zi1-Zi0) +γ(Wi1-Wi2)]
(3) P(Di=1) = F(α1-α0) + β(Zi1-Zi0) +γ(Wi1-Wi2)

Where F is the distribution function of єi0-єi1. If a normal distribution is to be assumed for єi0-єi1, then the
model turns out to be a probit or logit one (see Amemiya, 1985; Maddala, 1988).

Expressing Dit in terms of proportion of the volume of sales during first quarter after harvest (Sit) to total
sales during the year (Qit), i.e., Dit= Sit/Yit, one could note that the distribution of Dit is continuous but tends
to behave like probability; its value tends to lie between zero and one (i.e., 0≤ Dit ≤1). However, it is also a
conceivable possibility for Dit to assume negative values in particular cases where some farmers purchase
grains (instead of selling) during the 1st quarter after harvest. In fact this implies that the latent variable can
take negative values, in which case the tobit model is more appropriate to use.

The tobit model is given by:

α + βZ it + γWit + u it , if Dti > 0; i = 1, 2, 3, ..., n,

(4) Dit = 
0 if Dit ≤ 0

The problem with the estimation of this model is that Zi cannot be observed since it is the expected risk of
post-harvest grain loss. Hence, specification of a model that explains how farmers would form expectations
on the bases of actual and past post-harvest grain losses and other observable variables is essential.
Theoretically, this learning process in which farmers adjust their expectations as a function of the
magnitude of the mistakes they made in the previous period can be expressed as a weighted sum of all past
post-harvest grain losses with a geometrically declining weights.

∞
(5) Zite = ∑β Z
i =1
i it −1

This type of models are called distributed lag models of expectations since they consider the entire past
history of the decision maker with respect to the target variable. If βi are geometrically decreasing, we can
write

9
(6) βi = β0λi, 0<λ<1

The sum of the infinite series is β0 /(1-λ), and if this sum is equal to 1, then we should have β0= 1- λ. Thus
we get,

∞
(7) Zite = ∑ (1 − λ )λ Z
i =1
i
it −1

A further manipulation of equation (7) implies that

(8) Zite - λZite = (1-λ)Zit-1

We can use equation (8) to eliminate the unobserved Zite and estimate the resulting equation. That is,
lagging equation (5) by one time period and multiplying throughout by λ we get

(9) λDit-1 = αλβλZeit-1+ λγWit-1+ vit-1

Subtracting equation (9) from equation (5) and using the definition of adaptive expectations model as given
in equation (8), we get

(10) Dit* = α’+λDit-1+ β’Zit-1+ λWit-λ’Wit-1 + uit

Where α’=α(1-λ), β’=β(1-λ), γ’=γ λ, and ut= v6-λvt-1.

Hence, the tobit equation that will be estimated becomes,

 Dit * , if Dit > 0; i = 1, 2, 3, ..., n,

(11) Dit= 
0 if Dit ≤ 0

In fact Zi could also be disaggregated into a risk of loss in quantity and prices and the relative importance
of either of these elements of risk in explaining instant disposal of grains by farmers could be detected. In
order to identify which factors would determine the choice of a given (or a set of) post-harvest grain
management techniques, a binary (perhaps multinomial) logit model could be used. It is important to note
that higher level of Dit reduces household level of food security, especially if this is the result of some
involuntary processes; this is because it does not make any economic sense to dispose of a larger proportion
of ones produce immediately after harvest at the lowest price level.

Definition of Variables

Nine explanatory variables were included in the tobit regression estimation. The dependent variable is the
proportion of grain output that is actually sold during the first quarter (January to February) to the total
marketable surplus. Description for each explanatory variable is given below. Table 11 presents the mean
and the proportion of 1’s for continuous and dummy explanatory variables, respectively.

FSIZE: refers to household size; the expectation is that high demand for cash to meet non-grain or non-
agricultural consumption needs of the household members triggers larger sales during the first quarter.

SEX: is a dummy variable which takes on 1 if the household head is male and 0 otherwise. It is expected
that male-headed households tend to market their crops more instantly than female headed household.

EDUC: Refers to the highest level of formal education achieved by a member of the household. Education
is expected to influence grain sales decision in such a way that those households with better education level
are hypothesized to demonstrate a more informed sales decisions and vice versa. Proportion of grain sales
during the 1st quarter are expected to be lower for those households with better education.

CHEM: is a dummy variable which takes on a value of 1 if a household used chemicals to reduce post-
harvest grain loss and 0 otherwise. Application of pesticides is expected to reduce the risk of post-harvest
grain loss, thus reducing instant grain sales.

STCAP: refers to estimated total non-bag storage capacity owned by households. With larger capacities of
granaries (excluding bags which are temporary packing materials), the proportion of grains to be sold
during the 1st quarter is expected to be less.

TLU: refers to total livestock owned by households expressed in terms of Tropical Livestock Units.
Standard conversion factors have been used for different categories of livestock (see Annex I). Livestock
ownership is expected to reduce instance crops sales since households may resort to livestock (and/or
livestock products) sales to meet cash needs.

EXPLOSS: refers to the proportion of output that farmers expect to lose due to storage pests (or some
other factors) if grain crops were to be stored until prices rise substantially (in our case until the third
quarter. It is hypothesized that such fear of risk might force farmers to sale their crops sooner rather than
later.

PRICDIF: refers to the difference in average prices of grains that prevailed between third quarter (July to
September, when highest) and the first quarter (January to March, when lowest) measured in birr per
kilogram. Since we couldn’t generate grain price data at household level, we used the average of grain
wholesale prices for each location. Prices of the major crops in each location are considered since farmers
use these crops as ‘cash’ crops i.e. teff for Ada, wheat for Hetosa and maize for Bako. On the one hand, this
variable might capture the market risk element; i.e., as the difference increases farmers would postpone
their sales until later when prices actually get higher; and conversely, as the price difference tends to get
narrower, farmers might prefer to sale their crops sooner rather than later since waiting would be associated
with a higher risk premium specially when post-harvest grain loss could be larger. However, since the price
differences are the same for each location (not household specific), it won’t capture the risk perception by
each household; therefore the above interpretation might be misleading. On the other hand, since price
differences are the largest for teff and the smallest for Maize, this variable might capture the inter-crop
differences (could be considered as crop dummies). In this case, the hypothesis becomes that the variable
takes on a negative coefficient suggesting a tendency to instant sales for Bako and delayed sales for Ada
(capturing the inter-crop differences in susceptibility to post-harvest grain losses).

TAXLOAN: refers to the total monetary value of grains sold to cover dues such as repayment of input
loans and/or pay land taxes. It is expected that farmers’ instant crop sales are largely involuntary but
triggered by the need to meet primarily dues to the government such as input loan repayments and/or land
taxes
Table: Descriptive Statistics of the Explanatory Variables

Variable Mean (Percent with value 1 in case of SD

dummy variable)
SEX 94.50 -
FSIZE 8.00 2.6394
EDUC 7.58 3.6309
CHEM 56.60 -
STCAP 2.55 2.9747
TLU 8.44 5.4844
EXPLOSS 0.12 0.1827
PRICDIF 0.63 0.1419
TAXLOAN 654.40 753.5430