Economic Importance of Agriculture for Sustainable ... TABLE OF CONTENTS ECONOMIC IMPORTANCE OF...
-
Upload
truongdung -
Category
Documents
-
view
220 -
download
2
Transcript of Economic Importance of Agriculture for Sustainable ... TABLE OF CONTENTS ECONOMIC IMPORTANCE OF...
This paper was first presented to the Working Party on Agricultural Policy and Markets, 17-20 May 2010. Reference: TAD/CA/APM/WP(2010)23.
Global Forum on Agriculture
29-30 November 2010
Policies for Agricultural Development, Poverty Reduction
and Food Security
OECD Headquarters, Paris
Economic Importance of Agriculture for Sustainable Development and Poverty Reduction: The Case Study of Ethiopia Xinshen Diao, IFPRI, [email protected]
2
3
TABLE OF CONTENTS
ECONOMIC IMPORTANCE OF AGRICULTURE FOR SUSTAINABLE DEVELOPMENT AND
POVERTY REDUCTION: THE CASE STUDY OF ETHIOPIA .................................................................. 5
1. Introduction .............................................................................................................................................. 5 2. An overview of Ethiopian agricultural policy .......................................................................................... 6 3. Agricultural performance, food security and poverty .............................................................................. 8
3.1. Cereal production and productivity ................................................................................................. 10 3.2. Agriculture and poverty reduction .................................................................................................. 15
4. Agricultural-non-agricultural growth linkages in the Ethiopian economy ............................................ 20 4.1. Why agricultural growth linkages matter? ...................................................................................... 20 4.2. Measuring agricultural growth linkages in Ethiopia – a fixed price input-output model ................ 22 4.3. Results of an Ethiopia fixed price input-output model .................................................................... 25 4.4. Growth linkages in the Ethiopian economy – an economy-wide multimarket model .................... 34
5. Achieving agricultural growth ............................................................................................................... 46 Irrigation ................................................................................................................................................. 46 Adoption of improved seed .................................................................................................................... 47 Promoting modern technology in livestock production ......................................................................... 48 Halving the poverty: markets and non-agriculture matter ...................................................................... 49
6. Conclusions ............................................................................................................................................ 50
APPENDIX A ............................................................................................................................................... 54
A1. An illustration of the fixed price input-output models .................................................................... 54 A2. The fixed price, semi-input-output (SIO) models ........................................................................... 55 A3. The Ethiopia economy-wide multimarket (EMM) model ............................................................... 55
APPENDIX B: APPENDIX TABLES .......................................................................................................... 61
APPENDIX C: SENSITIVITY TEST OF THE SIO MODEL RESULTS ................................................... 69
REFERENCES .............................................................................................................................................. 70
Tables
Table 1. Cereal production in 2003/04 and 2007/08 .................................................................................. 11 Table 2. Cereal growth and growth contribution between 2003/04 and 2007/08 ...................................... 12 Table 3. Share of cereal areas and cereal yields by technology ................................................................. 14 Table 4. Yields in on-farm field trials vs. farmers' yield (tonne/ha) .......................................................... 15 Table 5. Poverty incidence and inequality ................................................................................................. 16 Table 6. National poverty rate by different daily 2005 USD PPP poverty line ......................................... 16 Table 7. Poverty rate by sector of employment of household head and livelihood ................................... 17 Table 8. Poverty rate by administrative region .......................................................................................... 18 Table 9. Agricultural and other income sources across four regions for two percentile household groups19 Table 10. Agro-ecological conditions across four regions for two percentile household groups .............. 20 Table 11. Agricultural growth linkages: international evidence ................................................................ 22
4
Table 12. Impact on Growth due to one unit of increase in selected sectors' output ................................. 27 Table 13. Impact on Income due to one unit of increase in sectors' output ............................................... 30 Table 14. Importance of services in measuring multiplier effect............................................................... 32 Table 15. Population and poverty rates in the three areas .......................................................................... 37 Table 16. Land size and cereal output per household in the three areas .................................................... 37 Table 17. Cereal yield and input use in the three areas .............................................................................. 38 Table 18. Agricultural and non-agricultural growth rate in the simulations .............................................. 39 Table 19. Agricultural growth is more pro-poor ........................................................................................ 44 Table A 1. The structure of the 2006/07 Ethiopian Social Accounting Matrix (SAM) ............................. 61 Table A 2. Income distribution in SAM .................................................................................................... 61 Table A 3. Household consumption spending patterns in Ethiopia ........................................................... 62 Table A 4. Composition of demand and supply by sector in SAM ........................................................... 63 Table A 5. Sensitivity test result – gains in GDP ....................................................................................... 64 Table A 6. Sensitivity test result – gains in total household income ......................................................... 65 Table A 7. Sensitivity test result - income ratio of rural poor household (rural poor household income in
SAM is 1) ................................................................................................................................................... 66 Table A 8. Sensitivity test result - income ratio of rural non-poor household (rural non-poor household
income in SAM is 1) .................................................................................................................................. 67 Table A 9. Agricultural commodities included in the economy-wide, multi-market model ..................... 68
Figures
Figure 1. Agricultural GDP annual growth rate (%) .................................................................................... 9 Figure 2. Agricultural GDP per capita (2000 constant USD) .................................................................... 10 Figure 3. Total cereal area according to the use of modern input (000 hectares) ...................................... 13 Figure 4. Food deficit, food balanced, and food surplus areas .................................................................. 36 Figure 5. GDP growth multipliers in staple and export agricultural growth scenarios .............................. 40 Figure 6. GDP growth multipliers in agriculture-led and non-agricultural-led growth scenarios ............. 41 Figure 7. National poverty rate (%) in agriculture-led and non-agricultural-led growth scenarios ........... 44 Figure 8. Comparison of effect of agricultural subsector growth on poverty reduction in the food deficit
and food surplus areas ................................................................................................................................ 45
5
ECONOMIC IMPORTANCE OF AGRICULTURE FOR SUSTAINABLE DEVELOPMENT AND
POVERTY REDUCTION: THE CASE STUDY OF ETHIOPIA1
1. Introduction
1. It has been more than two and half decades since the government of Ethiopia had formally
adopted Agriculture Development Led Industrialization (ADLI) as its development strategy in 1994. The
main goal of this strategy is to attain fast and broad-based development within the agricultural sector and to
make this sector's development to power broad economic growth. ADLI had been further rationalized as
the basis of the poverty reduction program subsequently adopted by the government in 2002 (MoFED,
2002), a program is officially known as Sustainable Development and Poverty Reduction
Program (SDPRP).
2. With 85% of the population living in the rural areas and depending on agriculture for livelihood,
there is no doubt for the economic importance of the agricultural sector for sustainable development and
poverty reduction in Ethiopia. The agricultural sector accounts for more than 40% of national GDP, 90%
of exports, and provides basic needs and income to more than 90% of the poor. A better performed
agricultural sector has provided growth to the overall economy, improved the food security and reduced
poverty in the recent years.
3. However, debate on the potential for the agricultural sector to lead industrialization and economic
transformation has been lasted for many years in the country, though the debate is often more political and
not well served by rigorous empirical evidence. As the second largest country in Africa and with extreme
high population growth, one of the key questions dominating the policy debate is the doubt for the future
development of agriculture given the increasingly small plots which farmers must earn their living. The
doubt also relates to the role of agriculture in the poverty reduction. While the majority of population lives
in the rural and most poor are the rural poor, there is a question about how much growth in agriculture that
can lead further and significant poverty reduction.
4. Even among those who believe the importance of agriculture in development and poverty
reduction, the debate exists on what kind agricultural growth should be pursued. Should the government
promote large-scale agriculture that is more advantage in adopting modern technology and hence more
productive and competitive? Or should the government focus on the growth of smallholder agriculture
from which a majority of rural population can get benefit? Among the agricultural subsectors, should
considerable specific policy supports or interventions focus on export-led agriculture that may get quick
outcome targeting niche markets, or emphasize the staple crop and livestock sectors that can bring a broad-
based growth for the country?
1. This report is the first draft of the Ethiopia country case study of an OECD project "The Economic
Importance of Agriculture for Sustainable Development and Poverty Reduction." The work should be
considered as work in progress. The principal authors accept responsibility for any errors. The authors are:
Xinshen Diao, Alemayehu Seyoum Taffesse, Bingxin Yu, and Alejandro Nin Pratt, International Food
Policy Research Institute (IFPRI).
6
5. Against this policy background, the objective of this report is to contribute to such debates by
focusing on the role of agriculture‘s future growth in economic transformation and poverty reduction. The
evaluation of such role will be conducted in a broad economic context, and the linkages of agriculture with
the other economic sectors and the possible differential contribution of agricultural growth at sub-sector
levels to the poverty reduction are quantitatively measured. In the following section (Section 2), we first
provide an overview of the agricultural policy evolutions in the last three decades in Ethiopia. The policy
outcome in terms of agricultural growth performance and poverty reduction are assessed in Section 3. In
Section 4 we provide a quantitative measure of agricultural linkages in the economy and assess the role of
future agricultural growth in further poverty reduction. Section 5 provides an assessment on some key
intervention areas that will promote agricultural growth, while the role of the government and policy
implications are concluded the report.
2. An overview of Ethiopian agricultural policy
6. While Ethiopia has been witnessed three major political regime changes in the recent history, the
importance of agriculture has been recognized by each government in this period. However, different
policies pursued by the different regimes have resulted in very different outcomes in agricultural and rural
development, particularly between the last two regimes in the past 35 years. In this period, the Derg regime
(1975-1991) has been characterized as an agrarian socialist regime with widespread government controls in
all economic spheres including agriculture. After overthrowing the imperial regime of Haile Selassie, the
Derg announced an agrarian reform program to declare all rural land to be the property of the state,
together with the nationalization of almost all other assets in the industrial and services sectors such as
manufacturing factories, financial institutions, big hotels and many residential buildings. While the
agrarian reform had prohibited all tenancy relations and provided a large number of rural households with
equal access to cultivation land according to their needs, the restriction on plot size per family, the
prohibition of hired agricultural labour, the intensification of collectivization, the establishment of large-
scale state farms, and a series of other anti-market and state-controlled economic instruments had not only
significantly negatively affected the incentives of farmers but also distorted the market mechanism in
guiding land allocation and promoting productivity improvement. While central planning types of
development strategies had identified agriculture as an engine of growth and targeted the improvement of
food security through agricultural productivity, most growth targets became just a piece of paper and had
never been able to achieve. Ethiopia suffered the worst famine on record in 1984 and the country's
economy was in the dismal state at the end of Derg Regime.
7. Bad policies and brutal political repression during the Derg period generated disastrous economic
outcomes and led to civil conflict. As a consequence the Derg regime collapsed in 1991 and the Ethiopian
People‘s Revolutionary Democratic Front (EPRDF) assumed power. The years that followed witnessed a
radical shift in overall government policy. Both the Transitional government (1991-94) and the EPRDF
government that followed initiated extensive economic reforms including significant market liberalization
and a structural adjustment program. Tariffs have been cut, quota constraints relaxed, licensing procedures
simplified, foreign exchange controls eased, compulsory cooperative membership and grain delivery
discontinued, subsidized rationing of manufactured consumer goods and fertilizers have been discontinued,
privatization of state-owned enterprises begun, private banks authorized, and interest rates decontrolled
and an inter-bank money market introduced. Consequently, the direct role of the state in economic activity
has declined.
8. The most important development strategy under the transitional government is the adoption of
Agriculture Development Led Industrialization (ADLI), which has been a central plank of the EPRDF
government's development program until recent years. The ADLI focuses on productivity growth on small
farms as well as labour-intensive industrialization. This strategy has been justified because agriculture is
the largest sector in terms of output and, particularly, employment and exports; the bulk of the poor live in
7
the agriculture-centred rural areas; considerable gaps exist between rural and urban across key dimensions
of human well-being including health, education and income; and there exists substantial potential to raise
agricultural productivity.
9. Consistent with the ADLI, in the mid 1990s, the government focus shifted from policy reforms
designed to "get the prices right" to public investment in agricultural extension aimed at boosting
productivity through the widespread introduction of modern technology (MoFED 2002). An extensive
extension program called the Participatory Demonstration and Training Extension System (PADETES) had
been implemented, and through this system, the government delivered off-the-shelf packages of fertilizer,
improved seed and credit, as well as information on input use and better agricultural practices to vast
majority of smallholders in the rural areas. The promotion of the credit-fertilizer package was accompanied
by a further liberalization of the fertilizer market. By 1997, fertilizer subsidies were completely removed
and retail prices were fully liberalized, which also resulted in higher fertilizer prices. The use of fertilizer
increased, though diffusion and adoption rates remained. Disappointing despite strong-handed promotion
of the credit-fertilizer packages existed at times. On average, agricultural output continued to fall behind
population growth.
10. Acknowledging the limited success of PADETES, the government revisited the program and
formulated an integrated rural and agriculture development strategy that was launched in 2002. The new
development strategy, which is officially known as ‗Sustainable Development and Poverty Reduction
Program (SDPRP)‘ (MoFED 2002), has centred on the principal goal of poverty reduction. In line with this
program, the government has introduced fiscal decentralization, judicial and civil service reform, and
public sector capacity building. After the continuing evidence of widespread food insecurity in the drought
of 2002/03, the government also initiated a strong focus on safety nets, programs to build the assets of food
insecure households, resettlement, and soil and water conservation (especially water harvesting).
11. The SDPRP, which covered the three years 2002/03-2004/05, was the first full Poverty
Reduction Strategy Paper (PRSP) developed and implemented by the Ethiopian government. It was
followed by the second PRSP titled ‗Plan for Accelerated and Sustained Development to End
Poverty (PASDEP).‘ The Plan formed Ethiopia‘s guiding strategic framework for the five-year period
2005/06-2009/10 (MoFED 2005). PASDEP aimed to significantly accelerate growth via the
commercialization of agricultural and the promotion of private sector development. It further focused on a
number of areas/issues in setting targets and designing interventions - a geographically differentiated
strategy, addressing the population challenge, unleashing the potential of Ethiopia‘s women, strengthening
the infrastructure backbone , managing risk and volatility, scaling up to reach the MDGs, and creating jobs
(particularly in urban areas) (MoFED 2005).
12. The agricultural growth agenda set by PASDEP consisted of the following elements:
shift to higher-valued crops;
promote niche high-value export crops;
a focus on selected high-potential areas;
facilitate the commercialization of agriculture;
support the development of large-scale commercial agriculture where it is feasible; and
better integrating farmers with markets – both locally and globally.
13. The instruments to achieve these in the context of PASDEP include (i) constructing farm-to-
market roads; (ii) development of agricultural credit markets; (iii) specialized extension services for
differentiated agricultural zones and types of commercial agriculture; (iv) the development of national
8
business plans and tailored packages for specialized export crops (such as spices, cut flowers, fruits and
vegetables); (v) area irrigation through multi-purpose dams; (vi) measures to improve land tenure security,
and to make land available where feasible for large-scale commercial farming; and (vii) reforms to
improve the availability of fertilizer and seeds.
14. PASDEP reflected deeper understanding of the role of agriculture in growth, with an
acknowledgement that the ADLI strategy needs to be enhanced. Indeed, the acknowledgement is explicit in
a draft version of the PASDEP document:
The approach remains basically sound.... However, the full potential of agricultural
growth has not yet been realized, and intensification of the strategy is needed. More
broadly, the overall growth performance has not yielded the hoped-for poverty-reduction
results over the long-term. The time is now right to mount a major effort to accelerate
growth, and this forms the main thrust of the PASDEP. (MoFED 2005).
15. Thus, PASDEP articulates a more comprehensive strategy which focuses on commercialization
and intensification of agriculture, emphasizes the importance of intersectoral linkages, favours a
geographically differentiated strategy, recognizes the dangers of volatile economic growth and rapid
population growth, and highlights the importance of the urban sector.
16. The analytical foundations of PASDEP are broadly consistent with the underlying causes of
poverty – low productivity symbiotically linked with:
limited capital stock (physical, human, infrastructural, natural) compounded by rapid population
growth;
weak institutions (imperfect or absent markets, weak civil service including agencies of public
service delivery, insecure property rights, sparse early-warning and emergency assistance
systems); and
regular shocks (erratic weather, volatile prices, ill-health episodes, and conflict).
17. Critical concerns relate to implementation and sustainability of achievements include:
Implementation - the extent to which priorities are set on the basis of decent diagnostics, the
feasibility of some of the ambitious targets in light of resource requirements (‗can‘ vs. ‗want‘);
Sustainability of achievements – institution building, the effort towards domestic resource
mobilization, industrial policy and incentives to the private sector, business environment...
18. The case of the livestock sector particularly in mixed-farming is a specific area yet to be fully
incorporated within the economic policy sphere.
19. The overview of Ethiopia's agricultural policy evolution in this section highlights the vision and
measures the current government has to promote agricultural growth. In the next section, we focus on
agricultural performance in the recent five years and poverty reduction in a relative longer period.
3. Agricultural performance, food security and poverty
20. Ethiopia has been well known for its agricultural development challenge given its large and rapid
growing population and limited and deteriorated land resource. These two factors together have caused
extreme land shortages in the highland Ethiopia, the area most population lives and most agricultural
production occurs. According to World Bank (2005), per capita land area in the highlands has fallen from
9
0.5 hectare in the 1960s to only 0.2 hectare by 2005, and a marginal productivity of labour is estimated at
close to zero.
21. Population pressure has led to expanded cultivation into forest areas and steep slopes. This
creates serious repercussion for the environment, which, together with fluctuation in rainfall, have made
agricultural production very vulnerable to weather shock. As indicated by Figure 1, growth in agricultural
GDP, which is predominated by crop and livestock production, has been extremely fluctuated. The positive
growth unlikely continued for more than two or three years before the next weather shock occurred and
resulted in agricultural GDP to decline. In the period of 22 years between 1982 and 2003, there were four
years in which agricultural GDP declined by more than or close to 10%, and growth was also negative in
the other five years. In fact, agriculture only contributed 0.3 percentage points on average to annual GDP
growth over the decade to 2003 (World Bank 2006). The serious drought that lasted for two years in 1984
and 1985, together with the bad policy under the Derg regime, resulted in the worst famine in Ethiopian
modern history. Per capita agricultural GDP declined by one-third and many people starved to die. Poor
farmers who survived from the drought often lose their livestock, causing permanent shock to their
livelihood.
Figure 1. Agricultural GDP annual growth rate (%)
-21
-16
-11
-6
-1
4
9
14
19
1982 1987 1992 1997 2002 2007
Agricultural growth
Agricultural growth
Source: World Development Indicator 2009.
22. In the recent years, the policy reforms, agricultural investments and public service provision have
provided a boost to agricultural production, primary cereals. After 2003, Ethiopia agriculture has witnessed
the most rapid growth in its history. As Figure 1 indicates such growth has already lasted for five years and
annual growth rate is also high. Even excluding the highest growth rate of 17% in 2004 that has certain
recovery pattern from the drought of 2003 in which agricultural GDP fell by 10.5%, the average annual
growth rate of 2005-08 is still as high as 10%. While there is a controversial debate about how realistic
such rapid growth is and a systematic, rigorous and comprehensive investigation into the sources of such
growth has yet be conducted, cereal production data aggregated from the country's agricultural sample
surveys is consistent with the total agricultural GDP growth. The recent growth has reversed the down slop
trend of agricultural per capita GDP, a trend had lasted for three decades under the Derg regime and the
1990s.
10
Figure 2. Agricultural GDP per capita (2000 constant USD)
R² = 0.6582
50
55
60
65
70
75
80
1981 1986 1991 1996 2001 2006
AgGDP pc
Trendline
Source: World Development Indicator (2009).
3.1. Cereal production and productivity
23. Cereals are the dominant staples for the majority of Ethiopians, are the source of 62% of average
Ethiopians' daily calorie intake, and account for about 45% of food expenditure for an average household.
Thus, cereals, including barley, maize, teff, wheat and sorghum, are the most important crops for Ethiopia's
agriculture. While 64% of agricultural value added comes from crops, more than 70% of crop land is
devoted to cereal production. More than 11 million smallholders engage in cereal production, and total
cereal production was 13.6 million tonnes in 2007/08, an increase of 4.8 million tonnes compared to
production in 2003/04 (Table 1). Total area allocated to cereals also expanded by 27% from 6.8 million
hectares in 2003/04 to 8.8 million in 2007/08. At the same time, average cereal yield exhibited a 22%
growth from 1.3 tonnes/ha in 2003/04 to 1.6 tonnes/ha in 2007/08.
24. Teff is the most favourable staple crop for both Ethiopian rural and urban consumers and for all
different income levels of households. Thus, teff occupied more land than the other crops. 30% of total
cereal land in 2007 was used for teff production. The second important food crop is maize, which occupied
20% of total cereal land, followed by sorghum (18%), wheat (16%), and barley (12%). While most cereal
crops are staple foods, barley is also used for local alcohol production. In terms of volume of production in
the same year (2007), maize actually ranked first with 3.8 million tonnes of output, and output of teff is
3 million. While teff occupied 30% of cereal land, output of teff is equivalent to 22% of cereal output. This
implies that teff is much less productive in land use. Indeed, national average yield of maize is 2.1 tonne
per hectare (mt/ha) in 2007, yield of teff is only 1.2 mt/ha, the lowest level of yield among all major cereal
crops. Teff is a crop only grown in few countries (mainly in Ethiopia) and its yield response to fertilizer is
relatively limited given that the technology to develop high-yield variety of teff is more difficult than to
develop other cereal crop varieties broadly grown in the world. On the other hand, teff is more favourable
than maize in Ethiopian diet and has higher income elasticity in demand. This combination indicates a
potential challenge to Ethiopia food security due to the inconsistency between the technology potential and
consumers' preference.
11
Table 1. Cereal production in 2003/04 and 2007/08
2003/04 2007/08
Area Output Yield
In total
cereal areas Area Output Yield
In total
cereal areas
000 ha 000 mt mt/ha % 000 ha 000 mt mt/ha %
Barley 911 1 071 1.18 13.4 985 1 355 1.38 11.4
Maize 1 300 2 455 1.89 19.1 1 767 3 750 2.12 20.4
Teff 1 985 1 672 0.84 29.1 2 565 2 993 1.17 29.6
Wheat 1 075 1 589 1.48 15.8 1 425 2 314 1.62 16.4
Millet 303 304 1.00 4.4 399 538 1.35 4.6
Sorghum 1 242 1 695 1.36 18.2 1 534 2 659 1.73 17.7
Total 6 816 8 786 1.29 100.0 8 675 13 609 1.57 100.0
With any modern inputs
Barley 390 492 1.26 42.8 499 688 1.38 50.7
Maize 783 1 518 1.94 60.2 1 106 2 441 2.21 62.6
Teff 1 059 919 0.87 53.4 1 763 2 076 1.18 68.7
Wheat 687 1 108 1.61 63.9 1 024 1 743 1.7 71.9
Total 2 919 4 037 1.38 55.4 4 392 6 948 1.58 65.1
Without modern input
Barley 521 579 1.11 57.2 486 667 1.37 49.3
Maize 517 937 1.81 39.8 661 1 309 1.98 37.4
Teff 926 753 0.81 46.6 802 917 1.14 31.3
Wheat 388 481 1.24 36.1 401 571 1.43 28.1
Total 2 352 2 750 1.17 44.6 2 350 3 464 1.47 34.9 Source: Authors’ calculation using CSA data.
25. Under ADLI, the government emphasized intensification to increase agricultural productivity,
especially a centralized extension push focusing on technological packages combining credit, fertilizers,
improved seeds and better management practices to raise productivity. Under this program, fertilizer use
almost doubled between 1990 and 2000 to 290 000 tonnes (World Bank 2007). However, the intensity of
fertilizer nutrient use per hectare stagnated in the latter half of 1990s and into this decades. According to
World Bank (2007), only 37% of farmers use inorganic fertilizer and application rates remain low at 16
kg/ha of nutrients (about 33 ka/ha for commercial products).
26. Following the drought of 2002/03, a large-scale food security program was rapidly scaled up in
poor and vulnerable areas, amounting to a significant part of total public spending. More recently,
weaknesses in the marketing system have been recognized and a new marketing strategy is being
developed based on scaling-up cooperatives, and establishing a commodity exchange and its associated
institutions. As a result of these various programs, public spending on agriculture, natural resources and
food security has risen from 5% of the total government budget in 1997/98 to over 10% in 2003/04, above
the CAADP's target for Sub-Saharan Africa.
27. Strong push on intensification seems to show certain results in the recent years. Consistent with
growth in total agricultural GDP, the period 2003/04 to 2007/08 has registered a recorded cereal
production growth in Ethiopian history. Total cereal production increased by 54.9% in four years, with a
more than 10% of annual average growth rate. Against the historical trend in which almost all increased
production can be explained by area expansion, only half of the recent growth is the result of area
expansion while the rest half is due to yield increase (Table 2). Table 2 also displays the growth by
individual crops, and growth contribution of area expansion and yield increase in each crop‘s production
12
growth. As Table 2 indicates, among the four main crops, there are two crops, teff and sorghum, for which
the growth in production is dominantly a result of yield increase, which contributes to, respectively, 63%
and 58.7% of teff and sorghum output growth. While for maize and wheat, area expansion is the main
source of their growth, increases in yield still contribute to 32% and 29% of their output growth,
respectively.2
Table 2. Cereal growth and growth contribution between 2003/04 and 2007/08
Total growth (%) Growth contribution
(Increased individual crop
output = 100)
(Increased total cereal
output = 100)
Area Output Yield Due to growth in
Areas Yield Output Areas Yield Output
Barley 8.1 26.5 17.0 30.6 69.4 100 3.0 6.9 10.0
Maize 35.9 52.7 12.4 68.1 31.9 100 18.8 8.8 27.6
Teff 29.2 79.0 38.5 37.0 63.0 100 8.1 13.9 22.0
Wheat 32.6 45.6 9.9 71.4 28.6 100 12.1 4.9 17.0
Millet 31.7 77.0 34.4 41.2 58.8 100 1.6 2.3 4.0
Sorghum 23.5 56.9 27.0 41.3 58.7 100 8.1 11.5 19.5
Total 27.3 54.9 21.7 49.7 50.3 100 49.7 50.3 100.0 Source: Authors’ calculation using CSA data.
28. The last three columns of Table 2 display growth contribution of each individual crop to total
cereal production growth. As the last column of Table 2 indicates, 27.6% cereal output increase is the
result of maize production growth, followed by teff, which contributes to 22% of cereal growth.
Considering the contribution of area expansion to the cereal production growth, increased harvest area in
maize production results in 18.8% of total cereal production growth, followed by wheat, of which area
expansion contributes to 12.1% of total cereal production increase. Considering the contribution of yield
increase to cereal production growth, 13.9% of such growth is the result of yield increase in teff, followed
by sorghum, of which yield increase contributes to 11.5% of total cereal production growth.
29. It has to point out that cereal cultivation is highly concentrated geographically. Almost 80% of
total area under cereals is in the Amhara and Oromia regions to the northwest, west, southwest and south
of Addis Ababa. This area includes a diverse set of conditions for agricultural production. Spatial
conditions for production and market access have been discussed elsewhere (see Diao and Nin Pratt, 2005;
Taffesse, 2006).
30. Table 1 also disaggregates production of the four main cereals by the use or without use of
modern input. Ethiopia has witnessed rapid increases in modern input use, particularly the use of fertilizer
in the result years. As Figure 3 indicates, fertilizer combined with local seed was applied to 2.5 million
hectares in 1995, reached to 3 million hectares in 2003/04, and jumped to 4 million hectares in 2007/08, a
level similar as the areas without use of any modern input.
2. It should notice that level of maize yield in Ethiopia is higher than that in many other African countries in
East Africa even in the early 2000s. According to World Bank, over the period of 1986 to 2000, the
average yield of maize increased from 1 mt/ha to 1.8 mt/ha as the result of National Maize Research
Programs.
13
Figure 3. Total cereal area according to the use of modern input (000 hectares)
0
500
1000
1500
2000
2500
3000
3500
4000
4500
1995/96 1996/97 1997/98 1998/99 1999/00 2000/01 2003/04 2004/05 2006/07 2007/08
Local seed & no fertilizer Fertilizer & local seed Fertilizer & improved seed
Source: Author’s calculation using CSA data (various years).
31. The government‘s fertilizer promotion policy focuses on the four main cereal crops, which are
relatively more responsive to fertilizer and have higher producer prices. By focusing on these four crops in
Table 1, it indicates that, 55% of the four crops‘ areas were applied by modern inputs, primarily fertilizer,
in 2003/04, while the per cent has increase to 65% in 2007/08. In these four years, total harvest areas of
these four crops increased by 28%. In the meantime, the areas without use of modern input did not change
and stabilize around 2.3 million hectares.
32. However, the yield difference between with and without modern input use is modest for all the
four crops. Except for wheat, the average yield for the areas with modern input use is less than 15% higher
than the average yield for the areas without any modern input. In the case of wheat, the yield gap is 30% in
2003/04, while it falls to 19% in 2007/08. Many factors affect fertilizer efficiency and in their 2007 study,
Byerlee et al. (2007) concluded the following major factors that affect the results of the intensification
program that the government has been promoted:
Low technical efficiency in the use of fertilizer due to the application of standard packages to
very diverse and risky environments, and the timeliness and quality of input supply.
Poor performance of the extension service resulting from limited human capital, competing
responsibilities, and entrenched routines and behaviours among extension agents.
Shortcomings in seed quality and timeliness of seed delivery.
Promotion of regionally inefficient allocation of fertilizer as promotion of fertilizer use tied to
credit programs is fed by government targets for fertilizer consumption at the local, regional and
national levels.
Input distribution tied to credit that limits the opportunity for the emergence of private sector
retailers.
Generation of an unlevel playing field in the rural finance sector by the guaranteed loan program
with below-market interest rates, undermining efforts to set up alternative institutions, branches
of commercial banks, or independent financial cooperatives.
14
Table 3. Share of cereal areas and cereal yields by technology
Share in each crop's total areas Average yield (tonnes/hectare)
2003 2004 2005 2006 2003 2004 2005 2006
Maize
Fertilizer with local seed 11.8 13.9 19.7 16.4
1.98 1.82 2.36 2.26
Fertilizer with improved seed 23.4 17.7 17.7 21.6
2.12 2.04 2.80 2.53
Improved seed without fertilizer 1.3 1.1 0.9 0.6
2.22 1.68 2.21 2.49
Without fertilizer & improved seed 63.6 67.3 61.7 61.5
1.78 1.68 2.10 2.14
Average 1.65 1.51 1.81 1.88
Teff
Fertilizer with local seed 45.2 47.2 54.4 53.5
0.87 1.06 1.06 1.22
Fertilizer with improved seed 0.3 0.5 0.5 0.6
0.95 0.99 0.99 1.24
Improved seed without fertilizer 0.3 0.2 0.1 0.1
0.67 0.61 0.90 1.47
Without fertilizer & improved seed 54.2 52.1 45.0 45.7
0.82 0.89 0.97 1.16
Average 0.45 0.47 0.44 0.54
Wheat
Fertilizer with local seed 50.1 50.4 60.6 53.8
1.62 1.81 1.83 1.69
Fertilizer with improved seed 3.7 3.4 2.6 2.0
1.54 1.79 1.85 1.59
Improved seed without fertilizer 0.9 0.6 0.5 0.9
1.27 1.22 1.41 1.49
Without fertilizer & improved seed 45.3 45.6 36.4 43.3
1.22 1.31 1.38 1.47
Average 0.62 0.67 0.56 0.68
Barley
Fertilizer with local seed 25.9 25.6 27.3 26.6
1.24 1.29 1.52 1.50
Fertilizer with improved seed 0.1 0.3 0.1 0.2
1.17 1.16 1.44 1.86
Improved seed without fertilizer 0.2 0.3 0.0 0.0
0.90 1.06 1.10 1.39
Without fertilizer & improved seed 73.8 73.8 72.5 73.1
1.09 1.04 1.19 1.38
Average 0.81 0.77 0.86 1.01 Source: Author’s calculation using part of CSA data (various years).
33. While all these factors have resulted in the low response of cereal yield to the use of fertilizer,
increased use of fertilizer without use of high-yield seed varieties seems to be the most important factor.
Table 3 shows cereal area allocation for the different technologies and the yield for the four cereal crops
under different technologies. Except for maize, the combined use of fertilizer and improved seed is applied
to 0.3% - 0.6% of harvested teff areas and 2.0 - 3.7% of wheat area. Only in the case of maize, the share is
18% - 23% high. Moreover, when a rapid increase in fertilizer use occurred in the last four years, growth in
the areas with the combination of fertilizer and improved seed has become stagnant. While the low yield
response to the combined use of fertilizer and improved seed in the case of teff and wheat seems to indicate
that the so-called improved seed is not really high yield varieties, a further assessment is necessary for
fully understanding farmer‘s behaviour as well as the constraints for promoting such combined technology
in Ethiopia. To show the potential of doubling Ethiopia cereal production by improving yields, we draw
from the World Bank (2006) to display what are the reachable yields for cereal crops in Table 4.
15
Table 4. Yields in on-farm field trials vs. farmers' yield (tonne/ha)
NAEIP* (1995-1999) SG2000** (1993-1999) Current farm
yield
2000-04
improved traditional improved traditional
Maize 4.73 1.57 4.60 1.57 1.82
Teff 1.43 0.85 1.62 0.64 0.82
Wheat 2.93 1.17 2.31 0.95 1.31
Barley 2.15 1.00 1.05
Sorghum 2.79 1.12 2.08 0.92 1.21 * NAEIP is the National Agricultural Extension Intervention Program. ** SG2000 is Sasakawa Global 2000 Program. Source: World Bank 2006.
34. Agricultural sector performance over the recent five years is also indicative of the new direction
of the country's development strategy (i.e. PASDEP), which indicates an evolution of strategy toward
market-driven diversification and commercialization, and increasing exports with a greater focus on private
sector investment. Following the opening of incentives for private investment in flower industry, a total of
over 100 investors have invested in the industry and flower exports increased to nearly USD 13 million in
2005. Other investments in high value products and supply chains are emerging, including the export of
green beans to Europe, the emergence of a contractual supply chain for the UK based bean industry, and
the supply of quality milk and poultry to urban centres. Several of these emerging industries involve out-
grower and/or contract arrangements with small farmers, often linked with an emerging indigenous
entrepreneurial class of farmers and agribusiness. Exports of oilseeds and pulses, two traditional cash crops
grown by many farmers, have also experienced impressive growth, increasing their value by a factor of ten
between 1997 and 2006 and demonstrating the increasing competitiveness of these sectors through area
specialization and the uptake of new technologies. While coffee is still the most important export crop in
Ethiopia, the combined exports of other crops and leather has passed the coffee in export value in the
recent years.
3.2. Agriculture and poverty reduction
35. Given that 85% of Ethiopians live in the rural area and more than 90% of the poor are in the
rural, agricultural growth direct transfers into poverty reduction if the growth is significant. While income
generation role of agricultural growth is important for poverty reduction, as many poor in rural Ethiopia
either mainly produce for own consumption need or are the net buyers of cereals, the direct consumption
effect of agricultural growth is equally important in poverty reduction in the country. While the country has
conducted agricultural production sample survey almost in each year, the national representative household
surveys in income, consumption and expenditure (HICES), which can provide a poverty measure, have
been conducted only in each five years in the recent 15 years. Moreover, the HICES is only available
publicly for the two runs of 1995/96 and 1999/2000, which have been widely used in poverty assessment
in the literature. The most recent HICES was conducted in 2004/05, but the survey data is not available for
analysis, besides the poverty rate reported by the government.
36. Table 5 first reports the government's official poverty assessment based on the three runs of
HICES and the poverty measure is based on the country's own standard, which is not necessary to be the
same as international standard measured by daily income. Instead, the country's own poverty measure
standard considers the minimum consumption level to meet the basic need of food and other spending and
also considers the different consumption expenditure needs (e.g., for housing) between the rural and urban.
For comparison, we also present the national poverty rate measured according to the World Bank's
16
standards in Table 6, in which the poverty line of USD 1.25 in 2005 PPP dollar per day is commonly used
as an international standard for poverty measure across countries.
37. While the level of national poverty rate for the same year differs between national and the
international measures, the trends of change in the poverty rate are similar. Poverty has declined over time,
and it declines more according to the measure using the international standard. Moreover, the poverty
declined more in both absolute level and per cent change between 1999/2000 and 2004/05, and this is true
for both measures. The third observation is that while poverty is falling, the country seems to be unlikely to
achieve the MDG1 of halving its 1990's poverty rate by 2015, if the poverty reduction trend, which is
calculated according to the country's own poverty measure, will be similar in the next 10 years after
2004/05 as that in the previous years between 1995 and 2005.
Table 5. Poverty incidence and inequality
Poverty rate (%) % Change in incidences
1995/96 1999/00 2004/05 1999/00 over
1995/96
2004/05 over
1995/96
2004/05
over
1999/00
Headcount index
National 45.5 44.2 38.7 -2.9 -14.9 -12.4
Rural 47.5 45.4 39.3 -4.4 -17.3 -13.4
Urban 33.2 36.9 35.1 11.1 5.7 -4.9
Gini coefficient (consumption)
National 0.29 0.28 0.30
Rural 0.27 0.26 0.26
Urban 0.34 0.38 0.44 Source: MOFED 2006.
Table 6. National poverty rate by different daily 2005 USD PPP poverty line
Poverty rate (%) Gini
USD 1.00 USD 1.25 USD 2.00
1981/82 48.6 66.2 89.9 0.32
1995/96 44.9 60.5 84.6 0.40
1999/00 36.3 55.6 86.4 0.30
2004/05 21.7 39.0 77.5 0.30 Source: World Bank PovCal Net.
38. Table 5 also reports the poverty rate for the rural and urban households separately. After
considering different expenditure patterns between the rural and urban households caused by location
factors, the rural poverty rate is still higher than the urban poverty rate in each survey year reported in
Table 5. However, the reduction in poverty rate is more rapidly in the rural than in the urban, indicating the
direct effect of agricultural growth on the poverty reduction. The urban poverty has in fact risen in the
period between the first and third survey years, although it fells slightly between the second and third
surveys. Because of this, the rural and urban poverty rate has converged in the recent years.
39. Gini coefficient that is used to measure inequality is included in both Tables 5 and 6. While Gini
coefficient has slightly increased in Table 5 according the country's own consumption-based assessment,
this coefficient fells significantly in Table 6 between 1995/00 and 2004/05 using the income measured by
the international standard. The slight increase in inequality reported in Table 5 is the result of significant
17
increase in the value of urban Gini coefficient. The urban Gini, which starts with a value higher than the
rural Gini in 1995/96, has increased from 0.34 to 0.44, a 10 percentage points increase, in the period of
10 years. In the same period, the rural Gini has actually fallen slightly, and has been at a low level of 0.27
to 0.26.
40. Significant declines in the rural poverty rate and low and relatively stable rural Gini coefficients
in Ethiopia seem to indicate that agricultural growth in the recent years has benefited the poor and growth
outcome has been shared by a majority of the rural population. Moreover, such growth outcome on poverty
and income distribution seems to be stronger in the first five years of the 21th century than in the previous
years. However, as the data of the recent HICES survey of 2004/05 is not available for our study, we have
to draw from the exiting analysis on the previous HICES surveys to further understand the relationship
between agriculture and poverty reduction in the rest of this section. In 2005, the World Bank published a
detail assessment on the role of agriculture in the well-being and poverty in Ethiopia (World Bank, 2005).
The relevant findings of this report are synthesized here to conclude this section.
Table 7. Poverty rate by sector of employment of household head and livelihood
Poverty rate (%)
Consumption per adult
equivalent Population share
1995/96 1999/00 1995/96 1999/00 1995/96 1999/00
By sector of employment of household head
Agriculture 40 38 1 592 1 600 85 84
Industry 32 43 1 980 1 707 1 6
Services 28 35 2 201 2 113 14 10
By type of livelihood
Mainly agriculture 41 38
Mainly cash crop producers 29 26
Coffee producers 42 40
Chat producers 19 33
Tea producers 41 24 Source: Tables 1.9 and 1.10 in the World Bank (2005).
41. Authors of the World Bank Report (2005) calculate the poverty incidence by sector of
employment of household head and by rural livelihood. As indicated in Table 7, poverty incidence among
households employed in non-agriculture is substantially lower than those employed in the agricultural
sector in the first run of HICES survey (1995/96). The poverty gap became significantly smaller between
these two types of employment and the poverty rate of a group of households whose heads are employed in
the industrial sector is actually higher than the poverty rate for agriculturally employed households. While
the poverty rate of the group households whose heads were employed in the service sector is still lower
than the poverty rate of agriculturally employed households in the second run of the survey, the gap has
become much smaller than that in the first run. Moreover, poverty rate increases in the non-agriculturally
employed household groups and it increases more in the industrially employed household group. This
finding is consistent with the change in the poverty rate by rural and urban household groups, which
further confirms the positive role of agricultural growth to rural poverty reduction.
42. In the second panel of Table 7, the poverty distribution is displayed by different types of rural
livelihood. The poverty rate for the group of households whose main livelihood activity is the engagement
of agriculture is similar as the poverty rate for the group of households whose heads employed in the non-
agricultural sector. However, if we only focus on the rural households whose main livelihood activities are
18
cash crop production, the poverty rate is significantly lower than the whole agricultural group in both runs
of the surveys. Among the cash crop producers, the households mainly involving in the production of
coffee, the most important export crop of Ethiopia, actually are similar poor as the agricultural household
group as a whole. The poverty rate for coffee growers is actually higher than for the agricultural producers
as a group in the second run of the survey (1999/00) due to the declined world coffee price. However, the
poverty rate for coffee growers has to be read in caution as coffee production areas are quite concentrated
in Ethiopia and the comparison between coffee and non-coffee households should make more sense
considering these areas only.
43. Geographically, poverty is widespread in Ethiopia, while the majority of the poor live in the four
large regions (Tigray, Amhara, Oromiya and SNNP) and Addis Ababa. Together, these regions account for
85% of population of the country (World Bank, 2005). The World Bank Report shows that the highest
rates of poverty among the major regions are found in Tigray and SNNP. Authors of the report calculate
their own regional poverty rate and two poverty lines are considered in the calculation. The two large
regions with the high poverty rate are characterized by lower than average arable land per capita, which
underscores the role of land scarcity in determining poverty.
Table 8. Poverty rate by administrative region
Lower poverty rate Upper poverty rate
1995/96 1999/00 1995/96 1999/00
Tigray 45 49 66 69
Afar 20 43 26 63
Amhara 45 36 65 55
Oromiya 28 32 46 52
Somali 8 15 18 33
Benishangul-Gumuz 49 54 72 71
SNNP 49 48 67 65
Gambela 35 66 48 79
Harari 25 29 43 47
Addis Ababa 34 41 50 57
Dire Dawa 47 49 65 68
National 38 38 57 57 Source: Table 1.12 in the World Bank (2005) and national poverty rate is from Table 1.2 in the World Bank (2005).
44. Regional disaggregation in poverty also shows that poverty rose between the two runs of surveys
in 9 of 10 regions and declined only in Amhara under the lower poverty rate assessment and slightly
declined in additions in the other two regions under the upper poverty rate assessment. It has to point out
that the poverty rate in Table 8 is from the World Bank report's authors own calculation based the same
data of HICES from which the official poverty rate is derived. According to them, both lower and upper
poverty rates for the country as a whole did not change between the two runs of HICES, which explains
why poverty rate rose in most regions and declined in few.
45. Focusing on the four large regions that account for 85% of total population in Ethiopia, the
World Bank report further assesses the relationship between livelihood and level of income. Table 9 is
drawn from the findings of the report. By considering only two percentile groups of population in each of
the four regions, it shows that for the poor one-percent of population (the 25th percentile), the level of real
expenditure is only equivalent to 63% of regional average for the three regions and only 48% for Amhara.
On the other hand, for the non-poor one-percent of population (the 75th percentile), the level of real
19
expenditure is 1.85 - 1.88 times of the expenditure level for the 25th percentile of population in the three
regions and 2.53 times in Amhara.
Table 9. Agricultural and other income sources across four regions for two percentile household groups
Tigray Amhara Oromiya SNNP
Real expenditure per adult equivalent in Addis 1995 price
Mean 1 181 1 981 1 396 1 550
25th percentile 740 956 874 984
75th percentile 1 389 2 416 1 647 1 817
Share of income from agriculture (%)
Mean 67 75 73 71
25th percentile 53 71 63 59
75th percentile 95 96 96 92
Share of income from other sources (%)
Mean 18 11 12 13
25th percentile 4 3 2 3
75th percentile 24 14 15 18 Source: Drawn from Table 4.7 in the World Bank (2005).
46. While there are many factors that associate to the income gaps between the poor and non-poor,
we focus only on the sources of income in Table 9. As indicated in the table, for the poor percentile group,
share of agricultural income is consistently lower than that for the regions as a whole and for the rich
percentile group across the four regions. It shows that for the rich percentile group, share of income from
agriculture is more than 95% for the three regions and 92% for SNNP. On the other hand, for the poor
percentile group, the share is below 60% for the two regions, and 63% and 71%, respectively, for Oromiya
and Amhara. The poor not only have lower share of agricultural income, but also few income generation
opportunities from other livelihood sources. This seems to indicate that many poor Ethiopian households,
particularly the extremely poor ones, could not earn enough income to meet their basic expenditures, and
supports of food and other types of aids are important component of their basic consumption. This finding
also indicates the positive relationship between agriculture as the main source of income and rural
households‘ position in income distribution ranking, and hence, the important role of agriculture even for
the relative non-poor rural households.
47. Similar as most African countries, Ethiopia is characterized as rainfall-dominated agriculture and
irrigation covers less than three per cent of agricultural crop areas. Understanding the relationship between
agro-ecological conditions and household income is helpful for identifying a group of policies that are
more location specific and targeted. Table 10 reports three important agro-ecological condition factors for
the four large regions and these factors are defined at the woreda level. As indicated in the table, the poor
group of households (25th percentile) lives in relatively low altitude areas, while the non-poor group
(75th percentile) lives in highland in all the four regions. While long run average rainfall varies across the
four regions, the poor seem to live in the areas with less rain than the areas where the non-poor live. On the
other hand, the variation of rain is larger in the areas the non-poor live than in the areas the poor live.
Agro-ecological condition information compiled in Table 10 indicates that the absolute disadvantage in
agricultural natural production condition partially characterizes the areas where the poor live, which is at
least the case for the one per cent of the poor (25th percentile) population considered in Table 9. While
agricultural potential in such areas is not necessary high and hence these areas can only pay limited role in
national wide agricultural growth, to promote the technology that aims at improving such disadvantage
20
condition in these areas such as land management and other farming practice is important for poverty
reduction. This policy issue will be further discussed in the following sections.
Table 10. Agro-ecological conditions across four regions for two percentile household groups
Tigray Amhara Oromiya SNNP
Mean altitude (m)
Mean 1 912 2 097 1 928 1 818
25th percentile 1 685 1 856 1 705 1 549
75th percentile 2 127 2 359 2 167 2 051
Long-run average rainfall (mm)
Mean 687 1 041 1 121 1 235
25th percentile 512 836 746 1 090
75th percentile 902 1 169 1 441 1 453
Long-run coefficient of variation of rain
Mean 0.28 0.26 0.23 0.20
25th percentile 0.21 0.17 0.16 0.15
75th percentile 0.33 0.33 0.28 0.23 Source: Drawn from Table 4.8 in the World Bank (2005).
4. Agricultural-non-agricultural growth linkages in the Ethiopian economy
4.1. Why agricultural growth linkages matter?3
48. The economic importance of agriculture for development has been quantitatively measured in the
literature and the linkage effect is often used in such measure. As agriculture grows, it stimulates series of
economic linkages with the rest of the economy. The resulting demand linkages fall into two broad
categories: production linkages, and consumption linkages.4
49. Production linkages include backward linkages – the input demands by farmers for farm
equipment, pumps, fuel, fertilizer and repair services – as well as forward linkages from agriculture to non-
farm processors of agricultural raw materials. In prosperous agricultural zones, these linkages prove
substantial as pump suppliers, input dealers, grain traders, processing industries and transporters emerge to
supply agricultural inputs and process and distribute farm output. Empirical work on these relationships
has focused on measurement of input-output coefficients to establish the strength of the forward and
backward supply linkages.
50. Consumption linkages include spending by farm families on locally produced consumer goods
and services. Early work in Green Revolution India indicated that higher-income small farmers spent about
half of their incremental farm income on non-farm goods and services as well as another third on
perishable agricultural commodities such as milk, fruit and vegetables (Mellor and Lele 1971). Thus,
consumption linkages from growing farm income can induce sizable second rounds of rural growth via
3. Although the discussion below focuses explicitly on agricultural growth and linkages thereof, the approach
and concepts therein apply to growth in other sectors as well.
4. It is important to emphasise that this study, and almost all studies of its kind, focus on demand linkages
described in the following paragraphs. Aside from demand linkages, there are other inter-sectoral linkages
in an economy. Briefly, these operate via saving and investment (private and public), labour flows, and
transfers including taxes.
21
increased consumer demand for non-agricultural goods and services as well as perishable, high value farm
commodities such as milk, meat and vegetables. In places like India, where many non-farm goods and
services are produced by labour-intensive methods, the spending multipliers not only accelerate growth but
also enhance the equity of agriculture led growth.
51. Following an initial spurt in farm productivity and incomes, production and consumption
linkages together induce second rounds of demand-led growth. Empirical evidence from around the
developing world suggests that a USD 1 increase in agricultural income will generate an additional
USD 0.30 to USD 0.80 income in rural non-farm economy. Linkages are even higher when consumption of
urban-produced products is included. In Africa and Asia, consumption linkages typically account for over
80% total spending linkages.
52. This evidence contrasts with Hirschman‘s claim of feeble agricultural growth linkages
(Hirschman, 1958). Where did Hirschman go wrong? He underestimated agricultural growth linkages in
two very fundamental ways. As Johnston and Kilby (1975) originally pointed out, agricultural technology
changed during the green revolution. The new high yielding varieties demanded pumps, sprayers, fertilizer,
cement, construction labour, and repair facilities from non-agricultural firms, thus generating substantial
backward linkages. Furthermore, considerable milling, processing and distribution of agricultural produce
took place in rural areas, thus generating important forward production linkages as well. The new
agricultural technology fundamentally altered input-output relationships.
53. Still more important were the consumption linkages that Hirschman had ignored altogether. As
Mellor and Lele (1971) originally pointed out, consumption linkages from growing farm income induce
sizable second rounds of rural growth via increased consumer demand. Where new technology or
investment in agriculture leads to increased income, farm families spend large increments of additional
earnings on high-value processed foods and on consumer goods and services such as transport, education,
health, construction and personal services.
54. Available evidence indicates, however, that demand linkages vary considerably across locations
and farm technologies. As Table 11 indicates, they typically prove highest in Asia and lowest in Africa
because of higher input linkages with Green Revolution Asian agriculture and because of higher income
levels which lead to more rapid diversification of consumer spending into non-foods.
55. The accumulating evidence briefly noted in the previous paragraphs induced a shift in the policy
prescriptions concerning sectoral and overall growth. Reversing the industry-first orthodoxy of the 1950's,
the results of Adelman‘s (1984) classic study suggest that agricultural demand-led industrialization can
generate superior growth and equity when contrasted with the alternative of export promoting
industrialization strategies. Better identified and measured growth linkages thus led to the recognition of
agriculture as a potentially powerful engine of economic growth.5
The same evidence also revealed that
this potential is neither present nor equally realizable everywhere.6
Consequently, it became vital to
consider two sets of policy relevant questions. Globally, it is necessary to answer, under what conditions
agriculture can become a leading sector to induce faster growth, how does agriculture grow, and do
government policies matter a lot to agricultural growth. Specific to each economy, it is indispensable to
establish the size of potential agricultural growth linkages and the extent to which it is possible to realize
them.
5. Prominent writers such as Irma Adelman, Peter Hazell, Peter Kilby, Bruce Johnston, Uma Lele, Michael
Lipton and John Mellor highlight the potential power of agriculture-led growth strategies, particularly in
the early stages of economic development.
6. A good summary of the relevant issues and evidence can be found in Sarris (January 2001).
22
Table 11. Agricultural growth linkages: international evidence
Initial Agricultural
Income Growth
Additional Income Growth Source of Linkages (%)
Other
Agriculture
Non-farm
Activities Total Consumption Production
Asia 1.00 0.06 0.58 0.64 81.00 19.00
Africa 1.00 0.17 0.30 0.47 87.00 13.00
Latin America 1.00 0.05 0.21 0.26 42.00 58.00 Source: Haggblade and Hazell (1989).
4.2. Measuring agricultural growth linkages in Ethiopia – a fixed price input-output model
56. A variety of economic models are available for measuring agricultural growth linkages. We first
introduce a fixed-price, semi-input-output (SIO) model as it is a rather simple and spreadsheet operated
model and can be easily used by the analysts in the government.
Fixed price input-output models
57. Early studies of growth linkages most commonly apply some variant of the linear input-
output (IO) model. In its most basic form, the input-output model uses fixed IO coefficients and assumes
fixed prices and perfectly elastic supply in all sectors. In such models, production sectors are linked by
input-output technical relationship as any sector's output can be employed as intermediate inputs in
production of other sectors' output. Second part of the linkages in this kind of model occurs between
consumer's final demand and sector's output. When consumers purchase any good, e.g., a shirt, it is
equivalent to purchase those products such as fibre and yarns used as intermediate inputs in shirt
production. The third part of the linkages is the relationship between income generated from factor
endowments employed in production process and received by consumers as owners of the endowments.
This occurs when a social accounting matrix (SAM) is used in the model and primary inputs (labour, land
and capital) are captured in sectoral production. When more primary inputs are employed by a sector's
production, without constraint on the supply of such inputs, there are income gains to consumers who own
such factor endowments.
58. Without investigating into the functional forms for each product's supply and demand, a linear
input-output relationship is assumed in an IO based model. This implicitly implies a Leontief technology in
production and non-substitution relationship in consumer's demand. While it is a rather strong (but not
unacceptable) assumption in the microeconomic theory, it avoids the arbitrary choice of elasticity either in
production or utility (welfare) function, i.e. there is no any substitution between inputs in a production
function nor substitution in consumption in a utility function. Given that an IO model often composes
disaggregate economic activities with many sectors and multiple consumers, it seems impossible to
econometrically estimate elasticity for all these sectors' supply and consumers' demand functions. Thus, to
assume a fixed relationship between input and output seems to be acceptable.
59. With the non-substitution assumption in both production and consumption, assuming that supply
of any good is not constrained by its inputs, any increase in demand (which can be export demand) leads to
more output with no change in price. From this point of view, a fixed price IO model is 'partial equilibrium'
in its factor market, while it is 'general equilibrium' in the commodity market. In Appendix A, to help
readers understand how the fixed price IO model works, we present a simple model to illustrate it.
60. Unconstrained supply in all sectors is, of course, an unrealistic assumption in any country,
particularly for some sectors. Bigsten and Collier (1995) and Haggblade, Hammer and Hazell (1991), for
example, note that the existence of a real multiplier hinges on the existence of slack resources which can be
pulled into productive activity. In a developing country, given high rates of seasonal labour
underemployment, typically low capital requirements and substantial rates of reported excess capacity in
23
many rural non-farm businesses, a highly elastic supply of rural non-farm goods and services is frequently
an appropriate assumption.7 In contrast, shortages of skilled labour, foreign exchange, and fixed capital
frequently constrain output in the formal industrial sector. Likewise in agriculture, seasonal labour
bottlenecks, land availability, soil fertility, input supply, marketing infrastructure and agro-climate
constraints frequently limit supply responses.
61. Even so, some analysts suggest that agricultural supply elasticities may be high, at least over a
certain range (Delgado et al. 1998; Thorbecke 1994). Anecdotal reports of piles of rotting fruit, unable to
find their way to market, and excess bags of grain unevacuated from specific remote regions bolster these
claims in some, limited circumstance. Yet apart from these episodic special cases, the overwhelming bulk
of empirical evidence points to a low aggregate supply response in agriculture (Binswanger 1989). If
farmers in the developing world could, in fact, increase crop output in unlimited amounts, agriculture
would indeed represent a powerful engine of economic growth, for both malnutrition and poverty would
vanish overnight as hungry farmers availed themselves of this perfectly elastic cornucopia.
62. By ignoring supply constraints altogether, unconstrained input-output multiplier models
exaggerate the size of the inter-sectoral linkages. Given that over half of the reported indirect effects in
these unconstrained models come from demand-induced growth in food grains and other allegedly
elastically supplied agricultural commodities, this questionable assumption biases anticipated indirect
income gains substantially upwards. Side-by-side comparison with alternative formulations suggests that
the unconstrained input-output models overstate agricultural growth multipliers by a factor of two to ten
(Haggblade, Hammer and Hazell 1991).
Fixed price, Semi-Input-Output (SIO) models
63. To better simulate real-world supply rigidities, semi-input-output (SIO) models classify sectors
into two groups, those that are perfectly elastic in supply and those with supply-constrained. The SIO
model permits output responses only in the supply responsive sectors, in which perfectly elastic supply
ensures fixed prices for their output. In the other group, i.e., the group of supply-constrained products,
perfect substitutability between domestic goods and imports guarantees that world prices will ensure fixed
prices for these goods as well. For these models to produce a reasonable approximation of reality, the
supply-constrained sectors often correspond to tradable goods whose domestic supply remains fixed at the
prevailing output price. In these supply-constrained sectors, increases in domestic demand merely reduce
net exports, which then become endogenous to the system and determined by the matrix of semi-input-
output multipliers.
64. As social accounting matrices (SAMs) have grown in popularity, SAM-based multiplier
estimates have emerged to complement and extend the early linkages work. In spite of such improvement,
given that SAM-based multiplier analysis does not capture price effect (and other imperfect substitution
relationship in production and consumption and imperfect elastic in supply, as a CGE model does), the
SAM-based multipliers are formally identical to the IO and SIO models. All require an input-output table
to calculate the production linkages; all adopt fixed prices, fixed input-output coefficients and fixed
marginal budget shares; and all come in unconstrained and constrained versions. The SAMs themselves
become convenient tools for summarizing the raw data and results. They also provide a basis for
incorporating investment, balance of payment, and government accounts. Frequently, given their origin in
poverty and income distribution analyses, the SAMs offer great detail on factor allocation and distribution
of income across household groups. What many in the literature call ―unconstrained SAM-based
multipliers‖ are formally identical to the unconstrained IO models. Similarly, the ―constrained SAM-based
7. Bagachwa (1981), Liedholm and Chuta (1976) and Steel (1977) report rates of excess capacity between
33% and 60% for the countries of Tanzania, Sierra Leone and Ghana. See Bagachwa and Stewart (1992)
for a detailed summary.
24
multipliers‖ are formally identical to the SIO models (Haggblade, Hammer and Hazell, 1991; Lewis and
Thorbecke, 1992). While we use a recently developed SAM for Ethiopia in the following SIO modelling
analysis, we consider only the linkages effect through production and consumption, and ignore the
government and investment that are included in the SAM as two individual institution accounts, given that
the fixed-coefficient method is not a proper tool to analyze investment behaviour and government decision.
65. What does a SIO model reveal? When researchers apply a SIO model for growth linkage
analysis, they need first to assume which sectors in the studied economy are supply-constrained and which
are supply-unconstrained. The analysis then focuses on those sectors that are supply constrained to discuss
their multiplier or linkage effect in the economy by assuming their supply curves shifted outward. For
instance, maize and sorghum, together with other agricultural subsectors, are assumed to be supply-
constrained in the Ethiopian economy in the given time, while trade and some other services, together with
selected manufacturing sectors, are supply-unconstrained. These assumptions imply that, without an
outward shift in maize's supply curve (which should be a vertical line with the assumption of supply-
constrained), output of maize (as well as output for other agricultural production) is fixed in the SIO model
for Ethiopia. Now, assuming that a new technology is adopted by maize producers such that maize supply
curve is shift rightward at the given prices for inputs and output. Then the linkages between such supply
shift in maize production and the rest of the economy are captured by the coefficient values in the column
of the matrix presented on the right side of Equation (4') in Appendix A corresponding to the maize sector
in this matrix. These coefficients indicate the additional demand on the other sectors' output and on the
primary factors with a one unit increase in maize production. Moreover, these input demand coefficients
capture both the direct demand on intermediate and primary inputs employed in the maize production and
indirect demand induced by the input-output linkages across all the sectors and consumer's demand. Such
second, third or more round effects (the cycle continues until all related growth possibilities are exhausted)
are typically focus of growth multiplier analysis. In this particular case, linkages or multiplier effects are
induced by supply shift in maize production curve given that there are sectors which are supply-
unconstrained and they are possible to have a local supply response to a shift in maize supply curve.
However, there will not have supply response for those sectors of which supply is constrained (e.g.
sorghum). For the supply-constrained sectors, increased demand for their products is met by reduced
exports or increased imports (which is often called a leakage effect to the rest of the world). In the present
case, when maize's supply curve shifts outward in Ethiopian economy, the supply response did not occur in
the other agricultural sectors and only occurred in the non-agricultural sectors (e.g. trade and other service
sectors).
66. Obviously, the assumption about which sectors to be supply-constrained and which are supply-
unconstrained affects the model results. In practice, tradability, technological and/or resource constraints,
and capacity utilization condition the extent of the rigidities of supply response. In the present case, given
that Ethiopia is a land-scarce economy, all agricultural production, including crop and livestock, is
assumed to be supply-constrained while most private services are assumed to be supply-unconstrained. In
the case of the manufacturing and other industrial sectors, the domestic manufacturing with less import-
dependency such as meat and dairy processing, grain milling, thread, yarns, fibre, lint and clothing,
together with the construction, and public administration subsectors, are deemed supply-constrained. In
contrast, the reported considerable unused capacity in some manufacturing suggests the possibility of
expanding supply without significant incremental costs.8
Therefore, the manufacturing other than list
above, mining, and utility, (together with most private services), are identified as supply-unconstrained.
8. For instance, CSA survey data (2003a) reveals that capacity utilisation within large/medium-scale
manufacturing averaged just above 50% during 1997/98-2003/04. More tellingly, this rate of capacity
utilization varies very little across years.
25
67. While the choice of which manufacturing sector as supply-constrained is rather arbitrary, it can
be used to capture different types of constraints in supply response across sectors if the local knowledge for
such constraints is available. However, it is common and must to be noted that a SIO model analysis does
not explain why or how the initial increase in a sector‘s output occurs or why certain sectors respond to this
trigger while others do not. For example, in the following section we will discuss the multiplier effect of
selected agricultural and non-agricultural subsectors, assuming their supply curves can be shifted outward
by adoption of any new technology. If such shift is unlikely to occur in reality, of course, the
corresponding linkage effect will not happen too. Moreover, we also conducted a series of sensitivity tests
in which different manufacturing and other industrial sectors are assumed to be supply constrained or
unconstrained. These sensitivity tests are used to evaluate the robustness of the model results.
68. The impacts of growth in the output of four staple crops (teff, barley, maize and sorghum), a crop
fully served as material inputs in textile manufacture(cotton), an export crop (coffee), two livestock sectors
(cattle and poultry), and two agriculture-related manufactures (grain milling and thread/yarns), and the
construction are examined under the conditions specified above. The aim is to provide a contrasting
analysis that can inform the dialogue on investment strategies in Ethiopia. More specifically, the results
allow a comparison across three broad investment strategies for the country: growth in staple crops
(cereals), growth in high value or export crops (cotton and coffee) and growth in manufacturing.
4.3. Results of an Ethiopia fixed price input-output model
Impact on growth
69. Under the assumptions summarized above, we first discuss the growth effect of a given sector by
assuming that this sector's supply curve can be shifted outward with adoption of any new technology. To
do it, we focus on the impact on GDP. Given the economic structure of Ethiopia, the fixed price SIO model
results show that a 1 Birr increase in teff domestic supply generates 1.42 Birr rise in total GDP. The
analogous change in GDP are 1.27 - 1.51 Birr for the other three grain crops, 1.56 and 1.54 Birr for cotton
and coffee, respectively, 1.56 and 1.55 Birr for cattle and poultry, respectively, 0.61 Birr for grain milling,
1.01 Birr for thread/yarns, and 1.08 Birr for construction (see Table 12).9
70. The contrast in the rise in total GDP due to growth in the agricultural sectors or in manufactures
and other industrial sector (construction in this case) is marked. This difference arises primarily because of
smaller value-added generated by the direct increase in manufacturing and other industrial output. As a
fraction of the gross value of output, material inputs used in manufacturing and other industrial range from
51% in thread/yarns to 74% in grain milling. Value added, thus, comprises on 25 to 49% of output only in
these three non-agricultural sectors. By comparison, the value added in total output is much higher in the
agricultural sub-sectors, ranging between 77 and 98%. Consequently, a 1 Birr increase in agricultural
output produces a bigger direct impact on GDP, i.e. GDP produced directed from the targeted sectors. In
the case of cereal production, such direct gains in GDP range between 0.75 and 0.87 Birr; 0.82 and 0.94
Birr for coffee and cotton production; and 0.93 and 0.94 Birr for poultry and cattle production,
respectively. On the other hand, an identical expansion in the three non-agricultural outputs can only
generate 0.20 to 0.43 Birr direct increase in GDP.
9. It is a misunderstanding that the linkage effect presented here tries to pick up a 'driving' sector and define
others as 'following' sectors (see, for example, Dercon and Zeitlin, 2009, for such kind of explanation). In
contrast, it is impossible for a SIO model to identify which sector is or should be a driver for the economy-
wide growth. Linkages can only occur when many sectors grow simultaneously. Rather, the linkage
analysis emphasizes second and third round demand side effects in stimulating growth. Such indirect or
general equilibrium effect of any sector's growth is unlikely to be observed in a typical partial equilibrium
analysis focusing on a particular sector.
26
71. We are more interested in the linkage effect between growth in the targeted subsector and the
gains in the rest of the economy, and such linkage effect is captured by the gap between the direct increase
in GDP and the total change in GDP (i.e. the difference between the first and second lows in Table 12).
Such indirect effect on total GDP is originated from the second, third, and more round linkages between
targeted subsector and the rest of the economy. The analysis shows that, when the SAM can be developed
with highly disaggregated production sectors, which is the case of the current Ethiopian SAM10
used in this
study, more indirect production linkages have been captured, particularly within manufacturing sectors.
For example, the SAM includes six textile subsectors along the textile production process. Some sub-
sectors are downstream activities of others in this process (e.g. fibre and lint are inputs of thread and yarns,
which, in return, are inputs of textile and clothing). With such details in manufacturing sectors, we can
further decompose production linkages into two parts: direct and indirect. We treat the direct impact on
GDP, the sector's value added as the direct production linkages, while the rest of production linkages that
generate through multiple rounds of intermediate demand as indirect linkages.
72. Given that Ethiopia's agricultural production is dominated by the use of primary inputs, indirect
production linkages induced by agricultural growth are quite small. As shown in Table 12, such linkage
effect accounts for 7.3% to 12.2% of total gains in GDP. On the other hand, indirect production linkages as
shares of total GDP gains are much larger for the non-agricultural growth, counting for 33.7% - 53.6% in
the three non-agricultural sectors. The strong linkages in the manufacturing production process are the
main reasons for Hirschman to emphasize the importance of industrialization in the development process,
as "the superiority of manufacturing in this respect is crushing" (Hirschman 1958, p110). Focusing
typically on such production linkages the governments of many developing countries emphasized the
importance of industrialization and have adopted an import substitution development strategy in the 1960s
and 1970s.
73. As some development economists pointed out (see, for example, Mellor, 1995), Hirschman
theory emphasizes only the linkages within production process, while consumption linkages have been
ignored. Consumption linkages can only be captured in a general equilibrium setup in which both
consumption and production are endogenous. Agricultural economists emphasize more such linkages, and
believe that they are particularly important in the early development stage, when food and other basic need
dominate consumers' consumption bundle, and when most such products are produced locally where a
majority of population resides in (Mellor 1995, p13).
74. To assess whether and how important consumption linkages are for Ethiopia's economy today,
the SIO model is used to further decompose the consumption linkages from the production ones. What SIO
model captures is constrained by the structure of the SAM, which represents the country's current
economic structure. Comparing the third and first rows in Table 12 it shows that consumption linkages
account for one-third total GDP gains created via growth either in agriculture or non-agriculture, though
the absolute level of income effect due to the consumption linkages is quite different between agriculture
and non-agriculture. However, when the indirect effect is considered, consumption linkages account for
85% to 93% of indirect effect of agricultural subsector growth on total GDP, except for coffee, in which
such linkages account for 78.9% of total indirect effect on GDP. In contrast, consumption linkages are
relatively weak in the non-agricultural growth, ranging between 47% and 66% in the three non-agricultural
sectors selected (Table 12, last row). Such results are consistent with findings in the literature using similar
method but focusing on other low income developing countries.
10. The current Ethiopian SAM includes 24 agricultural subsectors, 29 industrial and 11 service sectors.
27
Table 12. Impact on Growth due to one unit of increase in selected sectors' output
Cereals
Industrial
materials/Exportable
Livestock
Manufacturing Other industry
Teff Barley Maize Sorghum Cotton Coffee Cattle Poultry
Grain
milling
Thread
& yarns Construction
Change in GDP 1.42 1.27 1.47 1.51 1.56 1.54 1.56 1.55 0.61 1.01 1.08
Production linkages 0.90 0.80 0.93 0.95 0.98 0.97 0.98 0.98 0.38 0.63 0.69
Direct (sector's own
value-added) 0.83 0.75 0.87 0.86 0.94 0.82 0.94 0.93 0.20 0.43 0.25
Indirect (others) 0.07 0.06 0.06 0.10 0.05 0.15 0.05 0.04 0.18 0.20 0.44
Consumption linkages 0.52 0.47 0.54 0.56 0.58 0.57 0.57 0.57 0.23 0.38 0.39
Value-added
multiplier 1.60 1.58 1.57 1.65 1.56 1.75 1.56 1.55 2.68 2.10 3.49
Share in total indirect impact on GDP (%)
Indirect production
linkages 12.2 10.6 9.8 14.9 7.6 21.1 7.5 7.3 44.3 33.7 52.6
Consumption linkages 87.8 89.4 90.2 85.1 92.4 78.9 92.5 92.7 55.7 66.3 47.4 Source: Ethiopia SIO model results.
28
75. In short, almost all sectors generate large linkages. But those induced by agricultural activities are
larger, primarily because of larger initial value added (income). While the second round linkages across
sectors through increased intermediate demand are stronger in the industrial sectors than in the agricultural
sectors, given that current agriculture is dominated by the use of primary inputs, the third round linkages
through consumers' spending on local agricultural and non-agricultural goods and services dominate the
overall indirect effect of sector growth on total GDP gains. Moreover, consumption linkages are much
stronger in the agricultural growth than that in the growth led by the industrial sectors and such linkages
are often ignored in designing a country's development strategy. Obviously, for a low-income country with
large size of population (as in the case of Ethiopia), consumption linkages may play much larger role in
growth and development. Because this, a development strategy that can stimulate income growth to a
majority of population will definitely generate a large impact on the long term growth and development.
Impact on the level and distribution of incomes
76. Impact on the level of household income can also be analyzed using the SIO model results. In the
Ethiopian SAM, households are classified into four aggregated groups, rural poor households, rural non-
poor, urban poor households, and urban non-poor. The income levels used to define the country's poverty
rate in the rural and urban areas are the thresholds to distinguish the poor from not poor in both rural and
urban areas in the SAM. While income sources differ for the rural and urban households and between the
poor and not poor, consumption patterns are also different. In general, the rural households spend more
income on food, given the income level is lower for an average rural household than for an urban
household. In the same area (i.e. in rural or urban), the poor households spend more income on food than
the non-poor households. In order to capture the marginal propensity to consume, we apply marginal
budget shares (MBS), instead of average budget shares, in the SIO model, such that increases in
consumption due to income growth are not necessary the same as their current consumption patterns for
both poor and non-poor households in rural or urban areas. The MBS has to be econometrically estimated
using household survey data. We adapt the method developed by King (reference) and the estimate results
are in Appendix.
77. Table 13 first reports the change in income induced by growth among the four household groups.
The results show that, in general, rural households benefit more from agricultural growth as most income
gains go to the two rural household groups, while urban households benefit more from manufacturing and
other industrial growth. A 1 Birr increase in any of agricultural sector's output will generate 1.23 to 1.51
Birr rise in total household income, while an equivalent change in any of the three non-agricultural sectors'
output has much lower positive impact on household incomes.
78. To investigate the distributional effect of growth, we calculate the share of income gains across
the four household groups in the second panel of Table 13. We also calculate the ratio of this share over
corresponding household's income share in the SAM in the third panel of Table 13. When the ratio is
greater than 1 for a particular household, it indicates that this household gains disproportionally more than
what its income current shares in the economy. The second part of Table 13 shows that more than 50% of
additional income generated by agricultural growth goes to the rural non-poor household group, while the
rural poor household group gets about 13% to 33%. The share of additional income going to urban poor
household group is extremely low, between 2.7% and 3.3%. In total rural households receive 80% - 85% of
total gains in income when such income is generated from agriculture-led growth. On the other hand, the
income share of urban non-poor household group in total income gain more than doubles when growth is
driven by the non-agricultural sectors (the three sections reported in the table). In total about 40% of such
income goes to the urban households, while these households account for 15% - 20% of population in the
country.
29
79. The other way to assess income distribution effect of growth is to analyze income change for
each individual household group. We do it by a ratio of income share in the additional income generated
from growth over the share of this household's income in total income currently (in the SAM). The ratio of
one for a household indicates that this household is not affected by the change in income distribution,
while the ratio being greater (less) than one for a household tells that the new income distribution as an
outcome of growth is in favourable (against) this household. As shown in the third panel of Table 13, when
growth is led by staple agricultural subsectors, including staple crops and livestock products, the
distribution of additional income gain is in favourable of the rural poor household group, as ratio for this
group is consistently higher than 1 in the growth led by all the six staple subsectors (ranging from 1.01 in
the case of sorghum to 1.35 for cattle led growth). While income distribution led by staple crop growth
also favours rural nonpoor household group, this group disproportionally benefits more from export
agricultural growth such as coffee and cotton. When growth is led by the non-agricultural sectors, the new
income distribution moves to the direction benefiting the urban households, while benefits going to the
rural households are proportionally smaller than the size of these households' income in the economy in the
initial situation. While such different income distribution effects of growth are the outcome in which price
effect is ignored, it still provides us meaningful message in terms of a complicated relationship between
growth and income distribution. With the rural households forming the bulk of the poor, the potential
impact on the poverty, through income distribution effect, is thus greatest with growth in staple crops and
livestock production. A more direct measurement of the pro-poorness of different growth options will be
analyzed in the following section using an economy-wide model with endogenized prices.
30
Table 13. Impact on Income due to one unit of increase in sectors' output
Cereals
Industrial
materials/Exportable
Livestock
Manufacturing
Other
industry
Teff Barley Maize Sorghum Cotton Coffee Cattle Poultry
Grain
milling
Thread
& yarns Construction
Rural poor 0.40 0.38 0.38 0.37 0.20 0.20 0.51 0.50 0.12 0.20 0.21
Rural nonpoor 0.76 0.66 0.82 0.86 1.09 1.02 0.78 0.78 0.23 0.36 0.39
Urban poor 0.04 0.04 0.04 0.04 0.04 0.05 0.04 0.04 0.04 0.11 0.08
Urban nonpoor 0.17 0.15 0.17 0.19 0.18 0.21 0.18 0.18 0.19 0.29 0.31
Total 1.37 1.23 1.42 1.46 1.51 1.48 1.50 1.49 0.58 0.96 0.99
Share in total income gains (%)
Rural poor 28.8 30.8 26.8 25.2 13.4 13.8 33.7 33.4 20.6 21.1 20.9
Rural nonpoor 55.5 53.8 58.0 58.7 71.9 68.8 51.7 52.0 39.8 37.5 39.5
Urban poor 2.9 2.9 2.8 3.0 2.7 3.3 2.7 2.7 7.6 11.1 7.6
Urban nonpoor 12.7 12.5 12.3 13.1 12.0 14.2 11.9 11.9 32.0 30.3 31.9
Ratio to income share in SAM
Rural poor 1.15 1.24 1.08 1.01 0.54 0.55 1.35 1.34 0.83 0.84 0.84
Rural nonpoor 1.04 1.01 1.09 1.10 1.35 1.29 0.97 0.97 0.75 0.70 0.74
Urban poor 0.70 0.69 0.68 0.72 0.66 0.78 0.65 0.65 1.82 2.68 1.84
Urban nonpoor 0.73 0.71 0.71 0.75 0.68 0.81 0.68 0.68 1.83 1.73 1.83 Source: Ethiopia SIO model results.
31
Important role of service sector in determining the linkages effect
80. When a SIO model is used to measure the economic linkages or the multiplier effect, the model
does it by evaluating selected sector by sector. Because this, it seems to generate certain misunderstanding
in explaining the model results and some papers criticized that such analysis is 'too narrow and focused on
single-sector growth' (see, for example, Dercon and Zeiline, 2009). To clarify such misunderstanding, it is
necessary to further assess how the ability of supply response in the other sectors affects the direction and
extent of linkages effect of the studied sectors. In the above analysis, we assume that all private service
sectors are supply-unconstrained and they positively respond to the supply shift of studied agricultural and
manufacturing sectors. To assess how important service sectors in determining multiplier effect of growth
in the agricultural and manufacturing sectors, we conduct an additional scenario in which supply of the
four subsectors in the services, trade, transport, financial services, and the other business service, becomes
constrained while the assumptions on the other sectors are the same as those in the previous analysis. Such
constraints can be justified by lack of infrastructure in road, market and information such that market
accessibility of agricultural and non-agricultural producers is limited, or due to institutional and policy
barriers such that the difficulties to do business in the service sectors limit the possible response of the
private sector to increased supply and demand. In Table 14 we compare the model results with such
constrained services with those without such constraints reported in Table 12.
81. As expected, constraints in the supply in the services lower the gains in GDP induced by a one
unit increase in selected sectors' output. However, such constraints do not change the relative extent of
total GDP gains between growth in the agricultural and in the non-agricultural sectors. While lack of
supply response in the service sectors limits the economic linkage effect of growth in the production sector,
the constraint on the growth in the manufacturing and other industrial subsectors is more than on the
growth in the agricultural sectors. As shown in last panel of Table 14, the constraint in service supply
lowers total GDP gains by 13% - 24% when such gain is due to growth in the agricultural sectors, while it
lowers total GDP gains by 33% - 43% when the gain is due to growth in the non-agricultural sectors. When
GDP gains are decomposed into the two parts, i.e., due to production or due to consumption linkages,
Table 14 shows that lowered consumption linkages effect is the main channel for the service sectors to
constrain the linkages in the economy when agriculture is the trigger for such linkages, while both
production and consumption linkages are equally important channels through which the service sectors
constrain the linkages effect when the non-agricultural subsectors are the trigger for such linkages.
32
Table 14. Importance of services in measuring multiplier effect
Teff Barley Maize Sorghum Cotton Coffee Cattle Poultry
Grain
milling
Thread
& yarns Construction
Total gains in GDP
Supply-unconstrained
services 1.42 1.27 1.47 1.51 1.56 1.54 1.56 1.55 0.61 1.01 1.08
Supply-constrained services 1.21 1.09 1.26 1.23 1.34 1.17 1.35 1.35 0.35 0.68 0.66
Production linkages
Supply-unconstrained
services 0.90 0.80 0.93 0.95 0.98 0.97 0.98 0.98 0.38 0.63 0.69
Supply-constrained services 0.85 0.76 0.89 0.86 0.94 0.82 0.94 0.94 0.24 0.48 0.47
Consumption linkages
Supply-unconstrained
services 0.52 0.47 0.54 0.56 0.58 0.57 0.57 0.57 0.23 0.38 0.39
Supply-constrained services 0.36 0.33 0.38 0.37 0.40 0.35 0.40 0.40 0.10 0.20 0.18
Value-added multiplier
Supply-unconstrained
services 1.60 1.58 1.57 1.65 1.56 1.75 1.56 1.55 2.68 2.10 3.49
Supply-constrained services 1.37 1.37 1.37 1.36 1.36 1.36 1.36 1.37 1.51 1.43 2.03
Ratio between constrained and unconstrained services
Total gains in GDP 0.85 0.86 0.86 0.81 0.86 0.76 0.87 0.87 0.57 0.67 0.61
Production linkages 0.94 0.95 0.96 0.90 0.96 0.85 0.96 0.96 0.64 0.76 0.69
Consumption linkages 0.69 0.69 0.70 0.66 0.70 0.62 0.70 0.71 0.45 0.52 0.47
Value-added multipliers 0.86 0.86 0.87 0.82 0.87 0.77 0.87 0.88 0.56 0.68 0.58 Source: Ethiopia SIO model results.
33
Summary of agricultural linkage effect assessed by a SIO model
82. While the application of a SIO model cannot explicitly assess the source of growth, given the
constraints in land and other resources faced by the Ethiopian economy, it is obvious that technological
change and hence productivity must be the main source of sustainable growth. Alternative growth options
have significant differences in the direction and extent of growth linkages. Growth in agriculture produces
stronger linkages with the rest of economy than growth in non-agriculture. The potential benefits of
stimulating growth in agricultural production (albeit differentiated by products) are thus substantial.
Nevertheless, the size of this potential as well as the extent of its realization depends on a parallel
expansion in non-agricultural sectors (particularly in those associated with growing input or consumption
demand). Stronger linkage effect of agricultural growth is not only due to its use of primary input, but also
consumption linkages. The study shows that at Ethiopia's current development stage, consumption linkages
are more important than the second-run production linkages, which are high among the manufacturing
sectors. Consumption linkages are particularly high for growth led by the agricultural sectors, indicating
that the importance of agriculture to the overall economic growth is not only due to its size in the economy,
but also due the stage of development in which, at current, a majority of population in Ethiopia is poor who
consume most goods and services (including non-agricultural goods and services) produced locally.
83. When households can be disaggregated according to different income levels and rural and urban
location in a SAM, the SIO model can be used to assess income distribution effect of growth. While
ignoring price effect is a critical shortcoming of a SIO model, the analysis does provide us meaningful
message about the complicated relationship between growth and income distribution. Staples led growth
has shown to benefit more the rural households, particularly the rural poor households, while export crop
led growth seems to benefit rural non-poor households disproportionally. When growth is led by the
manufacturing sectors, urban households are major beneficial. Income distribution is worse and income
gap between the rural and urban households increases under such growth options. While the SIO model
cannot explicitly measure the poverty impact of alternative growth options, it implicitly indicates the most
effectiveness of staple led growth in poverty reduction.
84. It should note that the SIO model results depend on the assumption of constrained or
unconstrained supply elasticity among various non-agricultural sectors. By varying the set of non-
agricultural sectors assumed to be supply unconstrained show different results. In order to realize the
important role of services in measuring linkages effect, we altered the assumption about unconstrained
service supply in an additional scenario and assumed that trade, transport, financial and other business
services are constrained in their response to the growth in the studied agricultural and non-agricultural
subsectors. To do it we find that while the extent of linkages effect is lowered in all cases, growth induced
by the non-agricultural sector is constrained more by the service sectors. Barriers to growth due to
constraints faced by the service sector should be paid more attention in a development strategy in which
either agriculture or manufacturing is emphasized.
85. It has to point out that with globalization, growth linkages seem to be weakening in the
manufacturing sector, as export-oriented manufacturing does not need to depend on domestic markets for
input supply and output demand (reference for this argument). While it is true that many developing
countries want to pursue such growth path, following the success in China and other Asian countries'
experience, the current Ethiopian economy actually shows unutilized capacity among many manufacturing
sectors, even at the development level where manufacturing accounts for a small share of the economy.
The industrial survey shows that the main reason for unutilized factory capacity is due to domestic demand
constraint. When such firms cannot produce goods that are for domestic market, can they produce for
exports which often requires much better quality and more competitive in price? Moreover, globalization is
a two-way thing. When domestic firms have opportunities to export their products using imported goods as
input materials, they are also facing stronger competitions from imported goods they produce in their own
34
country's market. We observe that in many African countries' domestic markets have become increasingly
dependent on imported manufacturing goods and de-industrialization has occurred among some
manufacturing sectors. Obviously, competitiveness in export and domestic markets has become more or
less as a same thing. Without increasing domestic market demand and improving competition in domestic
market, it is unlikely to expect Ethiopia to jump into international market for manufacturing exports.
Production linkages effect analyzed in this study, thus, is still highly relevant to the current Ethiopian
economy.
86. While the SIO model is a useful tool for linkage analysis, its rigidity in prices is obvious an
improper assumption given that linkages are primarily through the interactions between supply and
demand in the domestic market. In order to overcome such shortcomings as well as the arbitrary
assumptions of constrained and unconstrained supply elasticity, an economy-wide model is developed and
applied to further assess the growth linkages and their impact on income and poverty reduction. In this
economy-wide model, not only domestic prices are endogenous and all economic sectors can respond to
changes in prices with different levels of elasticity, the economy is further disaggregated geographically.
Results from both models are compared and the complementary roles of the two models are also discussed
at the end of this section.
4.4. Growth linkages in the Ethiopian economy – an economy-wide multimarket model
87. Ethiopia is an open economy. However, as a land-locked country with high transportation and
other marketing costs, partly explained by considerable geographic distances and an inadequate road
network, prevent world market prices from automatically translating into domestic prices. As a
consequence, many commodity prices, especially those of agricultural products, are actually determined by
supply and demand conditions in the domestic market. For these reasons, it is necessary to take into
account for the interaction of prices and growth (the price effect) in analyzing agriculture-non-agriculture
linkages. Accordingly, in this sub-section we develop an economy-wide multimarket (EMM) model in
which prices of most agricultural and some non-agricultural products are endogenous variables. We apply
this EMM model to Ethiopian economy to further assess the importance of the agricultural sector in growth
and poverty reduction.
Why an EMM model?
88. It is possible to develop a CGE model based on the same SAM used in the SIO model analysis
discussed in the previous chapter. Indeed, the CGE approach is more preferable for economy wide analysis
and more comparable with above SIO analysis when the same SAM is used for both models. However, as
shown in the above analysis, the current SAM is for the national economy and there are only four
aggregate households at the national level. Given the current Ethiopian economy is still dominated by
agriculture and more than 80% of population live in the rural areas, special attention must be paid to the
structure of agriculture in the linkage analysis. It is well known that agricultural production systems are
typically characterized by the interactions between human behaviours and natural environment. With
heterogeneity in agro-ecological, social and economic conditions, agriculture in Ethiopia is highly
diversified. To analyze the economic linkages in the country, it is necessary to understand the role of
different agricultural subsectors at different locations of the country. Moreover, agricultural development is
constrained by market opportunities and conditions of market access provide different such opportunities
to different locations in the country.
89. To take into consideration of geographic factors such as agro-ecological conditions, population
distribution, production and market locations and connections and in order to better present the agricultural
sector and rural economy in the linkage analysis, we have developed a highly disaggregated EMM model
for Ethiopia. Most multimarket models focus on particular subsectors of agriculture or segments of
35
economies. The model developed for this study focuses on agriculture but puts the agricultural sector in an
economy-wide context, such that the model can be used for the economy-wide linkage analysis. The
original EMM model was developed and applied for the food security analysis in Ethiopia (Diao and Nin
Pratt, 2007). This model is extended and modified for this study in order to be consistent with the non-
agricultural economy described by the SAM discussed above, while detail agricultural sectors are still kept
as before.
90. Specifically, there are 32 agricultural commodities or commodity groups (see Table B1 in the
Appendix B for a list of agricultural commodities/sectors included in the model) in the EMM model. In
contrast with the SAM that represents the national economy, both agricultural production and consumption
in the EMM model are further disaggregated into sub-national regions in order to capture the geographic
heterogeneity of sectors and households. Limited by the data, the model captures totally 56 administrative
zones and all agricultural supply and demand functions are defined at the zonal level. Detail description of
the EMM model can be found in Appendix A, while a number of key results concerning impact on growth
and poverty are reported as following.
Three Ethiopias - areas of food deficit, food balanced, and food surplus
91. With highly spatially disaggregated information, Ethiopia can be examined according to sources
of domestic food availability, resulting in a division of Ethiopia into three categories: areas of food deficit,
food balanced, and food surplus (Figure 4). Based on data from Ethiopia‘s 2001/02 Agricultural Census,
woredas in which the average cereal equivalent output per rural household is 20% below the national
average fall into the food deficit area, those with output between 80% and 120% of the national average
form the food balanced area, and those with output 20% or more above than national average constitute the
food surplus area.11
11. The study includes 460 woredas. Cereal output equivalents were used to represent food availability.
Equivalents include cereals, pulses, oil crops, and root crops, and account for over 60% of household food
consumption in the urban and 70% in rural area. The conversion ratio for crops other than cereals was
based on their calorie content (see the FAOSTAT web site).
36
Figure 4. Food deficit, food balanced, and food surplus areas
Source: Constructed by authors based on Democratic Republic of Ethiopia (2002).
92. Almost 30 million of Ethiopians live in the food deficit area, where the annual food availability
averages only about 530 kilograms per household, even in a good year.12
This represents half the national
average (Table 15). In contrast, food availability per household in the food surplus averages
1 800 kilograms, which is 70% above the national average. The high proportion of cereals and other staple
crops in the food availability calculation (more than 70% of rural household food consumption) is
indicative of extremely low food availability and alarming food insecurity, in turn a reflection of very low
income levels per capita and a very high rate of poverty. Compared with the rural poverty rate of 46%
nationwide,13
the poverty rate in the food deficit area is 60%; in the food surplus area it is less than 40%.
Fifty per cent of the rural poor now live in the food deficit area; that area, however, only accounts for 37%
of the total rural population.
93. A major constraint to meeting food demand for the majority of rural households in the food
deficit area is extremely small farmland area. National farm size, including permanent and temporal crops,
averages about one hectare. In the food deficit area, however, farm size averages only 0.57 hectare
compared with 1.38 hectares in the food surplus area (Table 15). Of the 184 woredas constituting the food
deficit area, per household farmland is less than 0.4 hectares in half of them, and less than 0.3 hectares in
one-third of them. Cereal production yields are also lower than the national average, further eroding food
security in these areas. The average cereal yield in the food deficit area is about one metric tonne per
hectare, 20% below the national average and 30% below yields in the food surplus area (Table 16). Even
taking other staple crops into account, a significant yield gap in staple crop production still exists between
the food deficit and food surplus areas.
12. The calculation is based on the data for 2001/02, which is a good harvest year for most areas in the
country.
13. The poverty rate used in this study is consistent with data from HICES 1999/2000 (CSA 2000a).
The three areas are based on woreda-level ratios of cereal equivalent output per household to the national average:
Food deficit area—ratio of less than 0.8 Food balanced area—ratio of between 0.8 and 1.2 Food surplus area—ratio of greater than 1.2
37
Table 15. Population and poverty rates in the three areas
Indicator Food deficit areaa
Food balanced areab
Food surplus areac
National level
Total population 25.6 22.1 22.3 70.0
Rural 21.9 19.7 17.2 58.9
Urban 3.7 2.4 5.0 11.1
Share of population
Rural 37.3 33.4 29.3 100.0
Urban 33.0 21.7 45.3 100.0
Share of poor people
Rural 49.1 25.8 25.1 100.0
Urban 20.3 29.1 50.6 100.0
Poverty rate
Rural 60.5 35.4 39.0 45.8
Urban 22.6 49.2 41.0 37.0
Source: Calculated by authors from Federal Democratic Republic of Ethiopia (2002). aWoredas with cereal equivalent output per rural household at levels 20% below the national average.
bWoredas with cereal equivalent output per rural household at levels of 80%–120% of national average.
cWoredas with cereal equivalent output per rural household at levels 20% higher than the national average.
Table 16. Land size and cereal output per household in the three areas
Woreda-level rural household average Food
deficit area
Food
balanced area
Food
surplus area
National
level
Cereal land holding (hectares per household) 0.41 0.74 1.07 0.70
Farmland holding (hectares per household) 0.57 0.94 1.38 0.90
Cereal output (kilograms per household) 418 883 1 579 904
Cereal equivalent output (kilograms per household) 534 1 078 1 814 1 079
Source: Calculated by authors from Federal Democratic Republic of Ethiopia (2002).
94. Given a high population density in most of Ethiopia‘s rural areas, increasing land productivity is
the only feasible strategy for improving food security. The intensity of labour use and other inputs is often
linked to population pressure (Boserup 1965), a reality also reflected in international trends. Fewer modern
inputs are used in the food deficit area than in the food surplus area. For example, only 29% of cereal land
is fertilized in the food deficit area compared with a national average of about 40% and a food surplus area
rate of 56%. Returns to modern inputs, in terms of yield increases, are also low in the food deficit area
compared with those in the surplus area (Table 17).
38
Table 17. Cereal yield and input use in the three areas
Indicator
Food
deficit area
Food balanced
area
Food surplus
area
National
level
Cereal yield (tonnes per hectare) 1.08 1.19 1.44 1.28
Cereal equivalent yield (tonnes per hectare) 1.14 1.15 1.32 1.22
Cereal yield using fertilizer only (tonnes per hectare) 1.24 1.25 1.44 1.36
Cereal yield using fertilizer and improved seed (tonnes
per hectare) 1.65 2.20 2.63 2.46
Fertilizer use rate in cereals (per cent) 29.12 26.40 56.13 40.21
Fertilizer combined with seed rate (per cent) 3.08 3.15 4.88 3.91
Source: Calculated by authors from Federal Democratic Republic of Ethiopia (2002).
95. Certain agro ecological conditions, such as soil moisture, affect the feasibility and efficiency of
fertilizer use. Using the growth period as an indicator of agro climatic conditions, woredas were spatially
grouped according to two agricultural domains: high agricultural potential with a maximum growth period
of more than six months, and low agricultural potential with a maximum growth period of less than six
months. Surprisingly, 70% of woredas and 80% of rural households in the food deficit area were classified
as having high agricultural potential; this compared with 90% of both woredas and rural households in the
food surplus area. There is no significant difference in the ratio of fertilized cereal area to total area in the
two domains within the food deficit or food surplus areas. An econometric test further proves that
differences in the agricultural potential cannot explain the difference in fertilizer use or the cereal yield gap
between these areas.
96. Given the absence of household-level data, further analysis of factors affecting production
decisions by farmers, including input use, were not possible.14
Nevertheless, findings from woreda-level
data indicate a significant yield gap and, thus, potential for improving land productivity in those areas
dealing with severe food insecurity. We now turn to the EMM model analysis and the above spatial
patterns of Ethiopian agriculture are captured in the analysis.
97. A large number of previous studies have concluded that agriculture, especially food crops, have
strong growth linkages and multiplier effects; that is, increased agricultural (or food crop) production
would generate a disproportionately large increase in the country‘s total GDP, through increased demand
for inputs, and more importantly, through increased consumption demand as a result of higher agricultural
incomes.15
As the SIO model of the previous chapter, the EMM model is first used to derive sector-level
growth multipliers, deriving from total factor productivity (TFP) shocks in corresponding agricultural sub-
sectors.
98. Prior to the comparative analysis of agriculture-non-agriculture growth linkages, the EMM model
is employed to assess a business-as-usual scenario (also known as the ―baseline‖) in which the economy is
assumed to grow following its current trajectory through 2015 (and 2004 is the base-year used in the
model). The business-as-usual growth path is based on average agricultural and non-agricultural growth
trends for 1995–2004, during which time about 70% and 50% of the increase in total crop production and
cereal production, respectively, resulted from area expansion. Over the same period, the cereal production
14. Obtained Agricultural Census data were aggregated to the woreda level.
15. See Bell and Hazell (1980) for an early methodological discussion of alternative multiplier models used in
growth linkage analysis, and the discussion of Haggblade, Hammer, and Hazell (1991) on the improvement
in the multiplier models with limited price endogeneity.
39
growth rate was 2.9% per year – 0.4% higher than the 2.5% population growth rate – and the growth rates
of total staple crop and cereal yields were about 0.8% and 1.5% per year, respectively. Under the business-
as-usual scenario to 2015, and based on livestock production growth of 4.1% per year and non-agricultural
growth of 5.3% per year, GDP is projected to increase at 4.5% per year, and AgGDP at 3.7% per year.
99. On this basis, the livelihood of the majority of rural Ethiopians will not get significantly
improved by 2015. The national poverty rate will fall to 32.1% by 2015, from the high 2003 level of
44.4%. Given 2.5% yearly population growth during 2003–15, the decline in the number of people living
below the poverty line will only occur in the urban areas, while the number of the poor in the rural areas is
estimated to increase by 83 thousand by 2015.
Staples production has stronger growth linkages over time than export-oriented production
100. It is thus clear from the business-as-usual scenario without additional growth in both agriculture
and non-agriculture, it will be impossible for the country to meet the first MDG of halving the poverty rate
by 2015. On the other hand, achieving the objectives of halving poverty requires a greater understanding of
which sub-sectors can best induce the economy-wide growth and cut poverty faster. Hence, this section
focuses on an evaluation of two broad agricultural sub-sectors in terms of the country‘s growth and poverty
reduction strategy. The two sub-sectors are staples (cereals, root crops, pulses, oilseeds and livestock) and
exportables (coffee, selected fruits and vegetables, cotton, chat, sesame seed, sugar, and other horticultural
products). Specifically, these sub-sectors‘ contribution is assessed by exogenously increasing the
productivity growth rate of one sub-sector, while maintaining the growth of the others at their baseline
levels.
Table 18. Agricultural and non-agricultural growth rate in the simulations
Growth Rate
Base-
run
Staple crop led
growth
Export crop led
growth
Agriculture led
growth
Non-agriculture
led growth
GDP growth rate 4.5 5.5 5.5 5.5 5.5
Ag GDP growth rate 3.7 4.1 5.7 4.5 4.2
NonAg GDP growth rate 5.3 6.8 5.4 6.4 6.7
Total staple crop and livestock growth rate 3.1 4.9 3.0 4.3 3.2
Cereal output growth rate 2.9 4.5 2.9 4.2 3.0
Livestock output growth rate 4.1 6.4 4.0 5.2 4.2
Total high value crop growth rate 3.1 2.7 15.6 7.7 3.0
Traditional export crop growth rate 4.0 3.7 15.0 8.0 4.0
Non-traditional exports growth rate 8.0 4.8 31.2 18.3 6.4 Note: 'A sector-led growth' is defined by an exogenous productivity shock imposed in this sector, which endogenously induces growth in the other sector. For example, in the column called 'staple crop led growth', exogenous shock in sector's productivity is imposed on cereal and livestock production, while differences in the other sectors' growth rate in this scenario from that in the base-run are the endogenous results through linkages effect. Source: Authors calculation from the EMM model results.
101. With more than 80% of AgGDP and 40% of GDP, staples represent the largest agricultural sub-
sector in terms of value-added. In contrast, the export subsector constitutes quite small shares accounting
for about 10% of AgGDP and 5% of total GDP. Thus, the simulated additional annual growth for cereals‘
productivity was first determined, at 1.5%, which implies 2.2% additional annual growth in livestock. In
total, additional 1.8% of annual growth rate is obtained for the aggregated staple food sector (staple crops
and livestock). With such growth rate in the staple sector, total GDP will grow at 5.5% (partly through
strong linkage effects on the non-agricultural sector that will be discussed later). In order to produce the
same 5.5% of GDP growth rate, the agricultural exports sector needs to grow at 15.6%, with additional
12.5% of annual growth compared with the base-run (Table 18).
40
102. To make the impacts more clearly comparable growth multipliers are used. The multipliers are
defined as the total increase in real GDP divided by the increase in the shocked sector‘s total output, both
measured at the initial (base-year) level of prices. The resulting multipliers derived using an economy wide
and endogenous price models are in general relatively smaller than the standard fixed-price multipliers.16
Our model‘s simulation results show that the staple sector‘s growth multipliers are consistently greater
than one and increase overtime (Figure 5). These results imply that one unit (not one per cent) increase in
staple production will generate more than one unit of increase in total GDP. Moreover, such growth
linkages become stronger over time. For example, one unit of increase in staple production can have 1.03 –
1.12 units of increase in total GDP in the first five years in the simulation, while the same one unit of
increase in staple production will generate 1.29 units of GDP by 2015. On the other hand, the linkages
from agricultural export sector to total GDP is strong only in the initial five years, while the linkages
become weaker overtime, and the growth multipliers fall to below one by 2015.
Figure 5. GDP growth multipliers in staple and export agricultural growth scenarios
Both with 5.5% GDP annual growth
0.9
1.0
1.1
1.2
1.3
1.4
1.5
2004 2006 2008 2010 2012 2014
One-unit staple output One-unit export ag ouput
Source: Authors calculation from the EMM model results.
103. Different from the discussion in the previous section about the SIO model results in which the
linkages effect on the level of total GDP is decomposed into production and consumption linkages, in the
EMM model that is dynamic and runs for a period of 10 years, we focus on the linkages effect on the GDP
growth rate that can be decomposed into growth rate in the agricultural and non-agricultural sector. The
model result shows that the strong growth linkage effect in the economy induced by the staples growth is
mainly due to growth in the non-agricultural sector in response to the growth in the staples. In the scenario
in which exogenous productivity growth is only assumed for the staple sector, the non-agricultural GDP‘s
average annual growth rises to 6.8%, from 5.3% in the base-run (Table 18, third row, first and second
columns). The additional 1.5 percentage points of annual non-agricultural GDP growth is endogenously
induced by growth in staple sector, as there is no additional exogenous growth shock imposed on the non-
agricultural sector in this scenario. On the other hand, if additional economy wide growth is triggered by
additional growth in the agricultural export sector, at the same level of GDP growth, the annual growth rate
in the non-agricultural sector rises only to 5.4% (Table 18, third row, first and third column), with
additional 0.1 percentage points of annual growth compared with the base-run.
16. See Dorosh and Haggblade (2003) for a comparison of CGE and fixed-price multipliers for several
Sub-Saharan African countries.
41
Agricultural growth stimulates non-agricultural growth
104. Two more scenarios are run to further analyze the linkages between agricultural and non-
agricultural growth. In Scenario 3, it is assumed that an additional productivity growth shock occur within
the agricultural sector only, while in Scenario 4, a similar productivity growth shock is imposed on the
non-agricultural sector only. Again, the target is a 5.5% annual growth rate of GDP in both scenarios. To
generate such growth in total GDP (by taking into account the linkage effects), grain production grows at
4.5%, with 1.5 percentage points of additional annual growth from the base run, and livestock grows at
4.7%, additional 1.1 percentage points of annual growth. This results in a staple sector growth of 4.3%, an
additional 1.2 percentage points of annual growth. A relatively higher growth rate (7.7%) is needed for the
export agriculture with 4.5% of additional annual growth. This results in total agricultural output (not
AgGDP) to grow at 4.7% per year, instead of the 3.1% in the base run, and thus leading to a 1.6% of
additional annual growth.
Figure 6. GDP growth multipliers in agriculture-led and non-agricultural-led growth scenarios
Both with 5.5% GDP annual growth
0.6
0.7
0.8
0.9
1.0
1.1
1.2
1.3
1.4
2004 2006 2008 2010 2012 2014
One-unit agricultural outputOne-unit nonagricultural output
Source: Authors calculation from the EMM model results.
105. Growth in agriculture significantly stimulates the non-agricultural sector‘s growth. Measured by
total non-agricultural GDP, the annual growth rises to 6.4%, instead of 5.3% in the base-run (Table 18,
third row, first and fourth columns). The additional 1.2 percentage points of non-agricultural annual growth
are thus induced by the growth in agriculture. Calculated GDP growth multipliers are 1.05 – 1.13 in the
first five years and increase to 1.29 by 2015 in this scenario (Figure 6). That is to say, a one unit increase in
total agricultural output (measured at the base-year‘s prices) can generate more than one unit of total GDP
and its impact will reach 1.29 units of GDP by 2015.
106. In Scenario 4 the initial exogenous increase in productivity is assumed to happen in the non-
agricultural sector only. The non-agricultural sector grows by 5.3% in the base run. In order to induce total
GDP to grow at 5.5% annually, the sector‘s annual growth rises to 6.8%, with 1.5 percentage points of
additional annual growth. However, the non-agricultural sector can stimulate the growth in the agricultural
sector only slightly. Agricultural GDP's growth rate rises to 4.2%, instead of 3.70% in the base run
(Table 18, second row, first and last columns). The calculated GDP growth multipliers in this scenario are
less than one, 0.72 in 2004 and 0.68 in 2015. In other words, a one unit increase in non-agricultural real
output (not non-agricultural GDP) can only generate a less than one unit increase in total GDP (see
Figure 6 for the growth multiplier comparison between the two scenarios).
42
Demand linkages are important for non-agricultural growth
107. Strong demand linkages explain why faster agricultural growth, especially of staple production,
has such strong growth linkage effect on the non-agricultural sector. Empirical studies show that demand
linkages are the dominant factor for agriculture to stimulate non-agricultural growth in low-income
countries. For low-income countries, a majority of the population lives in the rural areas and mainly
depends on agriculture for its livelihood. This has two implications for the overall growth in the economy.
First, it implies that it is unlikely to have any broad-based growth without agriculture. Growth in
agriculture, especially in those sub-sectors in which a majority of farmers is engaged, can become an
engine strong enough to generate overall economic growth. This is the direct effect of agricultural growth
on the economy-wide growth. Second, it implies that rural demand is the dominant factor to provide
enough market opportunities for both industrial products and services produced in the economy. Increased
farm income has to be spent on both agricultural and non-agricultural commodities, including services.
Income elasticity of demand for the non-agricultural commodities and services is often greater than that for
the agricultural goods, which implies that farmer will spend more of the additional income on non-
agricultural consumption. This is the indirect effect – through demand linkages – of agriculture on the
overall economic growth.
108. Importance of the demand linkages has been doubted recently because changing local and global
conditions as a result of globalization. For some scholars (reference) an African country's demand for
agricultural good can now be met by imports and hence the importance of agriculture-induced demand
linkages in the economy seems to be much smaller in such environment. While agricultural trade has
grown more rapidly in the recent years due to market liberalization and globalization, few developing
countries, including those with much higher per capita income levels than Ethiopia, can primarily depend
on imports to meet food demand of domestic consumers. Even it is possible, it is often for the case in
which size of population is small and the countries that has been urbanized entirely. It is unlikely for a low-
income country like Ethiopia, which is poor in natural resource, underdeveloped in manufacturing, size of
total population is more than 70 million, and a majority of the population lives in the rural areas, to look
for a food-import development path.
109. With more than 85% of population living in the rural areas, it is obvious that rural demand is the
major component of total demand in the Ethiopian domestic market. Even though per capita rural income
is 40% lower than the urban income on average and each rural individual‘s non-agricultural consumption is
60% below the level for an average urban consumer. The household survey data shows that, because of the
dominant size of rural population in the country, more than 70% of non-agricultural products and services
are actually consumed by rural households.17
As many poor rural households can barely have additional
cash income left after meeting basic food consumption, increased income, especially cash income, will be
spent disproportionately more on non-agricultural consumption. The estimated marginal propensity to
consume shows that for each one Birr of additional income, rural household would spend 0.6 Birr on non-
agricultural consumption, though the average budget share of non-agricultural consumption is about 30%.
Moreover, the marginal budget share (MBS) of non-agricultural spending is higher among the low income
groups than that for the higher income groups among the rural population. For example, the MBS for the
non-agricultural products and services is as high as 0.74 for the lowest income quintile in the rural areas.
This share declines as income increases, and it is 0.47 for the highest income quintile (Wamisho and Yu,
2006). When a majority of population lives in the rural areas, demand for the non-agricultural goods and
services, induced by the agricultural income growth, is primarily met at the local market and supply is
provided by domestic firms and local enterprises. Demand linkages, hence, will continue to be the
important force to stimulate growth in the non-agricultural sector and rural households will provide steady
growing market opportunities for such growth.
17. Calculated from HICES for the data of 1999/2000.
43
Agricultural growth is pro-poor
110. It is widely anticipated that Ethiopia is unlikely to meet the first MDG of halving poverty by
2015 unless the country‘s growth performance improves dramatically. Ethiopia needs to not only
accelerate the level of growth, but also find ways to enhance the ‗pro-poorness‘ of growth. In other words,
identify the kind or composition of growth that is most effective at reducing poverty. In this regard, it is
necessary to consider the relative importance of agriculture and industry in helping the country achieve its
development objective of significantly reducing poverty.
111. The EMM model is thus used to examine how differences in the pattern of growth in the country
influence the rate of poverty reduction. More specifically, using the same four scenarios discussed above,
the growth-poverty linkages is also analyzed by comparing the effectiveness of additional growth in
reducing poverty under different growth assumptions. To make the results comparable, poverty-growth
elasticities are calculated for each of the four scenarios and growth in total GDP, rather than the sector's
GDP, is considered in the calculation.18
Table 19 shows that the poverty-growth elasticity is larger when
economy wide growth is driven by agriculture rather than non-agriculture. A one per cent annual increase
in per capita GDP driven by agriculture-led growth leads to 1.9% reduction in the poverty headcount rate
per year. By contrast, a similar increase in per capita GDP driven by non-agriculture leads to only 1.1% fall
in the poverty rate. These disparities in poverty-growth elasticities can translate into different reductions in
the poverty headcount over time. For example, with 5.5% of GDP growth in both scenarios, the poverty
headcount falls to 25% in the agriculture-led growth scenario, compared with 28% in the non-agriculture-
led growth scenario (Figure 7). Given its larger impact on poverty, agricultural-led growth in Ethiopia lifts
an additional 2.9 million rural people out of poverty compared to non-agricultural-led growth. Though in
the non-agriculture-led growth urban poverty population reduces more than that in the agriculture-led
growth, in total, there will be additional 2.6 million people lifting out of poverty in the agriculture-led
growth compared to non-agriculture-led growth, despite the fact that overall GDP grows at the similar rate
under the two scenarios.
18. The poverty-growth elasticity used in this study measures the responsiveness of the poverty rate to changes
in the per capita GDP growth rate. The formula for this elasticity is:
P P P GDPpc0 0 0
GDPpc GDPpc GDPpc P0
where P0 and
GDPpcare average annual changes (from the base-year) in the poverty headcount rate and
level of per capita GDP; and P0 and
GDPpcare the base-year poverty headcount rate and per capita GDP.
The poverty-growth elasticity measures the percentage change in the poverty headcount rate caused by a
one-percent increase in per capita GDP. This is not equivalent to a percentage point change in the poverty
headcount rate.
44
Table 19. Agricultural growth is more pro-poor
Baseline scenario Agriculture-led
growth scenario
Non-agriculture-led
growth scenario
Annual per capita GDP growth rate (%) 1.9 2.9 2.9
Annual non-agricultural production growth rate
(%) 5.9 6.1 7.7
Annual agricultural production growth rate (%) 3.1 4.7 3.2
Staples (staple crops and livestock) 3.1 4.3 3.2
Export crops 3.1 7.7 3.0
Poverty headcount by 2015 (%) 32.1 25.1 28.0
Difference in poor population in 2015 (1 000) -6,332 -3,682
Poverty-growth elasticity - -1.9 -1.2 Source: Authors calculation from the EMM model results.
Figure 7. National poverty rate (%) in agriculture-led and non-agricultural-led growth scenarios
Both with 5.5% GDP annual growth
24
26
28
30
32
34
36
38
40
42
44
2003 2005 2007 2009 2011 2013 2015
Ag Nonag
Source: Authors calculation from the EMM model results.
112. In summary, the findings of the EMM model regarding to the strong linkages effect of
agricultural growth is consistent with the findings of the SIO model for Ethiopia. However, the SIO model
has reached such conclusion by evaluating individual crops or subsectors one by one, while the EMM
model is more flexible and can be used to evaluate the linkages effect for a group of crops or broadly
defined agriculture. Moreover, the SIO model focuses on the one-time gains of GDP or household income,
the EMM model pays more attention to the growth rate effect of linkages. By doing so, the EMM model is
also able to assess the change in such linkages effect overtime and it finds that linkages effect becomes
stronger over time when growth is induced by the staple production's productivity but weaker when growth
is induced by the export agriculture.
113. The most important additional contributions of the EMM model analysis is the assessment of
linkages between a sector-led growth and poverty reduction. The effectiveness of agriculture-led growth in
poverty reduction further emphasizes the importance of agriculture to Ethiopia's development. While both
findings do not come out in surprising given that Ethiopia is one of the largest low-income countries in the
world and constrained by both resource poor and land-locked, it provides analytic evidence to supports the
strategy of 'Agricultural Development Led Industrialization' pursued by the Ethiopian government.
45
Different growth options at the regional level
114. In addition to the national-level analysis discussed above, the model allows for assessment of the
differential effects of the simulated growth options on poverty reduction across regions. For example,
growth in staple crops causes the rural poverty rate to fall in the food surplus area from 39% to 25.7%,
while in the food deficit area it only drops 4 percentage points, from 60.5% to 56.6% over the simulated
timeframe. While these results clearly show that staple crop growth is a strong driver of overall poverty
reduction, it will not be sufficient to redress poverty in the food deficit area given that some of the food
deficit areas do not have enough agricultural potential. Growth in other agricultural subsectors displays a
similar differential effect on rural poverty reduction at the sub national level (Figure 8).
Figure 8. Comparison of effect of agricultural subsector growth on poverty reduction in the food deficit and food surplus areas
a. Food deficit area
56
57
58
59
60
61
62
63
64
65
2003 04 05 06 07 08 09 10 11 12 13 14 2015
%
Business-as-usual scenario
Staple crop grow th scenario
Livestock grow th scenario
Nontraditional export grow th scenario
Coffee grow th scenario
b. Food surplus area
25
27
29
31
33
35
37
39
2003 04 05 06 07 08 09 10 11 12 13 14 2015
%
Business-as-usual scenario
Staple crop grow th scenario
Livestock grow th scenario
Nontraditional export grow th scenario
Coffee grow th scenario
Source: Authors’ model simulation results for Ethiopia.
46
115. The above analysis indicates the necessity for differential growth strategies across regions. A
balanced agricultural growth strategy appears necessary for improving food security and rural income in
the food deficit area, while growth in staple crops, especially cereals, will be the dominant driver in the
food surplus area. Increased cereal surplus, however, needs to be diverted to meet demand beyond the food
surplus area, making market and infrastructure development crucial, along with additional conditions to
reduce farmers‘ post-harvest risk (which, although not simulated in the model, is an extremely important
factor in growth and poverty reduction). In the absence of these preconditions, staple crop production
growth in the food surplus area would likely depress market prices, ultimately hurting rather than helping
farmers.
5. Achieving agricultural growth
116. The important role of agricultural growth in overall economic growth and poverty reduction
discussed in the preceding sections of this report will only be feasible with significant investments in staple
crops and livestock productivity. Hence it is important to assess the nature and extent of such investment.
Irrigation
117. Irrigation is naturally a critical component in reducing climate risk and improving crop
production. Reducing climate risk can also help to induce the use of modern inputs, such as fertilizers and
improved seeds, thereby further enhancing agricultural productivity. Irrigated area in Ethiopia totalled
about 200 000 hectares—slightly more than 2% of the total crop area. Of that irrigated area, 60% is planted
to cereal crops and 40% to other (mainly cash) crops. According to the agricultural sample surveys, the
yield gap between irrigated and rainfed crop production is 40%, meaning that, on average, irrigation has
the potential to increase cereal yields by up to 40%. Obviously, significantly increasing irrigation area
would stimulate cereal production, but given that only 2% of cereal production and slightly more than 2%
of other crop production is irrigated, it is unrealistic to expect that irrigation investment alone could
generate the levels of cereals growth modelled in the previous section. Moreover, many researchers (for
example, Fan and Hazell 2001) have shown strong diminishing returns to large-scale irrigation investment,
implying that caution is needed in promoting large irrigation projects.
118. An irrigated area growth scenario was formulated based on Ethiopia‘s Irrigation Development
Program, which is quite a moderate plan involving the development of about 274,000 hectares of
additional irrigated area by 2015, 50% of which will be allocated to cereal crop production. Simulation
results indicate that this level of expanded area will only increase irrigated cereal production to 3% to 5%
of total cereal production in 2015, representing minimal additional annual growth. It should be noted,
however, that given the medium- to long-term nature of the program (meaning that projects are only
completed toward the end of the simulation period), the potential returns are not fully captured within the
simulation timeframe.
119. In terms of cash crops, irrigated area under this scenario triples by 2015 and hence accounts for
5% of all cash crop area compared with 2% as of today. This in turn increases exports; horticultural
exports, for example, increase four-fold by 2015 over baseline levels, and coffee exports increase by about
the same amount. As already discussed, however, such productivity increases will only reach domestic and
international markets given improved infrastructure and market conditions. Consequently, the gains
projected under the irrigated area growth scenario should not be understood to result solely from irrigation
investment. Concurrent investments in markets and transportation are needed.
47
Adoption of improved seed
120. The low yields prevalent in Ethiopian agriculture are generally attributed to low usage and
efficiency of modern inputs. As discussed in Section 3 of this report, national survey data show that, while
about 40% of cereal production benefits from the use of fertilizer, only about 10% also gains from other
inputs, such as improved seed or irrigation. The average yield gap in cereal production due purely to lack
of fertilizer is actually quite small. Total cereal yields where fertilizer is used are about 1.4 metric tonnes
per hectare, 20% higher than yields without the use of any modern inputs. Many studies report similar
findings regarding fertilizer use. For example, based on a household- and plot-level survey conducted in
100 villages in the Tigary region, Pender and Gebremedhin (2004) find that fertilizer use is associated with
yield increases of 14% (with a weak statistical significance). Using the Ethiopia Rural Household Survey
(ERHS) for 1994, Croppenstedt and Demeke (1997) report fertilizer elasticities in the range of 0.03 to 0.09
in the production function. Yao (1996) reports elasticities in the range of 0.05 to 0.10, based on aggregated
time-series data.
121. There are many reasons for this disappointing outcome. Abrar, Morrissey, and Rayner (2004), for
example, find that average fertilizer application in Ethiopia falls within the low range of 10–50 kilograms
per hectare—considerably lower than the recommended rate of 150–200 kilograms. Pender and
Gebremedhin (2004) emphasize the complementary effect of fertilizer use with soil and water conservation
investment and land management. Both irrigation and stone terrace technology are associated with
increased fertilizer and other modern input use, and their joint effect on land productivity is significant in
Tigray. Farming practices also affect fertilizer efficiency. Howard et al. (2003) find ploughing four or
more times before planting can increase yields by 550 kilograms per hectare. Later planting reduces yields
by 280–315 kilograms per hectare, and failure to weed on time results in average losses of about
220 kilograms per hectare.
122. Lack of agricultural extension services may result in a knowledge gap for farmers when it comes
to adopting modern technologies, including fertilizer, properly. Ayele, Kelemework, and Alemu (2003)
report that even though the number of agricultural extension agents has significantly increased in Ethiopia
in recent decades, the national ratio of staff to farm households was still only 1:700. A high degree of
inefficiency of fertilizer use among cereal farmers was found by Croppenstedt and Mulat (1997). They
estimated mean efficiencies at 40% for fertilizer compared with 76% for land and 55% and labour. Badly
timed application may also contribute to low fertilizer use efficiency. This is partially due to the inability
of farmers to acquire fertilizer and fertilizer credit when needed. Production and price risk and resource
availability are all found to affect farmers‘ decisions regarding both fertilizer use and its proper
application. High price, output, and hence profit variability make investment in inputs risky for farmers
(Snapp, Blackie, and Donovan 2003.) Van den Broeck (2001) finds weather risk to be associated with
fertilizer use. In the case of good weather, fertilizer use can result in a 29% higher output value compared
with non-use; however, in the case of bad weather, it can lead to 30% lower output values.
123. If fertilizer is used with improved seed in cereal production, Agricultural Census data show that
average yields increase to 2.5 metric tonnes per hectare, doubling the level achieved without modern
inputs. This outcome is consistent with the findings of Howard et al. (2003). Based on a maize plot survey
in the Oromiya region, average maize yields were 70% higher when improved seed and fertilizer were used
compared with traditional seed and no fertilizer, and there is still a 40% potential for further improvement
based on results from research stations. The econometric analysis conducted by the authors also supported
their findings.
124. While significant gains in cereal production are possible from a combination of fertilizer use and
improved seed, survey data show that only about 4% of cereal area has been grown employing such
technologies. Some studies associate the low adoption of improved seed with the quality and price of seed,
48
which may result from lack of competition in both seed production and distribution (Crawford et al. 2003).
Further, adopting any modern technology often requires changes in crop or land management, and, once
again, in the absence of education, training, and extension services, farmers understandably find it difficult
to move beyond longstanding traditional farming practices. Learning new skills and monitoring input and
output prices are integral to modern technology adoption (Weir 1999).
Promoting modern technology in livestock production
125. Ethiopia has the largest livestock sector in East Africa, with a stock of 42 million cattle and
46 million sheep and goats. More than 60% of the cattle are raised in the highland area, following a typical
mixed crop–livestock system, and 60% of the sheep and goats are raised in the lowlands, which are
dominated by pastoral systems. The livestock sector plays multiple roles in the country‘s rural economy.
Live animals, especially cattle, are the most important source of cash income for many farmers; large
animals are the dominant asset; draught animals are virtually the only capital input in crop production for
most small farmers; and milk is one of the main sources of protein in the diet, especially for children.
126. Traditional technology plays a dominant role in livestock production. Except in Addis Ababa, the
number of hybrid and exotic cows is extremely low and grazing and crop residues are often the only
sources of animal feed. Because of the low use of modern technologies and inputs, livestock productivity is
extremely low. Yields from milking cows, for example, are among the lowest in East Africa. The average
yield in Ethiopia per cow is about 270 litres per year compared with 500 litres in Kenya, 480 litres in
Sudan, 400 litres in Somalia, and 350 litres in Uganda (Muriuki and Thorpe 2001). Once modern
technology is adopted, livestock productivity is significantly improved. In Addis Ababa, for example,
almost 50% of milking cows are of cross-bred and exotic varieties, while for the country as a whole the
ratio is less than 2%. Given the comparatively high ratio of modern technology adoption in Addis Ababa,
together with modern input use and favourable market conditions, yields from milking cow are two to three
times higher than the national average.
127. If the cross-bred milking cows were increased in line the existing 20% share in Kenya, the
scenario simulation of EMM model shows that it represents more than 10-fold growth. Achieving this
means an additional 4.5% annual growth in milk production. According to Fernandez-Rivera, Okike, and
Ehui (2001), the potential for increasing beef yields is significantly lower than the potential for increasing
milk yields. 40% of cattle in Ethiopia were draught animals — the most important source of beef — which
in part explains the low efficiency of beef production. Most draught animals can be kept 10 years or more
as working animals, and meat production is just a by-product. Because this, technology adoption rate is
lower for beef production than for milk production. Thus, similar growth through the adoption of modern
technologies in beef production as was assumed for milk production (approximately 20% per year), the
resulting overall growth in beef production in the model simulation is much lower (only about 0.5%).
128. A scenario of combination of milk, beef, and poultry production growth results in an additional
3.8% overall annual growth in livestock products. Milk is the dominant contributor to this result, while
beef and poultry play only marginal roles. This result implies that reasonably high growth in the Ethiopian
livestock sector is feasible by increasing the adoption of modern technology to 30% of total production
(compared with the 2001/02 level of only 10%). Furthermore, the model simulation shows that this
magnitude of livestock sector growth has the potential to induce additional 0.6% GDP growth and 0.8%
AgGDP growth per year over the projection period. However, the model simulation also show that there
seems to have comparatively weak linkages between livestock growth and poverty reduction, as increased
use of modern livestock technologies usually occurs in areas where such technologies are already in use—
generally areas where the poverty rate is below the national average. Modern livestock technologies are
rarely known of or applied by farmers in areas where poverty is particularly high. Thus, modern
technology adoption may not initially benefit the poorest people—which is consistent with the findings of
49
Hazell and Ramasamy (1991) for the early stages of the Green Revolution in India; specific targeting
policies that encompass increased education and extension, as previously discussed, will also be needed.
Halving the poverty: markets and non-agriculture matter
129. An agriculture-led growth strategy does not imply that investments should be in agriculture only.
Many studies have shown that poor infrastructure and dysfunctional markets prevent farmer access thereby
diminishing the profitability of agriculture (Kelly et al. 2003). It is important to remember that institutional
barriers also constrain farmers from becoming actively involved in market activities, and market
development does not solely imply infrastructure investment (Gabre-Madhin 2001). Nonetheless, this sub-
section focuses specifically on the growth and poverty effect of reducing transportation costs associated
with agricultural trade and improving market access for farmers.
130. The burden of geography - a combination of isolation, topography, and lack of past investments -
has left Ethiopia farmers especially poorly connected to roads, transport and other forms of
communications, even relative to other African countries. Ethiopian road density is 27 kilometres per
1 000 km2, slightly more than half the 50 kilometres per 1 000 km
2 average for Africa as a whole. Seventy
per cent of Ethiopian farmers are reportedly more than half a day‘s walk away from an all-weather road,
and farm households are on average 10 km from the nearest road, 18 km from the nearest public transport,
and 6 km from the nearest food markets. The combination of this poor market access and high
transportation costs significantly increases the gap between consumer and producer prices, which
ultimately lowers the farm-gate prices received by affected farmers. The average grain price gap is
estimated to be about 30% to 70% across regions, and domestic marketing costs can account for more than
50% of fertilizer prices paid by farmers (Jayne et al. 2003). The isolation raises the costs and risks of
market exchange, rendering commercialization strategies unattractive for many households who instead
opt to produce their own food. Better access to food and other markets would improve incentive to produce
more profitable cash crops and livestock, and raise farm incomes. Opportunities for economies of scale are
missed where markets are thinner or where trade flows are more sporadic, and trading networks suffer
from weaknesses in financing, information and the inability to assume risk.
131. Major improvements in the road network and market infrastructure are necessary to shit
agriculture to a more commercial footing, increase productivity, and spur diversification. Thus, we apply
the EMM model to simulate the decreased market costs resulting from increased investment in roads and
other market infrastructure. Constrained by available information on the quantitative relationship between
market costs and investment in such infrastructure in Ethiopia, two main assumptions were made:
(a) investment lowers the marketing margins between the food surplus and food deficit areas, and
(b) improved infrastructure will reduce the price gap between the food surplus and food deficit areas by
10% per year, such that market prices across zones will converge by 2015 (representing an overall decrease
in the price gap of 70%). It is further assumed that lower marketing costs are associated with improved
service sector productivity, and by 2015 such productivity increases by 15% over baseline levels (a 1%
increase per year).
132. Once growth in the agricultural sector is combined with improved marketing margins through
cross-sector linkage effects, there are 2.7% of additional GDP growth and 2.3% of additional AgGDP
growth per year than the base-run. That is to say, reducing marketing margins by 10% per year between
food deficit and food surplus areas would increase overall annual GDP growth by 0.7 percentage points.
Reducing marketing costs primarily benefits smallholders via the increased prices they receive for their
goods, increasing their income from the same level of output. Moreover, improving market conditions
creates a more efficient trading sector (as part of the service sector), which itself can generate greater non-
agricultural income at constant costs. Due to such cross-sector linkages and positive price effects, the
poverty rate under this scenario is significantly lowered, drawing the objective of halving poverty rate by
50
2015 within reach. Reducing marketing margins by 10% would reduce poverty by an additional
3.1 percentage points by 2015. Moreover, the pro-poor effect of the resulting growth is much stronger in
rural areas, where simulation results indicate the poverty rate drops to 25% by 2015 from the 2003 level of
45.8%.
133. While market improvement supports agricultural growth and generates additional non-
agricultural growth (though mainly in trade-related services), broad non-agricultural growth, including
manufacturing and other services, is also critical in meeting MDGs. Nonagricultural growth not only
creates nonfarm opportunities and rural income but also increases urban income; further, rural nonfarm
income creates market demand for agriculture. Cross-sector linkage effects induce more non-agricultural
growth over and above the agricultural growth discussed above.
6. Conclusions
134. Ethiopia has embarked upon an agriculture-based growth strategy to meet the challenges of
accelerating overall growth and poverty reduction. Three sets of key questions around the importance of
the agricultural sector were subsequently highlighted as relevant to the implementation of such a strategy.
These were: which sectors have large prospective linkages; what are the growth and poverty-reduction
potential of these sectors and constraints thereof; and what policy interventions are capable of unlocking
the growth potential.
135. We have applied two complementary approaches to address above questions and assess the
importance of the agricultural sector in Ethiopia. While to provide answers to such questions are important
for evidence-based agricultural strategy making, they are not enough for identifying binding constraints in
the process and hence to guide the strategy to be more targeted and more practical. From this regards, the
following factors that have not been discussed in depth need to be addressed before we conclude the report.
136. First, while the linkage analysis focuses on the interaction between sectors in the Ethiopian
economy and hence the linkages effect of a sector-led growth, and shows that agricultural growth has
stronger linkages or multiplier effect on the entire economy than the non-agricultural growth, lack of
incentives for smallholders to adopt modern agricultural technology and to participate in the productivity-
led growth may constrain such linkages effect to occur. Moreover, factors that affect smallholders'
incentives and constrain agricultural growth may not be necessarily within this sector. While we only
mentioned road condition that significantly increase the difficulty of market access for smallholders, many
other factors, including the land tenure system, doing-business conditions, are also important. Thus,
modern agricultural technology such as high yield seed varieties, fertilizer and other inputs suitable for
different locations and possible to be adopted or adapted by smallholders depends on actions often outside
agriculture.
137. Second, the multiplier analysis should not be misinterpreted as a single sector growth focus.
While the report displays that multipliers are useful measure to assess the linkages between a particular
production sector with the rest of the economy, such analysis does not imply that growth in any sector
alone can bring the overall economic growth to a level that can have significant welfare improvement to a
majority of population in this country. Rather, the existence of strong linkages within the economy
indicates the required growth in other sectors when a high multiplier is measured for a particular sector's
growth. Given that agricultural growth has stronger multiplier effect in Ethiopia's current economy, to
maximize the impact of policy reforms and public investment on the growth, linkages between the targeted
sectors (which can be those outside agriculture) and the agricultural sector have to be taken into
consideration. In certain situation multiplier effect occurs automatically through such linkages when right
policy is implemented or key bottleneck is removed by public investment. However, winner-picking
intervention for supporting any single sector at the cost of the growth in the other sectors has been proved
51
to be inefficient and often distorted the incentive system of the economy. Under such situation, linkages
effect are unlikely to occur.
138. Third, the report does not fully capture the series of dynamic factors that may rapid change the
economic structure when growth is accelerated. The linkages among different sectors in an economy and
multiplier effect of a sector on the rest of the economy are measured in a general equilibrium set up in this
report and a social accounting matrix (SAM) is the data source to conduct such analysis. While the
dataset (the SAM) captures the current economic structure of Ethiopia, it cannot predict the change in the
economic structure, given that the static dataset is used. Although the SAM used for this study has been
constructed with highly disaggregated sector structure and was based on the most recent economic data
available to the authors (the SAM represents the country in 2007), some small sectors (particular in the
manufacturing and non-traditional agricultural export sector such as cut flowers) and their input-output
relationship with the other sectors cannot be captured. In most cases, such small sectors have been
aggregated into the so-called 'other manufacturing sector' in the case of manufacturing or 'other crops' in
the case of cut flowers. If some of such small sectors are highly dynamic and will become important
economic activities soon in Ethiopia, the current study definitely cannot capture their potential role in the
economy in the future. Development is a process that can produce uncertain outcomes and is full of
surprises. The linkages captured in this analysis, thus, are limited by our current understanding of the
economic structure and therefore should not be seen as the only possibility to induce growth in the future.
139. Finally, the institutional structure of the economy is ignored in the analysis. Similar as in any
other general equilibrium model, perfect competitive markets for both products and inputs are assumed.
Linkages occur under an implicit assumption in which there is no any market imperfect and other
institutional barriers to constrain the private sector including farmers in response to the increased demand
to their products. If supply response is constrained by such barriers, linkages effect may not occur or
occurs with much more modest extent than what the model tells. While in the SIO model, it is possible to
control the supply response by assuming an inelastic supply function for a sector, it is impossible for the
model to identify the reasons to cause such inelasticity in supply. While understanding institutional barriers
of growth is an important component of understanding economic linkages, it is, however, beyond the scope
of this study.
140. Keeping these limitations of the report in mind, the following key messages can be drawn from
its analysis.
141. While two complementary approaches are applied in the report to assess agricultural growth
linkages in Ethiopia, both approaches reveal the same pattern of linkages, albeit with differences in the
magnitudes involved. Estimates based on a fixed-price semi-input-output (SIO) model indicate that large
growth linkages are generated by most of the sectors examined. Nevertheless, those induced by agricultural
activities are larger. This difference arises primarily because of much higher value-added created by the
direct increase in agricultural outputs. As a fraction of the gross value of output, value-added has high
share in agriculture. The flexible-price economy-wide multimarket (EMM) model, which also accounts for
spatial differences, produces comparable, albeit smaller, estimates for staples and agricultural exportables.
142. The EMM model results reveal that the impact of growth on poverty is larger when the economy
wide growth is driven by agriculture rather than non-agriculture. Given its larger impact on poverty,
agricultural-led growth in Ethiopia lifts more rural people out of poverty compared to non-agricultural-led
growth. Non-agriculture-led growth reduces urban poverty more than agriculture-led growth. In total,
however, there will be a larger number of people coming out of poverty in the agriculture-led growth
compared to non-agriculture-led growth, despite the fact that overall GDP grows at the same rate in both
scenarios.
52
143. Estimates also imply that growth in staple production will generate more than proportionate
increase in total GDP. Moreover, such growth linkages become stronger over time. On the other hand, the
linkages from agricultural exports to total GDP is strong only in the initial five years, the linkages become
weaker overtime, and the growth multipliers fall below one by 2015.
144. In light of the usual focus on exports, it is useful to briefly comment on why growth in staple
crops has such a significant effect. Cereals and other staple crops are the most important income source for
the majority of small farmers. Domestic supply of staple crops is the most important source of food energy
for both rural and urban poor consumers. Both of these features are likely to continue to apply for the next
10 years or so (the horizon being examined). Thus, raising productivity in staple crops will increase the
food supply, lower food prices, and help reduce the poverty rate in both rural and urban areas. Clearly,
better incentives and improved production conditions will give farmers more opportunities to diversify. As
a consequence, many presently subsistence crops grown extensively by poor farmers can become
marketable commodities and this shift would further increase poor farmers‘ cash income.
145. The main elements of the Ethiopia's government's approach to improving market connectivity
have focused on the crucial areas of liberalizing markets and improving roads, especially major highways.
Liberalization has clearly improved the functioning of grain markets after the interventionist policies of the
Derg, particularly in surplus grain producing regions (World Bank, 2006). The government has recognized
that a dramatically improved road network is a prerequisite for higher rate of growth in agriculture and to
foster urban-rural linkages. However, despite a significant effort over the past decade on road construction,
transport costs for farmers and for traders remain high, in part because most farmers lack feeder roads to
the main networks. And roads are a long term and costly investment that do not improve yield and returns
quickly. The model results show that growth in the agricultural sector is indeed constrained by the service
sector such as trade, transport, financial and other business services. Improvement in these market-related
services straightens cross-sector linkages and allows more gains in total GDP. Reducing such marketing
costs primarily benefits smallholders via the increased net prices they receive for their goods, thereby
raising their income from the same level of output. Improving market conditions also creates a more
efficient trading sector as well as other service sectors, which itself can generate greater non-agricultural
income without increasing costs.
146. But it is not just roads and topography - supportive policy measures are also needed to get
markets 'right'. Institution-building and attention to risk, information, and food aid-related distortions are
also critical (Gabre-Madhin 2001). While the government's recent emerging market strategy has built a
central commodity exchange to address problems of market information and transaction costs and risks,
and to exploit scale economies through cooperatives (Gabre-Madhin and Goggin 2006), the outcome has
so far not been as expected because other policy and institutional barriers that have constrained market
participation of the private sector. Many policy measures in other sectors, such as expansion of mobile
phones to rural villages or to subsidize the purchase of radios, could also have substantial impacts on the
market's responsiveness and connectivity. At the same time, innovative approaches are needed to help
farmers meet the new challenges of participating in global markets - particularly with regard to quality and
safety standards.
147. While market improvement supports both agricultural and non-agricultural growth, non-
agricultural growth, including manufacturing and other industries, is also critical. Non-agricultural growth
not only creates non-farm opportunities and rural income but also increases urban income; further, rural
non-farm income creates market demand for agriculture. Cross-sector linkage effects induce additional
non-agricultural growth over and above that generated by the agricultural growth and market
improvements discussed above. As a consequence, GDP grows faster and poverty declines more rapidly.
Competition is essential for developing the non-agricultural sector and the development of non-agricultural
53
sector requires greater participation by the private sector. A dynamic private sector is the key and its
existence needs policy supports and other institutional environments.
148. In summary, the key findings of this report that are highly relevant to the three policy questions:
Agricultural growth induces higher overall growth than non-agricultural growth. It also leads to
faster poverty reduction since it generates proportionately more income for farm households who
cover the majority of population in the current economy and represent the bulk of the poor. From
within agriculture, staple crops have stronger growth linkages.
Consumption linkages are much stronger than production linkages in the agricultural sector. In
most cases, the impact of increased consumption demand due to growth in agriculture is much
larger than that of the corresponding expansion in input demand.
Non-agricultural growth cannot be neglected, however. Such growth can, in its own right, have
large growth effects within the non-agricultural sector given that such growth creates more
production linkages than the growth led by the agricultural sector. More importantly, non-
agricultural sectors have to grow in order to match growing supply of agricultural products and
increasing demand for non-agricultural products. Otherwise, falling prices of agricultural
products may dampen the realized gains in growth and poverty reduction. Given the rather small
industrial sector, import-substitution investments in the relevant sectors appear necessary to
achieve success.
149. The key message is, therefore, that exploiting the potential growth linkages towards poverty
reduction and structural transformation require a diversified (or ‗balanced‘) growth strategy that
encompasses agricultural staples and exportables as well as non-agricultural sectors. On the one hand, the
explorations of this paper imply that the emphasis of ADLI and PASDEP on agricultural growth is, in
principle, warranted. On the other hand, the results also clearly show that exclusive focus on agriculture (or
insufficient attention to non-agriculture) is counter-productive. It would at best lead to unsatisfactory
outcomes in growth and poverty reduction. The greater comprehensiveness of PASDEP suggests that
policy-makers may have learnt that lesson.
54
APPENDIX A
A1. An illustration of the fixed price input-output models
We assume a small open economy in which there are two production sectors, each of which produces
a single good (Zi), and there is a single consumer who consumes both goods (C1 and C2). Both goods are
tradable and Ei represents net exports (imports) of good i if it is positive (negative).
(1)
(2)
where on the left side of Equations (1) and (2) Zi represents total supply of good i. The first two
components of the right side of the equations capture the intermediate demand in which aij is the fixed
input-output coefficient; the third component is for consumer's final demand for good i, which, in turns is
the composition of the two goods (to be employed in the production of good i) with fixed coefficient vi;
and the final component of the equations is the export demand (import supply) of good i, if Ei is positive
(negative).
Rearranging equations (1) and (2) and presenting them in a matrix format, we obtain the following
equation:
= (3)
In equation (3), the coefficient matrix is the I-M matrix, i.e.
= .
Let = Z and = E, the following relationship between Z and E holds:
(4)
is the inverse matrix of .
Since supply is assumed not to be constrained by inputs, sector's output and income are determined by
the level of exogenous demand (E) through the matrix in which M is called the multipliers.
55
A2. The fixed price, semi-input-output (SIO) models
To better simulate real-world supply rigidities, semi-input-output (SIO) models classify sectors into
two groups, those that are perfectly elastic in supply (e.g. Z1) and those with supply-constrained (e.g. Z2).19
As described in equations (3) and (4) above, the SIO model permits output responses only in the supply
responsive sectors (Z1). Perfectly elastic supply ensures fixed prices for these (Z1) goods. In the other
group, of supply-constrained products (Z2), perfect substitutability between domestic goods and imports
guarantees that world prices will ensure fixed prices for these goods as well. For these models to produce a
reasonable approximation of reality, the supply-constrained sectors often correspond to tradable goods
whose domestic supply remains fixed at the prevailing output price. In these supply-constrained sectors
(Z2), increases in domestic demand merely reduce net exports (E2), which then become endogenous to the
system and determined by the matrix of semi-input-output multipliers (M*). Specifically, Equation (3) has
to be rearranged as in the following form:
= , (3')
where is good 2 fixed in supply. Let on the left side of Equation (3') be
and and the right side of Equation (3') be B, Equation (3') is further
presented as
= B , such that
B . (4')
A3. The Ethiopia economy-wide multimarket (EMM) model
Supply functions of the EMM model
The structure of the EMM model is based on the neoclassical microeconomic theory and supply
functions are specified for each subsector within each zone. In the model, an aggregate producer represents
a specific zone‘s production of a specific sector. There are a total of 1,792 (32 sub-sectors x 56 zones) such
representative producers in the agricultural sector. Consistent with the setup of many other multimarket
models, the supply function, rather than the production function, is used to capture each representative
producer‘s response to market conditions. Specifically, the supply functions are derived under producer
profit-maximization and based on the producer prices of all commodities (including the prices for the non-
agricultural commodities). Risk and market imperfections are not taken into account and therefore do not
affect producers‘ profit-maximization decision in the model. In the crop sub-sectors, the supply functions
have two components: (i) yield functions that are used to capture supply response to the own prices given
farmland allocated to this crop; and (ii) land allocation functions that are functions of all prices and hence
are responsive to changing profitability across crops given the total available land:
19. See Bell and Hazell (1980), Kuyvenhoven (1978) and Tinbergen (1966) for further discussion of the SIO
method. In cases where equations are specified for all accounts of a complete social accounting matrix,
SIO models are also termed "constrained SAM multiplier models".
56
qiZR
tiZRqq
tiZR PY ,,
,,,tZ,i,R,,,,
(5.1)
j tjRqq
tiZR
qjZRPA ,,
,,ti,Z,R,,,,
and ,0,, J
j
qjZR (5.2)
In Equation (5.1) qtiZRY ,,, is the yield for crop i with technology q in Ethiopia's region R and zone Z, while in
Equation (5.2) qtiZRA ,,, is the harvest area for crop i defined in a similar way. In both equations PR,Z,i is
producer price for i in region R and zone Z. qtiZR ,,, and q
tiZR ,,, are shift parameters to capture growth in
yield and area expansion, which are exogenous and zonal specific.
The production of major staple crops and livestock products involves a variety of technologies, q.
For staple crops, modern inputs and their effects on crop productivity are captured through the
identification of 15 different technologies, maize production, for example, incorporates four primary
modern inputs—fertilizer, improved seeds, pesticide, and irrigation (individually or jointly)—and also
includes production without modern inputs. While the model captures the average difference in crop yields
across technologies, the marginal effect of increased use of an input for a given technology is not captured
because such input uses are not explicitly included in the supply function. The yield gaps, which is
represented by the differences in qtiZR ,,, in (5.1) across q within a zone for a same crop, are defined at the
zonal level and are consistent, by zone, with data from the national agricultural sample surveys for 1997
and 2000. Data on irrigation was also available for cash crop production and hence was employed in
supply functions for those crops.
The EMM model is dynamic and thus both yields and land change over time. To capture such
changes, growth rate in yields, qiZRY
g,,
, acts on the productivity shift parameter
q
iZRY
qqtiZR g
,,
1ti,Z,R,1,,, ,
while crop area expands as qtiZR ,,, is a function of an annual area expansion rate. Shocks to the model to
simulate improved production performance are introduced through changes in the productivity growth rate
at the zonal level and can be crop specific.
Combining equations 5.1 and 5.2 gives total supply for a specific crop in a region's zone:
q
tiZR
q
tiZRqtiZR AYS ,,,,,,,,, (5.3)
Supply function for the non-crop production is similar as the area function defined in (5.2). For
livestock, the model captures the productivity difference between traditional and modern technologies. For
example, three types of cattle are raised to produce beef: draught animals, from which beef is a by-product;
beef animals, using traditional technology; and beef stock, using improved technology. The productivity
(yield) gaps resulting from the use of different types of technologies in animal production are reflected in
the supply function. Moreover, the supply function also captures the difference in feed use between
traditional and modern technologies. Livestock production under modern technology requires feed grain,
while under traditional production it assumes feeding via grazing only. The feed-grain demand function is
therefore defined only for improved technology, and is a function of grain crop prices. Different
technologies are similarly defined for dairy, poultry, and sheep and goats.
57
Demand function of the EMM model
The demand side of the model is also defined at zonal level. Representative consumers are defined for
the rural and urban in each zone. This representative consumer‘s demand for each consumption good is
derived as following:
IiZRHjiZRH
tZRHj tjZRtiZRH GDPpcPCDpc ,,,,,,,
,,,,,,,,,,
, (5.4)
where DpcH,R,Z,i is per capita demand for commodity i in region R and Zone Z's rural or urban areas,
and PCR,Z,j is the consumer price for good j in the same zone. j = 1,2,…,42 (including both agricultural and
non-agricultural goods.) GDPpcH,R,Z is per capita income for the rural or urban consumers in region R and
zone Z. jiZRH ,,,, is price elasticity between demand for commodity i and price for commodity j, and
IiZRH ,,, is income elasticity.
Exports, imports, producer and consumer prices
As the name of the model suggests, a multiple market structure is specified. There is perfect
substitution between domestically and internationally produced commodities. However, transportation and
other market costs distinguish trade in the domestic market from imports and exports. For example, while
imported maize is assumed to be perfectly substitutable with domestically produced maize in consumers‘
demand functions, maize may still not be profitable to import if its domestic price is lower than the import
parity price less any transactions costs. Maize imports can only occur when domestic demand for maize
grows faster than domestic supply and the local market price rises significantly. A similar situation applies
to exported commodities. Even though certain horticultural products are exportable, if domestic production
is not competitive in international markets, either due to low productivity or high transactions costs, then
exports will not be profitable. Only when domestic producer prices plus market costs are lower than the
export parity price of the same product does it become profitable to export.
The model does not capture bilateral trade flows across sub-national region within Ethiopia. However,
the model does identify sub-national regions as being food surplus or deficit by comparing regional level
demand and supply for total food commodities. Thus, the model provides useful information to justify
possible intra-regional trade in the analysis. While producers and consumers in different regions operate in
the same national markets for specific commodities, prices can vary across regions due to differences in
transportation and market costs. For example, domestic marketing margins are defined at the regional level
according to the distance to Addis Ababa, which represents the central market for the country. For a food
surplus region, food crop prices faced by local producers are equal to the prices in the central market
subtracting market margins, while for a food deficit region local prices are higher than those in the central
market due to marketing margins.
For most agricultural commodities (except for coffee and a few non-traditional export crops, and rice
and wheat that are import commodities), some manufacturing goods and services, prices are endogenously
determined by the equilibrium between demand (including consumption, feed and other demand) and
supply in the country's domestic markets, at least in the early periods in the model. Such endogenized
prices are one of the key differences between the EMM and SIO models. On the other hand, the EMM
model does capture the price linkages between domestic and international markets, which occurs if
domestic prices convert to the import parity prices as an outcome that demand grows faster than supply or
to the export parity prices when supply grows faster than demand. In such situation either import or export
occurs, if even there is no trade for this commodity initially. Specifically, the following relationship
describes the possible linkages between import parity prices and consumer prices in the domestic markets:
58
,1, iiAddisti PWMWmPC
Mi > 0 if ―=‖ (5.5)
where Wmi is the trade margin for commodity i between border prices, PWMi, and consumer prices in
Addis market. When PCi is less than (1+WmR,i)PWMi, PCi is an endogenous price determined by domestic
supply and demand. The equation holds only when the imports are positive. In this situation, domestic
prices perfectly link with the world price and thus become exogenous.
The relationship between zonal-level and national market prices for consumers is as follows:
Addis
tiiZRtiZR PCDgapPC ,,,,,, 1 , (5.6)
where iZRDgap ,, is negative if Z is in food surplus area and positive if Z is in the food deficit area. If
PCi is endogenous in the national market, so does PCR,Z,i in the regional markets.
Similarly, there is following relationship between domestic producer prices and export parity prices:
iiAddisti PWEWeP 1, , Ei > 0 if ―=‖ (5.7)
where Pi is producer prices in Addis and PWEi is export border prices. If Pi is greater than (1-
WeR,i)PWEi, Pi is an endogenous price determined by domestic supply and demand. The equation holds
only when the exports are positive. Consumer and producer prices are not necessary the same, such that:
tiZRiZRtiZR PDmPC ,,,,,,,, 1 , (5.8)
where Dm is the margin between consumer and producer prices in domestic market. In summary, the
following relationship holds between domestic market and import/export parity prices:
iiAddisti
Addistiii PWMWmPCPPWEWm 1)1( ,, , and (5.9)
ZRH
tZRHtiZRHtititiZRZR
PoPDpcEMS,,
,,,,,,,,,,,,,
. (5.10)
Equation 5.10 solves for the price of commodity i if both M and E are zero in the country. Otherwise,
it solves for the value of M or E.
GDP and per capita zonal income function
Distinguished from most multimarket models that are usually partial equilibrium in nature, the per
capita income at the zonal-level is an endogenous variable in the EMM model, because all economic
activities (including the non-agricultural ones) are included. Per capita income is determined by the zonal-
level value added divided by population, rural and urban, respectively. Because of such setup, the model
has a general equilibrium nature, which allows production and consumption decisions to be linked at the
zonal level. Similar as a CGE model, intermediate inputs are explicitly included in the model through fixed
input-output relationship with sector‘s production. The IO coefficients are drawn from the SAM. The
aggregate of agricultural production value added equals agricultural GDP (henceforth, AgGDP), and the
sum-total of agricultural and non-agricultural value added equals national GDP. Both AgGDP and GDP are
endogenous in the model:
59
2,1, ,...;2,1,. ,,,,,,,,, iSPNonagGDPjSPAgGDPi
tiitj
tjZRtjZRtZR (5.11)
We assume that agricultural income at zonal level goes to the rural households within this zone and
urban households earn income from the aggregate non-agricultural sectors only. However, part of non-
agricultural income is also shared by the rural households and initial income levels for an average rural and
urban household, together with rural and urban population distribution, determine the share. Given this
share, per capita income is endogenously determined by changes in agricultural and non-agricultural GDP:
1, ,
,,,
,,,,,,
,,,
ZRrural
tZRH
tNZRHtZRZR
H
tZRH sPoP
NonagGDPsAgGDPsGDPpc
, (5.12)
where PoPH,R,Z represents zonal level rural or urban total population, and it grows exogenously
accordingly to the country's recent population growth rate.
Elasticities in supply and demand function
Similar as for any other simulation model, the EMM model results critically depend on the elasticities
applied in both supply and demand functions of the model. Ideally, the elasticities should be estimated
using the similar sources of data that the model is built on. While the supply elasticities for some food
crops (maize, teff, wheat, barley, sorghum and pulses) are estimated using log-linear functions based on
Ethiopian Agricultural Sample Survey data (1995/6 – 2000/01) and price information at the national level,
it is almost impossible to obtain similar elasticities for many other crops and for the other agricultural
subsectors, given that there are more than 30 such sectors included in the model, and there are no enough
data to estimate supply functions for most of them. Moreover, it is infeasible to estimate the elasticities for
the zonal level supply functions due to the relatively short time series. Thus, a similar elasticity is
employed for the supply function of a similar commodity across zone, and 0.2 (the average level of
estimated elasticities for above six crops) is chosen as the own price elasticity in the supply functions for
those commodities that are unable to estimate due to lack of the data. The negative cross-price elasticities
in the supply function are then derived from the own-price elasticity multiplied by the value share of each
commodity (at the zonal level). The homogeneity of degree zero condition is imposed on the supply
function such that, within each time period, there is no supply response if all prices change proportionally
in a same way. The constraint on crop area function is also imposed to avoid a simultaneous expansion of
all crop areas over a given time period.
While there is similar elasticity in supply functions for a same crop across zones, due to the
combinations of 15 different technologies vary across zone, the aggregate supply response to the price
change in a similar crop can be different across zone. For example, maize produced in the food surplus
areas of Ethiopia often uses more fertilizer than that produced in the food deficit areas, which implies that
the supply function representing the traditional technology dominates total supply of maize in the deficit
areas, while the supply functions employing both traditional and modern technologies are equally
important in some zones in the surplus areas. Due to differences in the technical coefficients of the supply
functions representing different types of technology, the aggregate supply responses to increases in the
maize prices will be stronger in the surplus areas than in the deficit areas (the technical coefficients in the
supply function representing the modern technology are, on average, 50% to 100% greater than those in
the same functions representing the traditional technology).
On the demand side, income elasticity can be estimated using household survey data. We
econometrically estimated such income elasticity using Ethiopia's Household Income, Consumption, and
Expenditure Survey (HICES [CSA 2000]). The estimation method is drawn from King and Byerlee, 1978
60
(see appendix for the discussion of estimation method). The price elasticities are then derived from the
linear expenditure of demand system using the current expenditure share and income elasticity such that
the budget constraint is satisfied for each demand function. That is ,0,,,,,,, J
j
IiZRHjiZRH and
,1,,,,,, J
j
IjZRHjZRHsh where iZRHsh ,,, is the expenditure share of commodity i for household H in region
R and zone Z.
Both income and price elasticities for any specific commodity vary across zone due to different
consumption patterns and income levels. The variations in income level across zone affect the ratio of
subsistent consumption over market demand for a specific commodity, while the variations in consumption
patterns affect the average budget share of each commodity in total expenditure. These two factors
determine that for a demand function of similar commodity there are different income and price elasticities
across zones. For a representative consumer of a zone with lower level per capita income and large budget
share of a specific staple crop, both income and own price elasticities in the demand function for this
commodity are relatively lower than those in the demand function for another representative consumer
from a zone with higher level of per capita income and smaller budget share of the same commodity.
The EMM model thus characterized is deployed to explore growth linkages and impact on poverty
reduction. The initial stimulus to growth is introduced as an exogenous productivity growth. Moreover, a
'baseline' and four different productivity growth scenarios are considered. To make the resulting linkage
estimates comparable, the impact of sector's size and growth potential need to be take into account.
Assuming similar growth rates at the sub-sector level, greater economy-wide growth will be generated by
the larger sub-sectors, in turn producing a (generally) larger effect on poverty. On the other hand, small
sub-sectors have greater capacity to grow rapidly and require the investment of fewer resources to do so.
Thus, in determining whether a sub-sector will ultimately drive growth, both the linkage effects on the
economy and poverty as well as the growth potential (determined by supply and demand factors) must be
considered. In order to ensure comparable quantitative measurement across the agricultural sub-sectors
modelled, those exhibiting similar total GDP growth but different productivity growth were examined to
assess the growth effect of each on overall economic growth and poverty reduction.
To analyze the growth-poverty effect, the nationally-defined poverty line is adopted in the model
rather than using the World Bank‘s ‗a-dollar-a-day‘ measure. National poverty lines are typically measured
by household total expenditure, since household income is often significantly underreported in developing
countries. The household level expenditure data from HICES is used to develop a micro-simulation model
to capture detailed household consumption patterns. This micro-simulation model is linked with the EMM
model for calculating poverty rates at the regional or national level. The calculation of poverty lines and
fraction of populations below the poverty lines in the simulations and other detailed mathematical
descriptions of the model can be found in Diao et al. (2005).
61
APPENDIX B: APPENDIX TABLES
Table A 1. The structure of the 2006/07 Ethiopian Social Accounting Matrix (SAM)
Agricultural sectors/commodities Nonagricultural sectors/commodities
(1) Teff, (2) Barley, (3) Maize, (4)
Sorghum, (5) Wheat
(6) Pulses, (7) Enset, (8) Oil seeds
(9) Vegetables, (10) Fruit crops, (11)
Sugar cane and sugar beet, (12) Cotton,
(13) Tea, (14) Chat, (15) Coffee, (16)
Tobacco, (17) Cut flowers, (18) Other
crops
(19) Cattle, (20) Poultry, (21) Other small
livestock, (22) Raw milk
(23) Forestry
(24) Fishing
(25) Coal and other mining
(26) Meat and oilseed products, (27) Dairy products, (28) Grain
mill products, (29) Grain mill services, (30) Sugar and sugar
confectionary, (31) Beverages, (32) Tobacco products, (33)
Manufacturing of tea, (34) Other processed food products
(35) Cotton lint, (36) Tread and yarns, (37) Fibre, (38) Other
textiles, (39) Wearing apparel, (40) Leather products
(41) Wood products, (42) Paper products publishing, (43)
Chemicals, rubber and plastic products, (44) Non-metal products,
(45) Metal products, (46) Motor vehicles and parts, (47) Other
transport equipment, (48) Electronic equipment, (49) Machinery
and equipment, (50) Other manufactures
(51) Electricity, (52) Water, (53) Construction
(54) Trade and repair services, (55) Hotels and restaurants, (56)
Transport services, (57) Communication, (58) Financial services,
(59) Business services, (60) Real estate and renting services, (61)
Recreation and other private services
(62) Public administration and defence, (63) Education, (64)
Health
Source: Ethiopian SAM 2006/07.
Table A 2. Income distribution in SAM
Household Type
Income (in billion Birr)
Share (%)
Rural poor households 32.4 25.0
Rural not poor households 68.0 53.4
Urban poor households 6.6 4.2
Urban not poor households 27.3 17.5
Total 134.4 100 Source: Ethiopian SAM 2006/07.
62
Table A 3. Household consumption spending patterns in Ethiopia
Rural poor
Rural non-
poor Urban poor Urban non-poor
Food 62.7 54.1 56.1 32.7
Cereals 25.4 19.5 17.1 7.6
Enset, pulses and oilseeds 6.6 6.0 3.7 1.5
Sugar, coffee, tea, tobacco
and chat 5.4 5.9 6.9 4.1
Fruits and Vegetables 2.2 1.8 1.7 0.9
Other Crops 4.2 3.0 4.1 1.8
Livestock, livestock
products and fish 16.0 15.0 14.4 11.7
Other processed food 2.9 2.9 8.3 5.1
Non-food agriculture and
industrial goods 18.3 24.6 22.5 37.0
Forestry, wood and energy 5.6 7.0 3.5 3.0
Textile and clothing 4.2 5.8 4.6 6.2
Other manufacturing 7.7 11.2 11.7 24.8
Utility and construction 0.8 0.7 2.6 3.0
Services 19.0 21.4 21.4 30.3
Private services 18.3 20.8 19.7 28.8
Public services 0.7 0.6 1.7 1.6 Source: Ethiopian SAM 2006/07.
63
Table A 4. Composition of demand and supply by sector in SAM
Demand (total demand = 100)
Supply (total supply = 100)
Households
Intermediate
inputs Others Exports
Domestic
production Imports
Food 71.0 18.8 2.0 8.3 96.3 3.7
Cereals 86.7 12.3 0.0 1.1 92.9 7.1
Enset, pulses and oilseeds 64.5 18.1 1.5 15.9 99.2 0.8
Sugar, coffee, tea, tobacco
and chat 48.0 15.8 5.5 30.6 98.7 1.3
Fruits and Vegetables 88.4 6.8 0.1 4.7 99.5 0.5
Other Crops 82.2 8.9 0.9 8.1 95.6 4.4
Livestock, livestock products
and fish 69.6 23.8 3.2 3.4 97.4 2.6
Other processed food 59.5 40.5 0.0 0.0 95.7 4.3
Non-food agriculture and
industrial goods 30.6 35.0 33.8 0.7 68.2 31.8
Forestry, wood and energy 67.2 30.3 2.5 0.0 95.4 4.6
Textile and clothing 79.6 12.7 0.0 7.6 72.3 27.7
Other manufacturing 28.0 47.8 24.2 0.0 47.9 52.1
Utility and construction 5.8 16.0 77.9 0.2 100.0 0.0
Services 41.0 32.0 25.3 1.7 90.8 9.2
Private services 54.4 43.4 0.0 2.2 87.5 12.5
Public services 5.8 1.7 92.2 0.3 99.6 0.4 Source: Ethiopian SAM 2006/07.
64
Table A 5. Sensitivity test result - gains in GDP
Teff Barley Maize Sorghum Cotton Coffee Cattle Poultry
Grain
milling
Thread
&
yarns Construction
Result used in Table 4.1 1.42 1.27 1.47 1.51 1.56 1.54 1.56 1.55 0.61 1.01 1.08
Most industrial sectors
constrained in supply 1.23 1.11 1.28 1.35 1.40 1.36 1.40 1.39 0.50 0.80 0.51
Most industrial sectors
with unconstrained supply 1.48 1.32 1.52 1.57 1.62 1.59 1.61 1.60 0.64 1.26 1.13
Difference from results in Table 4.1 (% difference from Table 4.1)
Most industrial sectors
constrained by supply -13.6 -13.0 -12.9 -11.1 -10.5 -11.4 -10.3 -10.3 -18.4 -20.9 -52.2
Most industrial sectors
with unconstrained supply 3.7 3.6 3.6 3.6 3.6 3.7 3.6 3.6 4.3 24.3 4.5 Source: Ethiopia SIO model results.
65
Table A 6. Sensitivity test result - gains in total household income
Teff Barley Maize Sorghum Cotton Coffee Cattle Poultry
Grain
milling
Thread
&
yarns Construction
Result used in Table 4.2 1.37 1.23 1.42 1.46 1.51 1.48 1.50 1.49 0.58 0.96 0.99
Most industrial sectors
constrained by supply 1.20 1.08 1.25 1.31 1.37 1.33 1.36 1.36 0.48 0.78 0.50
Most industrial sectors
with unconstrained supply 1.41 1.26 1.46 1.50 1.55 1.53 1.55 1.54 0.60 1.19 1.03
Difference from results in Table 4.2 (% difference from Table 4.2)
Most industrial sectors
constrained by supply -12.3 -11.8 -11.7 -10.0 -9.3 -10.2 -9.2 -9.2 -16.9 -19.3 -49.1
Most industrial sectors
with unconstrained supply 3.1 3.1 3.1 3.1 3.1 3.1 3.1 3.1 3.8 23.7 4.1 Source: Ethiopia SIO model results.
66
Table A 7. Sensitivity test result - income ratio of rural poor household (rural poor household income in SAM is 1)
Teff Barley Maize Sorghum Cotton Coffee Cattle Poultry
Grain
milling
Thread
&
yarns Construction
Result used in Table 4.2 1.15 1.24 1.08 1.01 0.54 0.55 1.35 1.34 0.83 0.84 0.84
Most industrial sectors
constrained by supply 1.20 1.29 1.11 1.03 0.51 0.52 1.40 1.39 0.83 0.85 0.87
Most industrial sectors
with unconstrained supply 1.14 1.22 1.07 1.01 0.54 0.56 1.33 1.32 0.83 0.84 0.84
Difference from results in Table 4.2 (% difference from Table 4.2)
Most industrial sectors
constrained by supply 4.3 4.7 3.4 2.2 -5.3 -5.3 4.0 4.0 0.3 1.0 3.8
Most industrial sectors
with unconstrained supply -0.9 -1.0 -0.8 -0.6 1.5 1.4 -1.2 -1.2 -0.1 -0.3 -0.1 Source: Ethiopia SIO model results.
67
Table A 8. Sensitivity test result - income ratio of rural non-poor household (rural non-poor household income in SAM is 1)
Teff Barley Maize Sorghum Cotton Coffee Cattle Poultry
Grain
milling
Thread
&
yarns Construction
Result used in Table 4.2 0.73 0.71 0.71 0.75 0.68 0.81 0.68 0.68 1.83 1.73 1.83
Most industrial sectors
constrained by supply 0.55 0.55 0.54 0.61 0.55 0.68 0.55 0.55 1.81 1.68 1.71
Most industrial sectors
with unconstrained supply 0.76 0.75 0.74 0.78 0.72 0.84 0.72 0.72 1.83 1.74 1.83
Difference from results in Table 4.2 (% difference from Table 4.2)
Most industrial sectors
constrained by supply -24.2 -23.6 -23.8 -17.9 -19.1 -16.1 -18.9 -18.8 -1.0 -2.9 -6.4
Most industrial sectors
with unconstrained supply 4.9 5.1 5.1 4.7 5.3 4.1 5.4 5.4 0.1 0.6 0.2 Source: Ethiopia SIO model results.
68
Table A 8. Agricultural commodities included in the economy-wide, multi-market model
Maize, Teff, Wheat, Sorghum, Barley, Millet, Oats, Rice,
Potatoes, Sweet potatoes, Enset, Other root crops,
Beans, Peas, Other pulses,
Groundnuts, Rapeseed, Sesame, Other oil crops,
Domestic vegetables, Bananas, Other domestic fruits,
Exportable vegetables, Other horticultural crops, Chat, Cotton,
Coffee,
Sugar, Beverages and spices
Bovine meat, Goat meat and mutton, Other meat,
Milk and dairy products,
Poultry and eggs, Fish
69
APPENDIX C: SENSITIVITY TEST OF THE SIO MODEL RESULTS
We have conducted two series of sensitivity tests to assess the robustness of the SIO model results.
We found that while different level of linkage effects are obtained by varying supply elasticity in the
sensitivity tests, the general direction of linkages effects under different growth options holds constant. In
the first series of tests, we assume all industrial sectors are supply inelastic except for mining and a few
food processing sectors, while in the second series of tests, supply in all the industrial sectors are highly
elastic, except for the investigated three sectors. In both cases, most private services are supply elastic.
When supply is inelastic, it implies the sector's output will not change and changes in demand for the
sector's good are met by trade, while when supply is highly elastic, it assumes that there is no resource and
other constraints such that output changes in response to the changes of demand for it.
We report selected results that are shown in Tables 4.1 and 4.2 to discuss the sensitivity test. Four
indicators are selected: changes in GDP, changes in total household incomes, ratios of income share for the
rural poor and urban non-poor household groups. These four indicators are reported in Tables A5-A8,
respectively.
The test results show that assumptions of the constraints in industrial supply affect the magnitude of
the changes in GDP and total income for growth led by any agricultural and non-agricultural sector, but
such difference in the magnitude will not modify the major findings we discussed in Chapter 4. Change in
GDP led by agricultural growth is consistently higher than that led by non-agricultural growth. This is also
true for additional total income gains for households generated from a sector-led growth. As expected,
when supply response of industrial sectors is constrained, gains in both GDP and total income are smaller
than when the supply response is unconstrained. While magnitude of the gains in GDP and total income
led by the growth in the industrial sectors in general is more sensitive to the assumption of supply elasticity
than growth led by the agricultural sectors, within non-agriculture, growth led by the sector with more
detail up-and down-stream disaggregation in the SAM (which is thread and yarns in this case) is more
sensitive to the assumption of supply elasticity than the others.
The sensitivity test on the income distribution effect reveals quite different results between the rural
and urban households. While the assumption of supply elasticity in the industrial sectors has very modest
effect on the income ratio for the rural poor household group, its effect on the income ratio of urban non-
poor household group is much larger, particularly when most industrial sectors are constrained in supply
response. These results make sense as the rural poor households learn more income from the agricultural
sectors while the urban non-poor households learn incomes from the non-agricultural sectors only.
70
REFERENCES
Abrar, Suleiman, Oliver Morrissey, and Tony Rayner (2004). ―Aggregate Agricultural Supply Response in
Ethiopia: A Farm-Level Analysis.‖ Journal of International Development 16: 605–620.
Adelman, Irma (1984) ‗Beyond Export-led Growth,‘ World Development, 12, 9: 937-949.
Ayele, Gezahegn, Fasil Kelemework, and Dawit Alemu (2003). ―The Provision of Rural Services and
Agricultural Productivity in Ethiopia: A Characterization.‖ Phase 1 report for the 2020 Vision
Network‘s East Africa regional project, ―Assessing the Impact of Policies for Rural Service
Provision on Agricultural Productivity and Food Security,‖ EARO/IFPRI, June 2003. Washington,
D.C.: International Food Policy Research Institute.
Bagachwa, M.S.D. (1981): ―The Urban Informal Enterprise Sector: The Case Study of Arusha Town‖,
ERB Paper, No. 81.4, University of Dar es Salaam.
Bagachwa, M., and T. Stewart (1992). ―Rural Industries and Rural Linkages in Sub-Saharan Africa: A
Survey,‖ in F. Stewart, S. Lall, and S. Wangwe (eds) Alternative Development Strategies for Sub-
Saharan Africa. Macmillan, London.
Bell, C. L. G., and P. B. R. Hazell (1980). ―Measuring the Indirect Effects of an Agricultural Investment
Project on its Surrounding Region,‖ American Journal of Agricultural Economics, 62 (February):
75–86.
Bigsten, A., and P. Collier (1995). ―Linkages from agricultural growth in Kenya,‖ in J. W. Mellor. (ed.)
Agriculture on the Road to Industrialization, Johns Hopkins University Press for the International
Food Policy Research Institute: Baltimore, Md., U.S.A.
Binswanger, Hans (1989). ―How Agricultural Producers Respond to Prices and Government Investments,‖
Annual Bank Conference on Development Economics, World Bank, Washington, D.C.
Boserup, E. (1965). The Conditions of Agricultural Growth. New York: Aldine.
Byerlee, Derek, David Spielman, Dawit Alemu, and Madhur Gautam (2007). Policies to Promote Cereal
Intensification in Ethiopia: A Review of Evidence and Experience. IFPRI Discussion Paper 00707.
Washington DC: International Food Policy Research Institute.
Crawford, Eric, Valerie Kelly, T. S. Jayne, and Julie Howard (2003). ―Input Use and Market Development
in Sub-Saharan Africa: An Overview.‖ Food Policy 28: 277–292.
Croppenstedt, Andre, and Mulat Demeke (1997). ―An Empirical Study of Cereal Crop Production and
Technical Efficiency of Private Farmers in Ethiopia: A Mixed Fixed-Random Coefficients
Approach.‖ Applied Econometrics 29: 1217–1226.
CSA (Central Statistic Authority) (June, 2008). ―Agricultural Sample Survey 2007/2008 (2000 E.C.):
Volume I - Report on Area and Production Crops (Private Peasant Holdings, Meher Season),‖
Statistical Bulletin 417, Central Statistical Agency (CSA), the Federal Democratic Republic of
Ethiopia, Addis Ababa.
____ (May 2007). ―Household Income, Consumption and Expenditure (HICE) Survey 2004/5: Volume I -
Analytical Report,‖ Statistical Bulletin 394, Central Statistical Agency (CSA), the Federal
Democratic Republic of Ethiopia, Addis Ababa.
71
____ (Jul 2006). ‗Report on Area and Production of Crops (Private Peasant Holdings, Meher Season),‘
Ethiopian Agricultural Sample Survey (2005/06 (1998 E.C.)) – Volume I, Statistical Bulletin 361,
Central Statistical Agency (CSA), Addis Ababa, Ethiopia.
____ (September 2006). ‗Report on Farm Management Practices for Belg/Second Season Crops,‘
Ethiopian Agricultural Sample Survey (2005/06 (1998 E.C.)) – Volume VI, Statistical Bulletin 361,
Central Statistical Agency (CSA), Addis Ababa, Ethiopia.
____ (August 2006). ‗Report on Area and Production of Belg Season Crops for Private Peasant Holdings,‘
Ethiopian Agricultural Sample Survey (2005/06 (1998 E.C.)) – Volume V, Statistical Bulletin 361,
Central Statistical Agency (CSA), Addis Ababa, Ethiopia.
____ (October 2005). ‗Report on Area and Production of Belg Season Crops for Private Peasant Holdings,‘
Ethiopian Agricultural Sample Survey (2004/05 (1997 E.C.)) – Volume V, Statistical Bulletin 331,
Central Statistical Authority (CSA), Addis Ababa, Ethiopia.
____ (November 2004). ‗Report on Area and Production of Belg Season Crops for Private Peasant
Holdings,‘ Ethiopian Agricultural Sample Survey (2003/04 (1996 E.C.)) – Volume V, Statistical
Bulletin 302, Central Statistical Authority (CSA), Addis Ababa, Ethiopia.
____ (2003a). ‗Report on Large and Medium Scale Manufacturing and Electricity Industries Survey 2002,‘
Statistical Bulletin, No. 281, Ethiopian Central Statistical Authority, Addis Ababa, Ethiopia. CSA
(August 2008). ―Agricultural Sample Survey 2007/2008 (2000 E.C.): Volume III - Report on Farm
Management Practices (Private Peasant Holdings, Meher Season),‖ Statistical Bulletin 417, Central
Statistical Agency (CSA), the Federal Democratic Republic of Ethiopia, Addis Ababa.
____ (July, 2003b). ‗Report on Socio-Economic Characteristics of the Population in Agricultural
Households, Land Use, and Area and Production of Crops,‘ Ethiopian Agricultural Sample
Enumeration (2001/02 (1994 E.C.)) – Results at the Country Level, Part I, Central Statistical
Authority, Addis Ababa, Ethiopia.
____ (July, 2003c). ‗Report on Farm Management Practices, Livestock and Farm Implements,‘ Ethiopian
Agricultural Sample Enumeration (2001/02 (1994 E.C.)) – Results at the Country Level, Part II,
Central Statistical Authority, Addis Ababa, Ethiopia.
____ (February 2002). ‗Report on Area and Production of Belg Season Crops for Private Peasant
Holdings,‘ Ethiopian Agricultural Sample Survey (2000/01 (1993 E.C.)) – Volume V, Statistical
Bulletin 245, Central Statistical Authority (CSA), Addis Ababa, Ethiopia.
____ (2000a). Household Income, Consumption, and Expenditure Survey 1999/2000. Ethiopian Central
Statistical Authority (CSA), Addis Ababa: Ethiopia.
____ (November 2000b). ‗Report on Area and Production of Belg Season Crops for Private Peasant
Holdings,‘ Ethiopian Agricultural Sample Survey (1999/2000 (1992 E.C.)) – Volume V, Statistical
Bulletin 244, Central Statistical Authority (CSA), Addis Ababa, Ethiopia.
____ (December 1999). ‗Report on Area and Production of Belg Season Crops for Private Peasant
Holdings,‘ Ethiopian Agricultural Sample Survey (1998/99 (1991 E.C.)) – Volume V, Statistical
Bulletin 217, Central Statistical Authority (CSA), Addis Ababa, Ethiopia.
____ (March 1999). ‗Report on Area and Production for Major Crops (Private Peasant Holdings, Meher
Season),‘ Ethiopian Agricultural Sample Survey (1998/99 (1991 E.C.)), Statistical Bulletin 200,
Central Statistical Authority (CSA), Addis Ababa, Ethiopia.
____ (December 1998a). ‗Report on Area and Production of Belg Season Crops for Private Peasant
Holdings,‘ Ethiopian Agricultural Sample Survey (1997/98 (1990 E.C.)) – Volume V, Statistical
Bulletin 196, Central Statistical Authority (CSA), Addis Ababa, Ethiopia.
72
____ (December 1998b). ‗Report on Farm Management Practices for Belg Season Crops for Private
Peasant Holdings,‘ Ethiopian Agricultural Sample Survey (1996/97 (1989 E.C.)), Statistical Bulletin
171, Central Statistical Authority (CSA), Addis Ababa, Ethiopia.
____ (September 1998c). ‗Report on Farm Management Practices (Private Peasant Holdings, Meher
Season),‘ Ethiopian Agricultural Sample Survey (1997/98 (1990 E.C.)), Statistical Bulletin 193,
Central Statistical Authority (CSA), Addis Ababa, Ethiopia.
CSA and NFIA (October 1998). Report on National Fertilizer Benchmark Survey of 1996/97 (1989 E.C.)
and 1997/98 (1990 E.C.), Central Statistical Authority (CSA) and National Fertilizer Industry
Agency, the Federal Democratic Republic of Ethiopia, Addis Ababa.
Delgado, Christopher L., Jane Hopkins, Valerie A. Kelly, with Peter Hazell, Anna A. McKenna, Peter
Gruhn, Behjat Hojjati, Jayashree Sil, and Claude Courbois (1998). Agricultural Growth Linkages in
Sub-Saharan Africa, International Food Policy Research Institute, Washington, D.C.
Dercon, Stefan and Andrew Zeilin. 2009. Rethinking agriculture and growth in Ethiopia: A conceptual
discussion. Paper prepared as part of a study on agriculture and growth in Ethiopia.
Diao, Xinshen, and Alejandro Nin Pratt (2007). "Growth Options and Poverty Reduction in Ethiopia: A
Economy-wide Model Analysis," Food Policy 32: 205-228.
Diao, Xinshen, and Alejandro Nin Pratt with Madhur Gautam, James Keough, Jordan Chamberlin,
Liangzhi You, Detlev Puetz, Danielle Resnick, and Bingxin Yu (July 2005). "Growth Options and
Poverty Reduction in Ethiopia: A Spatial, Economy-wide Model Analysis for 2004–15," DSGD
Discussion Paper, No. 20, Development Strategy and Governance Division, International Food
Policy Research Institute (IFPRI).
Dorosh, Paul, and Steven Haggblade (2003). ―Growth Linkages, Price Effects and Income Distribution in
Sub-Saharan Africa,‖ Journal of African Economies 12(2): 207-235.
Fan, Shenggen, and Peter Hazell (2001). ―Returns to Public Investments in the Less-Favored Areas of
India and China.‖ American Journal of Agricultural Economics 83 (5): 1217–1222.
Federal Democratic Republic of Ethiopia (2002) Agricultural Census 2001/02. Addis Ababa: Central
Statistics Authority.
Fernandez-Rivera, S., I. Okike, and S. Ehui (2001). ―The ―Livestock Revolution‖: Implications for the
Demand for Feed in developing countries.‖ Paper presented at the Forage Adoption Workshop,
International Livestock Research Institute, Addis Ababa, June 2001. Gabre-Madhin, Eleni Z. 2001.
Market Institutions, Transaction Costs, and Social Capital in the Ethiopian Grain Market. IFPRI
Research Report No. 124. Washington, D.C.: International Food Policy Research Institute.
Gabre-Madhin, Eleni and Ian Goggin (2006). ―Does Ethiopia Need a Commodity Exchange?: An
Integrated Approach to Market Development.‖ Paper presented at the Ethiopia Strategy Support
Program Conference, 6-8 June, Addis Ababa.
Haggblade, S., and P. Hazell (1989). ―Agricultural Technology and Farm-NonFarm Growth Linkages,‖
Agricultural Economics, 3: 345-364.
Haggblade, S., J. Hammer, and P. B. R. Hazell. 1991. ―Modeling agricultural growth multiplier,‖
American Journal of Agricultural Economics, 73(2): 361–374.
Hazell, Peter, and C. Ramasamy (1991). The Green Revolution Reconsidered. Baltimore: John Hopkins
University Press.
Howard, Julie, Eric Crawford, Valerie Kelly, Mulat Demeke, and Jose Jaime Jeje (2003). ―Promoting
High-Input Maize Technologies in Africa: The Saskawa-Global 2000 Experience in Ethiopia and
Mozambique.‖ Food Policy 28: 335–348.
73
Hirschman, Albert O. (1958). The Strategy of Economic Development. Yale University Press, New Haven.
IFPRI (2005). The Future of Small Farms: Proceedings of a Research Workshop, International Food Policy
Research Institute (IFPRI), held in Wye, UK, June 26-29, 2005. Washington, DC.
Jayne, T. S., J. Govereh, M. Wanzala, and M. Demeke (2003). ―Fertilizer Market Development: A
Comparative Analysis of Ethiopia, Kenya, and Zambia.‖ Food Policy 28: 293–316.
Johnston, B.F., and P. Kilby (1975). Agriculture and Structural Transformation: Economic Strategies in
Late-Developing Countries. New York: Oxford University Press.
Kelly, Valeric, Akinwumi A. Adesina, and Ann Gordon (2003). ―Expanding Access to Agricultural Inputs
in Africa: A Review of Recent Market Development Experience.‖ Food Policy 28: 379–404.
King, Robert P. and Derek Byerlee (1978). ―Factor Intensities and Locational Linkages of Rural
Consumption Patterns in Sierra Leone.‖ American Journal of Agricultural Economics, May 1978:
197-206.
Kuyvenhoven, Arie. (1978). Planning with the Semi-Input-Output Method: With Empirical Applications to
Nigeria, Martinus Nijhoff Social Sciences Division, Leiden.
Lewis, Blane D., and Erik Thorbecke (1992). ―District-Level Economic Linkages in Kenya: Evidence
Based on a Small Regional Social Accounting Matrix,‖ World Development, 20, 6:881-897
Liedholm, C. and E. Chuta. 1976. The Economics of Rural and Urban Small-Scale Industries In Sierra
Leone. African Rural Economy paper No. 14, Department of Agricultural Economics, Michigan
State University, East Lansing, Michgan.
Mellor, John W and Uma J. Lele (1971). ―A Labor Supply Theory of Economic development,‖ Occasional
Paper, No. 43, Department of Agricultural Economics, Cornell University, Ithaca, NY.
MoFED (July, 2002). Ethiopia: Sustainable Development and Poverty Reduction Program, Ministry of
Finance and Economic Development (MOFED), Federal Democratic Republic of Ethiopia (FDRE),
Addis Ababa, Ethiopia.
MoFED (2005). Ethiopia: Building on Progress: A Plan for Accelerated and Sustained Development to
End Poverty (PASDEP) (2005/06-2009/10), Ministry of Finance and Economic Development
(MoFED), Federal Democratic Republic of Ethiopia, Addis Ababa.
Muriuki H.G. and W. Thorpe (2001). ―Smallholder Dairy Production and Marketing in Eastern and
Southern Africa.‖ In the Proceedings of the South-South Workshop on Smallholder Dairy
Production and Marketing - Constraints and Opportunities. March 12-16. Annand, India.
Pender, John, and Berhanu Gebremedhin (2004). ―Land Management, Crop Production and Household
Income in the Highlands of Tigray, Northern Ethiopia,‖ Environment and Production Technology
Division Discussion Paper. Washington, D.C.: International Food Policy Research Institute.
Sarris, Alexander H. (January 2001). The Role of Agriculture in Economic Development and Poverty
Reduction: An Empirical and Conceptual Foundation, Paper prepared for the Rural Development
Department of the World Bank, mimeo.
Snapp, S. S., M. J. Blackie, and C. Donovan (2003). ―Realigning Research and Extension to Focus on
Farmers‘ Constraints and Opportunities.‖ Food Policy 28: 349–363.
Steel, William F. (1977). Small-Scale Employment and Production in Developing Countries: Evidence
from Ghana, Praeger, New York.
74
Taffesse, Alemayehu Seyoum (2006). ―Product Flows and Growth Linkages in Ethiopia – A Note,‖ in the
Proceedings of the Planning Workshop on Rural-Urban Linkages (RUL) Thematic Research Area of
the Global Mountain Program (GMP), held on August 29-30, 2006 at the International Livestock
Research Institute (ILRI), Addis Ababa Campus.
Thorbecke, Erik, (1994). Intersectoral Linkages and their Impact on Rural Poverty Alleviation: A Social
Accounting Matrix Approach, UNIDO, Vienna.
Tinbergen, J. (1966). Economic Policy: Principles and Design, North-Holland, Amsterdam.
van den Broeck, Katleen. and Stefan Dercan (2001). ―Agricultural Production Changes in Ethiopia.‖ Draft
version www. bu.edu/econ/ied/neudc/papers/Vandenbroeck-final.pdf.
Weir, Sharada (1999). The Effects of Education on Farmer Productivity in Rural Ethiopia. Working Paper
S99-7. Oxford, U.K.: Centre for the Study of African Economies, Department of Economics,
University of Oxford.
Wamisho, Kassu, and Bingxin Yu (2006). ―Estimates of Consumer Demand Parameters for Ethiopia,‖
ESSP/IFPRI, mimeo.
World Bank (2006). Ethiopia: Policies for Pro-Poor Agricultural Growth. Washington, DC: World Bank.
World Bank/EDRI (Ethiopian Development Research Institute), April 2004.
__________ (2007). Ethiopia: Accelerating Equitable Growth - Country Economic Memorandum, The
World Bank.
__________ (2009). World Development Indicators, The World Bank.
Yao S. (1996). ―The Determinants of Cereal Crop Productivity of the Peasant Farm Sector in Ethiopia,
1981–97.‖ Journal of International Development 8 (1): 69–82.