1. Introduction: Planning for the role of agriculture in growth
This paper sets out a conceptual framework for understanding the role of smallholder
agriculture in Ethiopian growth prospects. It seeks to answer two questions: first, what
economic conditions are required to foster growth in smallholder agriculture? And second,
how does growth in agriculture relate to growth in other sectors: to what extent can
agriculture spur growth elsewhere, and to what extent does it depend on other sectors for
its own continuation?
The current growth and development strategy in Ethiopia has its genesis in the policy of
agricultural development‐led industrialization (ADLI). This strategy is well captured by its
name: in essence, growth has to start from growth in agriculture, which then stimulates
growth in other sectors of the economy, most notably industry. It was first very eloquently
articulated in a paper by the then Ministry of Planning and Economic Development in 1993
Much of the intellectual justification for a model of industrialization via agriculture
stems from various interpretations of the historical accounts of the industrial revolution in
Europe, not least in England . Much recent policy work, including the 2007 World
Development Report effectively underwrites the primacy of agricultural development to
stimulate growth in many poor countries such as Ethiopia . Agriculture has retained a
leading role in the recent evolution of the Government of Ethiopia’s conceptualization of
growth, embodied in the Planned for Accelerated and Sustained Development to End
Poverty (PASDEP). In this current framework market mechanisms have taken on increasing
prominence. For example, PASDEP proposes a two‐pronged strategy to the achievement of
a “massive push to accelerate growth”: the commercialization of agriculture, and the
acceleration of private‐sector development (, p. 45).
In this paper, we offer a conceptual framework for understanding whether and how
agriculture‐led growth can work in Ethiopia. We combine theory and evidence from other
growth experiences to discuss both some of the key factors relevant for stimulating growth
in smallholder agriculture in Ethiopia and the role it can play in overall growth.
Much of the evidence on the primacy of agriculture stems from research arguing that
multipliers from growth in agriculture are higher than from any other sector [33, 53]. We
argue in this paper that this approach is incomplete. It is misleading to formulate policy by
comparing multipliers from hypothetical, exogenous increases in the growth rates of
different sectors for two reasons. First is a simple issue of calibration: since we are
ultimately interested in the multipliers stemming from government policy, marginal growth
rates should be weighted by the fiscal cost of bringing about these increments to growth.
Second and more fundamentally is the presumed exogeneity of growth in a ‘driving’ sector:
where this growth comes from cannot be asked independently from what is happening in
To address the second of these, we draw upon the workhorse dual economy models that
have their roots in the seminal work of W. Arthur Lewis . We consider alternative
specifications of such models, their relevance for the Ethiopian case, and their consequences
for the interplay between agriculture and other sectors in the growth process. We also draw
upon the vast empirical literature that estimates (typically static) linkages between sectors.
Putting these elements together, we present a case that modest consumption linkages from
agriculture to industry necessitate policies that actively promote other sectors, ideally
strengthening inter‐sector linkages in the process. The economic incentives required to
sustain agricultural growth require either high and rising urban demand or access to export
markets. Conversely, growth in urban economic activity requires low food prices, which can
be sustained by strategies of imports (which frees agriculture to focus on higher value‐added
products, but which requires improvements to the effective tradability of cereals); by raising
land productivity in cereals production; or by creating more efficient links between urban
and rural sectors. The latter in particular will help provide the incentives required to sustain
agricultural growth, while sustaining low food prices required for this agricultural
productivity to impact the growth of the economy as a whole.
The remainder of the paper proceeds as follows. In the remainder of this section, we
describe the theoretical underpinnings of ADLI. Section 2 distinguishes alternative
mechanisms to growth in agriculture. Section 3 considers the relationship between
agricultural growth and growth in other sectors. Brief conclusions – and a discussion of
policies that aim to change the structure of relationships between sectors, rather than taking
them as given – are presented in Section 4.
Background: the role of agriculture under ADLI
Current policies towards the development of the agricultural sector and its role in the
Ethiopian economy as a whole are guided by the strategy of Agricultural Development Led
Industrialization (ADLI), put forward by the Government of Ethiopia in 1993. ADLI has as its
aim “to bring about a structural transformation in the productivity of the peasant agriculture
and to streamline and reconstruct the manufacturing sector, so that it makes extensive use
of the country’s natural resources and manpower” . Since 1993, this strategy has been
developed further and fine‐tuned, most recently in the more nuanced Plan for Accelerated
and Sustained Development to End Poverty (PASDEP), Ethiopia’s strategic framework for the
five‐year period 2005/06‐2009/10.
This emphasis on structural transformation is reflected in ADLI’s phased approach to
development policy. Under ADLI, the government set out: first, to improve agricultural
technologies, particularly seeds; second, to expand irrigation, infrastructure, and the use of
modern inputs, including fertilizers and pesticides; and third, to expand rural nonagricultural
opportunities. (, p. 19).
What is the understanding of growth process that gives birth to this framework? The basis is
put forward unapologetically as the agricultural sector: “agriculture should be the starting
point for the structural transformation of the economy” (, p.5). A key assertion is that
the primary driver of demand for industrial output will be domestic, rather than foreign,
demand, based on first initiating growth in agriculture. With an eye toward domestic
demand as a constraint on the growth prospects of that sector, it is suggested that incomes
should be a focus (presumably, both of agricultural policy, which is not mentioned, and of
industrial policy). The implication is that labour‐intensive technologies play an important
role. “While agriculture constitutes the primary market, the services sector should not be
ignored as a market for industrial products.” (p. 21). But the role of industry is evidently
quite secondary: it is relegated to creating employment for “redundant” rural labour, and to
providing an “expanding market for other sectors of the economy.”
However, part of the agricultural orientation of Ethiopian development strategy under ADLI
reflects the view that the process of industrialization should build on domestic inputs. While
such a strategy is doubtlessly useful as a means to add value to agricultural production,
recent global trade patterns cast doubt on whether processing of domestic products will
suffice to gain Ethiopia a more substantial point of entry into industrial trade. In particular,
the model of domestic processing is contrary to the growing pattern of trade in tasks ,
whereby trade in manufactured goods has increasingly relied on imported inputs. Recent
research has shown that when countries are unable to export goods that are produced using
imported inputs, their ability to export is significantly hampered: Collier and Venables 
show that differences in Rules of Origin across trade agreements have significant effects on
trade flows. Against this backdrop, the case for focusing export activity only on those
products that can be produced using entirely domestic inputs is weakened – a strategy of
exporting only entirely domestically produced goods is akin to self‐imposing the very same
restrictions that have been shown to restrict exports when imposed by other countries.
Given the growing importance of trade in tasks, it may be sensible to free the focus of
agricultural policy from its role as supplier of inputs to industry.
In sum, ADLI can be seen as a phased development strategy, starting in the agricultural
sector, which then will offer labour, inputs and a source for demand for the non‐agricultural
sector. Growth in the agricultural sector is expected to be driven by yield (output per
hectare) increases, stemming from technology first, and then further infrastructure,
irrigation and other modern input increases. In what follows we will discuss the theoretical
and empirical basis of each of these key aspects in more detail: how growth in agriculture
comes about and how growth in agriculture and the non‐agricultural sector can relate.
Planning for strategic support in a dynamic context
In reconsidering both the role of agriculture in Ethiopian growth, and the role of policy in
improving this performance, a conceptual distinction appears. There is an important
difference between the drivers of growth, on the one hand, and the optimal government
decisions to invest in one sector or another. Many analysts of growth prospects have
concentrated on the growth multipliers of narrow, single‐sector growth, to argue that a
given growth rate in one sector has the greatest impacts on the growth rate of the economy
as a whole. But this does not imply that this is the best area for government intervention:
policy intervention should be guided by comparison of the returns – at the margin – to
transferring an economy’s physical resources across sectors.
This approach represents a departure from traditional use of CGE models to guide policy.
Consider for example a recent study by Diao et al. , who present a pair of economy‐wide
models for Ethiopia, based on parameterizations of input‐output coefficients, demand
elasticities, and other estimated parameters of the economy. These models are used both
to explore the growth impacts of a given, exogenous rate of technological improvement in
one sector or another, and to compare growth rates required in specific subsectors in order
to achieve a given target rate of growth for the economy as a whole. However, these
models do not set out to consider the causes of growth in these agricultural subsectors; nor
(since they generally make simplifying assumptions about the elasticity of supply from
industry or other sectors) do they emphasize the tradeoffs in resource allocation and the
complementarities in demand and production implied.
The causes of agricultural productivity growth cannot be divorced from the study of growth
outcomes and linkages – fundamentally, incentives to innovate in any sector depend on
relative prices, themselves a function of growth rates elsewhere in the economy. Moreover,
and equally important as a guide to policy, it is important in a world of finite budgets to
consider the tradeoffs made in allocating government investments across sectors. The
relevant question is not one of which combinations of sectoral growth rates suffice to
produce a given target growth rate. Rather it should be the question of which allocation of
the physical and human resources available – both in the private sector and in the
government ‐ produces the highest possible growth rate.2 The answer to this question will
depend on equating marginal growth impacts across potential investments, for which it is
necessary to understand the multipliers from policies to single‐sector growth rates.
The choice to allocate public resources across sectors involves several complications over
and above the problem of choosing policies to achieve growth within a single sector. For
instance, the choice faced by policymakers is not simply one of choosing between
investments that have direct impacts on a single industry: indirect effects on other
industries, operating through consumption, production, and investment decisions, are
equally important. Second, there may be important complementarities between alternative
forms of investment – so that gains are only achieved by policies that are well coordinated,
both in terms of sectors, geography, and sequencing. Third, whereas equity considerations
will tend to promote balanced total investments across geographic space, in an environment
of increasing returns it is not necessarily the case that this will lead to balance in sectorspecific
spending across space. Instead, equitable growth may best be achieved by focusing,
e.g., on education in one location and on infrastructure in another. Finally, public
investments may in general play one or both of two functions: to foster growth in a specific
sector, or to strengthen growth linkages across sectors. The second‐order effects of
strengthening linkages may be equally important to overall growth prospects; such public
goods may also be particularly poorly served by the market.
While most analysts have taken these linkages, or multipliers, as given features of the
economy, it may be the case that important growth implications of particular types of
investment – transportation infrastructure being perhaps the most obvious example –
function to strengthen linkages across sectors. This point will be explored in greater detail
The remainder of this paper proceeds as follows. Section 2 discusses routes to growth in
agricultural output. A key contribution of the theoretical framework considered here is to
recognize the endogeneity of agricultural growth: alternative routes to growth in the
agricultural sector, broadly defined, may result from different incentives in the economy.
Not all of these are equal in their implications for long‐run growth within the sector, or in
their implications for growth multipliers to other sectors. Section 3 takes up this issue,
investigating the relationship between growth in agriculture and that in other sectors. This
framework has the broad implication that, while agricultural growth must be a part of any
long‐term strategy for growth in Ethiopia, a balance may be struck between the desire to
increase agricultural production and to increase agricultural earnings and wages; moreover,
given the need to harness market forces to incentivize agricultural growth, there is an
important role to be played, both by policies that link agricultural output to domestic and
international markets, and by policies that grow urban incomes, which will play an
increasingly important role as a source of demand for agricultural goods.
It should be clear by now that simple viewpoints on the role of agriculture in development
are then based on fallacies, misunderstanding essential economic interlinkages. For example,
an ‘agriculture first’ viewpoint on growth priorities cannot be simply justified by stating that
most of the labour force is active in the agricultural sector at this moment or because the
share of agriculture in GDP is high. Even more sophisticated analyses rarely take into account
the economic incentives required to induce growth in agricultural sectors. That agriculture
should be the main driver of growth and that it should be the focal point for government
investment should be proven, not stated. Such a proof requires us to examine the means by
which agricultural growth is sustained as well as its ability to stimulate growth in other
2. Sources of growth in smallholder agriculture
Central to the role of agricultural growth in Ethiopia is an understanding of the mechanisms
by which the agricultural sector itself can grow. This is a partial‐equilibrium exercise, taking
such factors as prices and demand in other sectors as exogenously given, in order to
understand the relationship between the general economic context and the progress of the
agricultural sector. Taking agricultural growth to be endogenous – and understanding how 7
incentives in other sectors drive the direction of agricultural growth – is fundamental: these
incentives provide the multipliers that sustain impacts of well designed policies over time.
Within smallholder agriculture, we focus on two sources of growing productivity: adoption
of new technologies – be these inputs or crops – by smallholders, and increasing of the landlabour
ratio among smallholders.3 The economic conditions associated with each are
discussed in turn below.
Changes in technologies and changes in factor ratios can both bring about productivity gains,
in the sense of increasing the returns to land or labour. But these sources of growth do so in
distinct ways. Technological change can increase the productivity of both factors, though in
practice it will of necessity be biased in one direction, changing the ratio between returns at
the margin to one factor or the other. Large‐scale mechanization, for example, may
decrease the marginal return to labour relative to capital and land inputs. By contrast,
adoption of high‐value crops, or of inputs such as fertilizer that require intensive cultivation
techniques, may be seen as labour‐augmenting. Like technology, changes in factor ratios
alter the relative returns to land and labour. An increase in the land‐labour ratio in
agriculture implies rising (average and marginal) productivity of labour, but falling land
productivity (assuming other inputs are held constant). Such changes in factor inputs have
characterized a perhaps surprising share of recent agricultural changes in Ethiopia . The
Government of Ethiopia has long recognized the importance of the land‐labour ratio to the
growth of rural incomes in the face of mounting population pressure; in ADLI , for
example, it is argued that the expansion of non‐farm activities should be promoted “… there
by [sic] increasing holding sizes for rural families.” [p. 19]
This distinction will be important in considering the form of stimulus that agriculture will
provide to other sectors. These will be discussed in detail in Section 3, but the importance of
labour‐versus‐land bias in technological change is briefly foreshadowed here. When labour
markets are competitive, increases in labour productivity represent growth in and of
themselves: they result in higher wages in both rural and urban sectors.4 Increases in the
productivity of land that are not labour‐augmenting, such as mechanization, do not have the
same effect on wages across the economy. Indeed, this seems to have been the case in the
Philippines. Describing the process of rapid agricultural growth in that country, Bautista 
writes that “[a]gricultural labour has not been helped by the substantial mechanization of
some farm operations, particularly the preparation of land for rice and threshing of the crop”
(p. 126). However, a greater supply of agricultural output may itself have effects on the
growth potential of other sectors by maintaining favorable domestic terms of trade that 8
preserve high returns to capital, which are subsequently reinvested to create industrial
growth. This effect may obtain even if rural wages do not rise (for example, if the increased
agricultural revenue accrues to owners of capital or land in a mechanized agricultural sector).
Given the focus of this paper on the role of smallholder agriculture in Ethiopian growth, we
will not discuss in any detail the determinants of large‐scale, commercial farming. Here we
will simply reiterate the basic point that the growth framework must not take the expansion
of commercial farming as exogenous, just as it must seek to understand the determinants of
smallholder productivity growth. At times, similar forces may drive the two: growth of
commercial farms will also require appropriate and sustained price incentives, for example.
Recent emerging successes in Ethiopia and elsewhere do permit a few suggestive
generalizations, however. Commercial farming can employ technologies that achieve
economies of scale not accessible with smallholder technologies; they may also be profitable
under some agronomic and environmental conditions that characterize parts of Ethiopia
where smallholder agriculture has not been viable. For some crops, well integrated
outgrower schemes may provide a means to increasing the impact of high‐value crop
production for rural incomes. As a source of increased cereals output, it is important to
recognize the limitations of commercial agriculture – which is typically characterized in this
case as a land (but not labour)‐augmenting technological improvement. Commercial cereals
production may not be a large source of employment, and as such may not have large
impacts on rural wages; however, as discussed in Section 3 the increased supply of food may
be crucial to the growth of other sectors.
Adoption of new technologies and crops by smallholders
Drawing on Green Revolution experiences, the standard view of agricultural growth is
founded upon intensive improvements by smallholder farmers. This may take the form of
farmers either adopting new technologies for the production of existing crops, as in the
Green Revolution case, or of farmers adopting new crops altogether (e.g., moving into higher
value‐added crops, such as fruits and vegetables).
What exactly is required for this to take place? We highlight two conditions: the supply of
appropriate technologies, and the magnitude of economic returns to adoption by
First, as the experience of the Green Revolution showed, the availability of agricultural
innovations plays an important role. The obvious first requirement is that scientific work to
develop agricultural innovations takes place. As Kei Otsuka  has emphasized, the
availability of innovations requires more: not only agricultural innovations at the level of
global public goods, but also the adaptation of the resulting modern varieties (MVs) to local
growing conditions. In the case of Vietnam, Otsuka argues that the “serious efforts to
nurture… researchers at the national programs” of IRRI and CIMMYT played an important
role in translating global innovations into locally profitable technologies. Similarly, while the
orthodox story of England’s own agricultural revolution typically emphasizes the role played
by enclosure, an important role was played by the development of new systems of crop
rotation and other techniques (see, e.g., Timmer  for an early discussion of the historical
Beyond the research stage, the effective supply of agricultural innovations to smallholder
farmers clearly requires both a network of extension officers and an efficient distribution
network, which reliably delivers desired seeds to farmers.
Of note here is the importance of simultaneous improvement in the distribution systems
across seeds, fertilizers, and other inputs. Evidence on the returns to improved seed
adoption in Africa (see Otsuka and Kalirajan [39, 40] for a helpful overview, and see Dercon
and Hill  for recent evidence) suggests that these returns are significantly higher when
farmers also adopt fertilizer. For example, Kei Otsuka and Takashi Yamano  find that in
Kenya, maize yields on high yielding variety (HYV) areas are 24% higher when farmers use
typical levels of organic fertilizer than when they do not, and 34% higher when farmers use
typical levels of chemical fertilizer than when they do not. The economic returns to sustain
improved seed adoption may be present when a broad set of complementary inputs are
available, but may not be obtained by farmers in the absence of efficient and reliable
delivery channels for fertilizer and other inputs.
This first requirement, the availability of inputs, is hardly controversial. A second
requirement is often overlooked: the first stages of agricultural innovation in a population
require strong economic incentives. The returns to adoption of new technologies and crops
must be high; the returns required to induce adoption are higher than those required in
order to sustain use of new technologies once they have secured a firm foothold in the
population. Input prices, fertilizer requirements, uncertainty about production techniques:
all of these factors combine to make adoption of new agricultural technologies a costly and
risky proposition, particularly for early adopters.
There are various ways of creating these high returns, most notably high yields, high output
prices or low input prices. The Green Revolution in India and Southeast Asia appears to
deserve its name not least because of the massive yield gains that were on offer by using
new varieties. Wheat technologies developed at CIMMYT, with suitable complementary
inputs, offered a more than fivefold increase in yields on small farmer demonstration plots
in India . Rice yields on offer via IR8 rice were early on estimated to be 5 times the
traditional variety using traditional methods and 10 times using optimal input packages .
Often, such high yield gains are not immediately available, or difficult to immediately
replicate on farms. Accordingly, periods of widespread improvement in smallholder yields
are often led by high prices, which generate the required high returns to adoption.
Historical examples are legion: Robert Allen  argues that the expansion of Paris in the
nineteenth century led to intensification of production techniques in its surrounds, and,
similarly, that the growth of London caused enclosure of grain lands and their conversion to
pasture in its surrounds. Turning to modern cases, China’s agricultural growth since 1985
was precipitated by measures that offered a strong increase in returns on offer to farmers
via (i) the liberalization of agricultural prices, resulting in a significant increase at the outset
of the economic reform process, and (ii) the freeing of households to specialize in crops with
high returns . In Vietnam, the highly fertile and widely irrigated land created agronomic
conditions conducive to large returns at the outset of that country’s green revolution .
In the Philippines, Hayami  shows the green revolution from the 1960s, linked to the
high yield rice varieties gradually resulted in rice price declines, resulting in only slow
agricultural income increases, but started at a relatively high level of prices.
A further caveat is important to understanding the relationship between returns and
adoption: the returns required to induce early adopters may have to be higher than the
returns required to sustain a pattern of aggregate productivity growth in agriculture. The
aggregate process of adoption itself is characterized by increasing returns. This is true for
Importantly, farmers learn from one another, leading to well established S‐shaped curves
characterizing adoption. These have been documented since the seminal work of Zvi
Griliches  on the adoption of hybrid corn in the USA in the first half of the twentieth
century. Munshi’s recent empirical work on the green revolution in India  and has
affirmed the importance of learning externalities in the adoption of HYVs in that case,
although this work also suggests that cross‐farmer learning varies by crop. Munshi argues
that rice‐growing conditions in India were more variable across individuals than were wheatgrowing conditions, making it more difficult to learn from the experiences of others in thecultivation of this crop. This provides a useful lesson for the diffusion of improved varieties
and new crops in Ethiopia: the more variable are growing conditions for these technologies,
the higher the sustained returns required in order to further their diffusion.
In light of the value of learning from neighbours’ experiences, some farmers may
strategically delay the process of adoption in order to benefit from watching those around
them . This further explains why early adoption is more difficult to induce than late, as at
later stages of adoption the strategic incentive for delay is reduced.
Even in the absence of learning externalities across farmers, it may still be true that farmers
– once they have adopted an individual technological improvement – find it easier to
continue to adopt newly improved varieties as these become available. In Vietnam, for
example, aggregate productivity gains in rice were sustained over a twenty‐year period. The
speed of diffusion suggests that this sustained growth was not attributable to the steady
process of bringing new farmers ‘into the fold’. Instead, the consensus  for that country
is that its green revolution was brought about by a series of improved technologies,
introduced in succession. Fixed costs in learning new production techniques may make it
easier for farmers – having undertaken one form of agricultural innovation – to progress to
Taking these three broad considerations into account, our partial‐equilibrium story of
agricultural growth is thus: at early stages, there are two key constraints to technology
adoption, in the form of the availability of appropriate technologies on the ground and
appropriately high returns to the adoption of this technology. Once initial adopters have
been induced, then cross‐farmer learning about techniques and profitability, within‐farmer
increasing returns to innovation, and alleviation of strategic adoption considerations all relax
the level of returns required to sustain adoption. Dercon and Hill , the companion piece
on sources of agricultural growth in Ethiopia, show that despite some rhetoric to the
contrary, the availability of appropriate and high return technologies on the ground is
limited at present in Ethiopia, affecting the scope for fast agricultural growth via adoption,
although current high cereal prices should offer a window of opportunity for adoption
Increases in land labour ratio among smallholder farmers
A second route to rising labour productivity is through an increase in cultivated land per
worker in the smallholder agricultural sector. By definition, this can come about either
because of an expansion in cultivated land or because of a reduction in the agricultural
labour force. While many zones will have only limited scope for area expansion, it is possible
that rising labour productivity may come about in this sector nonetheless, as non‐farm
economic activity draws labour out of the agricultural sector.
Adenew , Seyoum Taffesse  and Dercon and Hill  critically address the extent to
which expansion in cultivated area is responsible for recent improvements in agricultural
output in Ethiopia. This evidence suggests that a considerable share of recent cereal
production growth has come from area expansion, even though this remains contentious.
Furthermore, the area expansion has not led to larger farms per worker compared to 10
years ago, as during the same period the rural population continued to grow considerable.5
Finally, even if output growth was largely driven by area expansion, however, resource
constraints imply that the continued expansion of cultivated area is not a route to sustained
growth in agricultural labour productivity.
In the absence of area expansion, rising land‐labour ratios may be the result of the transfer
of labour to other sectors. Some of this may occur through rural non‐farm activities; as
studied by Loening et al.  and discussed by Rijkers et al. , rural non‐farm activities are
integral to the Government’s Plan for Accelerated and Sustained Development to End
Poverty (PASDEP) .
Such a reduction in the agricultural labour force can indeed be associated with increased
agricultural incomes per capita. Allen [2, 3] has shown the effect of the removal of labour on
agricultural productivity in England. This is also shown by historical evidence in Italy ,
where the impact of the Black Death in the middle of the 14th century is associated with a
rise in labour productivity in agriculture.
Growth implications of a rising land‐labour ratio
Either of these mechanisms to increasing the land‐labour ratio will raise the average product
of labour. In the case of land expansion, total output will increase, whereas in the case of a
reallocation of land resulting from out‐migration, the effect on total output is ambiguous, as
will be explained below.
In general, the marginal product of labour will rise in both cases. As will be discussed in
Section 3, when labour markets are competitive across urban and rural areas, such an
increase in the marginal product of labour will result in rising real wages across the
economy; this in and of itself can be a source of growth.
In the case of increasing the land‐labour ratio through labour transfer, an important
question is the implication for total agricultural output. Declines in total agricultural output
can, as discussed in Section 3, retard growth in other sectors by driving up wages paid there.
Extensive debates of the 1960s and 1970s regarding the existence of pure surplus labour –
the extreme case in which the marginal product of labour is zero over some range ‐ have
largely been regarded as a “red herring” . If economies of scale are sufficiently powerful,
it is even possible the consolidation of small farms can generate a positive net effect on total
agricultural output when labour moves out of that sector. But what is important, and
broadly accepted as a characterization of labour markets in many developing countries, is
that the marginal product of labour in smallholder agriculture lies below its opportunity cost
in other sectors. Factors including the availability of risk‐coping mechanisms create
incentives for household members to remain in that sector, even when a transfer of labour
would increase aggregate output in the economy. Irreversibility in the decision to relocate
to urban areas – which may be brought about, for example, by imperfections in the market
for land – will also exert a pull, keeping labour in rural areas.
In an economy that is not completely open, a decline in total agricultural output will tend to
turn terms of trade against other sectors. This results in rising nominal wages in, e.g., the
industrial sector; even if the economy is closed, this decreases (accounting) profits, while if
industry competes on international markets then its competitiveness will be hurt. Either
mechanism has the potential to retard growth of non‐agricultural sectors.
For this reason it is important to understand whether and to what extent the withdrawal of
labour will reduce total agricultural output. If there are economies of scale in smallholder
farms, it may in fact be the case that these will offset the reduction in labour supply – so that
total output can increase as a consequence of labour withdrawal. As Dercon and Hill 
shows, there is inconclusive evidence for the presence of such economies of scale in
Ethiopia; despite a few papers showing their presence, there is no convincing evidence for
scale economies. 6 If anything, this same evidence suggests that scale economies are
stronger in food deficit areas, which would support a strategy of supporting off‐farm
activities in these areas.
What factors bring about an increase in the land‐labour ratio?
As already noted, there has been an increase in cultivated land over the past decade.
However, the potential for continued area expansion as a source of growth is limited.
Incentives for a shift in labour from agricultural to non‐agricultural occupations, on the other
hand, may be harnessed as a means to increasing agricultural incomes among those who
remain in the sector.
Returns to non‐farm activities – both rural non‐farm activities and urban industrial and
service employment – will naturally increase over the course of the development process, as
domestic demand for non‐agricultural goods rises. Policies that can accelerate this process
include investment in education: for example, Mogues et al.  find substantial
consumption increases to be correlated with such public investments, although these results
should be treated with caution given the challenges inherent in drawing causal inference
from correlations in this context. Given the importance of informal rural insurance and
other safety nets particular to rural settings [12, 26], the migration of labour to urban areas
may be incentivized by government policies that strengthen safety nets in urban areas.
There may be a role for land policy here: since land is among the most important assets for
income security, measures that protect migrants’ claims to land may reduce the risks
inherent in rural‐urban migration.
3. Agricultural growth in an economy wide context
Agricultural growth does not occur either in isolation or exogenously. Multipliers from
public investments in agriculture will depend not only on their ability to generate growth
within that sector, but also on linkages to other sectors. Moreover, given the scale of the
Ethiopian economy and given finite government resources, policies promoting growth
should seek to harness market forces that can provide a continued source of incentives for
innovation. In the tradition of endogenous growth models, one may view the highest
thereby speeding the arrival of the point at which this marginal product exceeds the
institutional wage paid to surplus labourers, and real wages can rise. But in early stages of
development, it will be only improvements in land productivity (or, equivalently, the average
product of labour) that matter, given the surplus labour assumption.
By contrast, in a neoclassical dual economy, the surplus labour condition does not hold.
Workers in the agricultural sector have positive marginal products. Moreover, labour
markets are assumed to be competitive in the neoclassical model. Under this condition, this
marginal product of labour in agriculture determines incomes. With agricultural incomes
determined this way, increases in the marginal product of labour in either the urban or rural
sector plays an immediate role in the growth of nominal wages. The importance of the
marginal product of labour as a source of real income growth is a key distinction of the
The way in which this translates into growth in real wages depends on income levels.
Neoclassical models such as those of Jorgenson [27, 28] and Eswaran and Kotwal  have
typically assumed that workers have hierarchical preferences: below some threshold level of
consumption, they are unwilling to substitute industrial output for agricultural output, that
is, to “eat shirts”. Consequently, workers do not benefit from increases in labour
productivity in the industrial sector, which merely increase the supply of a good that they do
not consume. On the other hand, advances in agricultural productivity are required to
release labour into other sectors, and these productivity advances – by increasing the supply
of food – are a source of real wage growth at low levels of income .
Below we examine the empirical relevance of the surplus labour assumption. In Section 0
we re‐examine the role of agriculture in an open economy, and consider evidence of the
openness of Ethiopia’s cereals market. In Section 0 we also consider the implication of
Hirschman’s  notion of linkages in this context. Such linkages are often overlooked in
models of dual economies.
As highlighted above, the contrast between classical and neoclassical models of a dual
economy highlights the importance of the question of surplus labour in determining
agriculture’s contribution to growth. If labour markets are competitive, then increases in
rural incomes can translate into real wage growth in urban areas. Otherwise this will not
occur until a combination of labour transfer to other sectors and agricultural technical
change increase the marginal product of labour above current income levels.
If there is a surplus of rural labour, growth in other sectors may be accelerated by increasing
the marketed surplus of agricultural output. If this marketed surplus increases less than
proportionally with the growing urban labour force, relative food prices will rise, and urban
growth with slow. Two key reasons explain the failure of the marketed agricultural surplus
to rise in this way. First, if the marginal product of labour(ers) in agriculture is not zero, then
growth; the figure plots a regression line and 95% confidence interval for this relationship
among regions outside of Addis.
In regions where there has been significant transfer of labour from rural to urban areas, food
prices have risen more sharply than national trends. This is consistent with the view that (a)
the withdrawal of labour has had a non‐negligible impact on the marketed surplus from
agriculture, and (b) urban consumers are unwilling to substitute the consumption of services
or industrial goods for agricultural output.
total output of that sector will decline with the migration of the rural workforce. Second, if
rural smallholders choose to consume rather than to sell the increment in the surplus of
food per capita that accrues when family or neighbors migrate, then this absorption by the
rural sector will decrease the supply of food per capita to urban areas.7
Below we present evidence from Ethiopia. We argue that, while it is possible to increase the
allocation of labour to rural, non‐farm activities without diminishing agricultural output, the
movement of labour to urban economic activities does come at a significant cost to
agricultural output (or at least to marketed agricultural output per urban worker). There is
limited evidence on the relationship between agricultural and non‐agricultural incomes in
Ethiopia, but what is available is consistent with the view that labour markets are not
Marginal product of labour
In theory, it is possible that the withdrawal of labourers from the agricultural sector would
not diminish its output. Two mechanisms may support this. On the one hand, an increase in
the labour inputs of those who remain behind may compensate for lost labourers – this is
the argument first put forward by Sen . Alternatively, consolidation of rural landholdings
may result in economies of scale that have a net positive impact on production. This
possibility is discussed further by Dercon and Hill , but as reported above, the evidence is
at best mixed.
Evidence suggests that marginal product of labourers is not zero in Ethiopia. Rural‐urban
migration, in the absence of technological progress in agriculture, will result in a decrease in
total agricultural output. The evidence examined below is consistent with either of the two
mechanisms suggested above: that agricultural production declines in response to the
transfer of labour, or that the marketed surplus simply fails to rise in proportion with the
growing urban labour force. However, significant transfer of labour to the rural non‐farm
sector appears to be possible without decreasing the effective supply of labour to
Competitiveness of labour markets
The potential for increases in the marginal product of labour to be a source of real wage
growth in other sectors depends on the competitiveness of labour markets. This
competitiveness may potentially operate differently between rural farm and non‐farm
labour markets, on the one hand, and between the rural and urban labour markets more
generally, on the other. Rijkers, Söderbom and Teal  shed some light on the tightness of
the rural (farm and non‐farm) labour market. They argue that seasonal participation in nonfarm
enterprises is common. Consequently, these authors find no strong evidence of a
relationship between wages in the rural farm and non‐farm sectors. Given their different
seasonal demands, the rural farm and non‐farm sectors do not compete for the same
effective pool of labour: these labour markets are segmented by time. So long as this is the
case, increases in the marginal product of labour in one type of rural activity need not
translate into real income gains in the other.
If the ability of labour to move flexibly from the urban (non‐farm) sector to rural sectors is
not fully constrained, then the labour‐market equilibrium may create a competitive link
between sectors. Barriers to labour mobility may break this link; e.g., Deininger and Jin
provide some evidence of a relationship between migration and land rental markets in China
. Greater costs involved in rural‐urban migration suggest that expansion of urban
employment will draw meaningful amounts of labour away from rural farm and non‐farm
enterprises – and thereby will decrease total agricultural output, at least in the absence of
substantial technological change.
Agricultural growth increases the supply of educated labour to other sectors
Finally on the supply of labour, there may be an important and often‐overlooked
relationship between agricultural productivity growth and the supply of semi‐skilled labour
to industry. If part of the increased agricultural revenues are invested in children’s
education – whether this investment is undertaken by households or by the government –
then this will ultimately increase the supply of skilled labour to other sectors. Mao and
Schive  argue that this was important to the absorption of Taiwan’s immigrants in nonfarm
employment during that country’s agricultural boom (a boom in which migration from
rural to urban areas grew from 19,000 individuals per year during the period 1952 – 1964 to
more than 93,000 individuals per year in the period 1963 – 1973). See Estudillo and Otsuka
 and Hayami  for related evidence from the Philippines.
Conclusion: classical domestic markets for food and labour
Taking this evidence together, we suggest that much of the evidence is consistent with the
characterization of the Ethiopian economy as a classical, rather than neoclassical, dual
economy. High transaction costs involved in migration to urban centres – including limited
rural land markets – allow a disconnect between wages in these sectors. The household
basis of smallholder production means that the incomes of smallholder farmers are linked to
the average product of their labour (since the returns to land typically accrue to the same
individuals); consequently, labour markets will not equate marginal products of labour
across sectors as in the competitive, neoclassical model. The empirical relationship
between labour transfer from rural to urban areas and local food price inflation is also
consistent with the importance of total agricultural output as a determinant of urban wages.
In this case, technological change in agriculture plays two roles. In the short term, its most
important contribution to growth is through an increase in total food supply. This helps to
keep down cost of living in other sectors, allowing greater capital accumulation in industry
even when its goods are produced for the domestic market alone, and maintaining
competitiveness if industrial output is exported. Second, agricultural technological progress,
by increasing the marginal product of labour, helps to accelerate the arrival of what Ranis
and Fei  call the ‘commercialization point’: the point at which this marginal product
exceeds the institutionally determined (i.e., not competitive) rural wage. It is only from this
point, at which the labour surplus is exhausted, that gains in productivity in any sector
translate into real wage growth.
The supply of educated labour to industry provides an important caveat to this argument
about domestic markets for labour and for agricultural products. First, the market for skilled
labour may be considerably more competitive, given that educated workers may find it
easier to migrate in search of employment. Second, and perhaps more importantly, since the
proceeds from agricultural productivity gains are partially invested in human capital,
progress in agricultural technology may further contribute to growth through the supply of
It is appropriate to qualify these arguments with poverty reduction in mind. In Ethiopia,
growth is obviously needed for sustainable poverty reduction, but the nature of the growth
process will matter for poverty reduction. While we may need agricultural output growth as
part of a balanced growth strategy to keep the growth of nominal industrial wages in check,
poverty reduction requires real wage growth. The extent to which agricultural growth and
labour migration involves growth in the marginal product of labour is then relevant.
Furthermore, to ensure more poverty reduction for the next generation, education, health
or other means of increasing the skills of the labour force will contribute to both growth and
poverty reduction, as it will reduce the segmentation of labour markets and increase
migration from rural to urban areas.
Open economy considerations
Openness to trade changes the relationship between agricultural productivity and growth in
both the classical and neoclassical models. In the classical dual economy model of a closed
economy, increases in agricultural (land) productivity were important in order to stave off
deterioration in domestic terms of trade. When the classical dual economy becomes an
open one, then relative prices of agricultural and industrial goods are pinned down by
international markets. In a perfectly open economy, if rural‐urban migration brings about a
decrease in the marketed surplus available per non‐agricultural worker, this will no longer
result in rising relative prices for food.
This reduces the importance (from a growth perspective) of technical changes in agriculture,
such as mechanization, that increase land productivity and the average product of labour
but have a limited effect on the marginal product of labour. On the other hand, this
decoupling of food prices from domestic supply frees smallholder agriculture to act as a
source of real wage growth in and of itself. Labour‐intensive technological changes, which
may include the adoption of fertilizers and shifts to high value added crops, can speed the
growth in labour’s marginal product, which will ultimately drive real wage growth in
agriculture and beyond.
A priori, there is no reason why Ethiopian cereal markets should not be integrated with
world markets, so an open dual economy model could offer guidance. However, this ignores
a key feature of Ethiopia: that it is landlocked with a relatively large distance from the main
cereal and demand areas from the nearest port. For example, until the last few years, maize
prices in Addis Ababa were fluctuating around $15 per quintal for many years, with transport
and handling costs to Djibouti near $6 per quintal or 40% of the mean market price. The
result is that even if there were no trade distortions, for a wide range of domestic market
conditions, prices would move independently from world prices. As a result, within some
broad bounds, cereals are best considered non‐tradable goods, suggesting that the
discussion in the previous section, on the importance of domestic food production to food
price inflation, remains relevant.
The evidence supports this. As argued by Rashid et al. , domestic cereals in Ethiopia are
generally considered to be a non‐tradable sector. Substantial within‐country variation in
cereal prices, as presented in Figure 1 and in evidence for unexploited spatial arbitrage
opportunities presented by Negassa and Meyers  further supports the view that prices
for cereals, at least outside of Addis Ababa, do not effectively compete with imports.
This does not mean that cereal prices have not exceeded import parity prices.
Rashid et al.  show that, after accounting for price distortions created by food aid,
domestic prices for wheat and maize have on several occasions exceeded import parity
prices since the mid‐1990s. This is definitely the case since September 2007 up to now (see
Dercon and Hill ). Trade barriers, market imperfections, macroeconomic and foreign
exchange constraints are responsible for these deviations (as food imports should have been
able to keep prices on parity prices). It also shows that the status of cereals as de‐facto nontradable
commodities can be changed via a review of policy parameters, and trade barriers.
But as a land‐locked country with considerable transportation costs to the nearest harbours,
food prices will be largely domestically determined and the analysis in the previous section
will remain valid.
Production and consumption linkages
Traditional dual economy models focus on endowments of land, labour, and capital as
determinants of long‐run growth. They often overlook static multipliers created by linkages
between the production of intermediate goods and the market for the consumption of
domestic output. Their focus is typically on the supply side. However, the linkages
highlighted by Hirschman  may also play a role in determining agriculture’s contribution
to growth. Modern authors, while departing at times from Hirschman’s conclusions (see, for
example, Mellor ), have focused on two broad types of linkages from this sector to
• production linkages, whereby increases in agricultural output lead to greater supply
of inputs used in non‐agricultural production or, conversely, greater demand for
non‐agricultural outputs used in agricultural production;
• consumption linkages, whereby increases in agricultural income lead to increased
demand for non‐agricultural (as well as agricultural) consumption goods.
Much of the debate on the role of agriculture has focused on the magnitude of these
linkages relative to parallel linkages stemming from growth in other sectors. On one side,
Hirschman  has argued that weak backward production linkages and limited forward
production linkages in agriculture strengthen the argument for growth led by
manufacturing: according to Hirschman “the superiority of manufacturing in this respect is
crushing” (p. 110). On the other hand, proponents of agriculture‐led growth have
responded by emphasizing the importance of consumption linkages; for example, Mellor
(, p. 13) argues that “it is consumption linkages that are most important. … [rural
people] will spend at least 70 percent of their incremental income on consumption goods.”
There is no reason to presume a universal answer to this debate. The strength of production
and consumption linkages will depend on many country‐specific factors. Some of these may
be taken as given – natural resource endowments, for example – while the strength of other
forms of linkages may themselves be amenable to policy intervention. The latter is an often
overlooked consideration in policy evaluation: not all forms of agricultural growth will have
the same multiplier effects, and these multipliers may depend on the way in which
agricultural growth is achieved. Growth attained by investment in roads that increases
returns to agricultural innovation, for example, may also affect the underlying linkages23
between sectors. In light of the country‐ and policy‐specificity of these linkages, we visit
below the theory and evidence for some key forms of linkage in Ethiopia, putting these in an
international context where possible.
Production linkages from agriculture can be ‘forward’ – where agriculture provides inputs
employed in other sectors – or ‘backward’ – where agricultural production creates demand
for the output of other sectors. There is a broad consensus that backward linkages from
agriculture have little power.13 In considering the forward linkages from agricultural
production to other sectors, it is common to focus on the direct use of agricultural inputs as
factors of production in industry. For this reason the ADLI strategy  focuses primarily on
the supply of raw materials from agriculture to industry. It sets out food self‐sufficiency and
the production of raw materials for industry as its first two policy priorities for the
Perhaps the most obvious mechanism by which agriculture can stimulate growth in other
sectors is by providing raw materials or intermediate inputs for industrial production. For
this reason, ADLI  (p. 27) emphasizes improvements in the quality and quantity of hides
and skins and raw materials for the textile industry as priorities for agriculture. A social
accounting matrix (SAM) developed by IFPRI (as presented in Diao et al. , p. 35) shows
the rationale for continuation of these efforts: 50.6% of forestry and fishing output is used as
an intermediate good, with coffee, tea, and chat (33.6%) and livestock (31.9%) not far
behind in terms of their use in other sectors.
However, two cautionary notes are important here. While relatively high shares of domestic
production of these raw materials may be used as inputs elsewhere, this does not
necessarily imply either (a) that these constitute a large share of the overall intermediate
goods requirements of industrial production; or (b) that domestic markets for industrial
goods provide opportunities for further expansion in sectors that rely heavily on local raw
materials. To the extent transport costs allow, export possibilities may help to resolve the
latter difficulty. However, trade and industrial production data suggest that Ethiopia has
made little progress since the early 1990s in reducing its dependency on imported inputs. As
of 2002, imported industrial inputs remain approximately 48% of total input content, almost
exactly their share (49%) in 1993.14 This is only slightly below the 2002 share of imported24
inputs for all of Sub‐Saharan Africa, although the remainder of the region shows a marked
upward trend not present in Ethiopia .
However, a simplified characterization of trade in manufactured goods as representing the
final output of sectors that produce using only domestic inputs seems no longer to fit well
with the global trade patterns. As discussed in Section 0, trade is increasingly done ‘in tasks’,
with countries relying heavily on imported intermediate inputs . Recent work has shown
that, when economies are constrained to export only those products which are composed
entirely of domestically produced inputs, this dramatically reduces export volumes . This
suggests that, under the modern fragmentation of trade, significant opportunities will be
missed by a strategy that does not allow industry to draw upon imported inputs. More
optimistically, this suggests that it is increasingly unnecessary to orient agriculture toward
the production of raw materials for use in industry as a precondition for successful industrial
A central tenet of the argument that agriculture leads development is that consumption
linkages, operating via farmers’ demand for industrial (and service) goods, create strong
linkages from agricultural growth to growth in other sectors. Mellor  epitomizes this
“[I]t is the consumption linkages that are most important. The savings rate of rural people
receiving increments to their incomes tends to be high but rarely exceeds 30 percent. Thus
they will spend at least 70 percent of their incremental income on consumption goods.
Albert O. Hirschman’s widely quoted statement that agriculture has very poor linkages to
other sectors and thus is not a sector to emphasize reflects a common tendency of
development economists to ignore the stimulative role of effective demand for consumption
goods and services.” (p. 13)
Mellor goes on to claim that “the producers of agricultural commodities spend little of that
incremental income on increased consumption of basic food staples.” If true, this suggests
that agricultural growth will tend to stimulate production in other sectors, via its demand
effects. Looked at in a dual economy framework, this could be interpreted as a solution to
Lewis’s deteriorating terms of trade: a small increase in the subsistence incomes of the
farming population would suffice to induce very large Engel effects – which would swing
domestic terms of trade in favor of industry once again.
However, it is important to recognize that the extent of such consumption linkages cannot
be expected to be present uniformly across all countries. Preferences of consumers may be
different. When income elasticities of demand are not constant, the starting incomes of the
agricultural sector will matter as well: marginal propensities to consume agricultural goods
are likely to be higher at lower levels of income. Moreover, transportation costs and the
physical and institutional linkages of markets will impact upon the marginal propensity to
consume out of industrial goods in rural areas: poor access to urban markets will dampen
any potential stimulative effects.
Evidence for Ethiopia suggests that consumption linkages are not particularly strong.
Estimates presented by Diao et al.  provide average and marginal budget shares for a
range of goods separately by occupational category.15 They find that the marginal budget
share of cereals is 19.2% for farmers, as compared to 4.8% for wage earners and 2.8% for
self‐employed ‘entrepreneurs’. Marginal budget shares for other types of food are similarly
high among farm households, who allocate 21.6% of their budget at the margin to such
expenditures. And farm households allocate only 22% of their marginal income to industrial
goods, as compared with 40.3% allocated at the margin by wage earners and 26.5% by the
self‐employed. Marginal expenditure on services is reasonably strong, at 37.4%, which
compares favorably with the 31.5% of marginal expenditure spent on services by wage
earners, but unfavorably with the 56.3% of marginal expenditure spent on services by the
self‐employed. Thus the effects on demand for industrial output are considerably lower for
growth in agricultural productivity (leading to increased agricultural incomes) than they are
for growth in urban incomes.
Perhaps surprisingly, marginal expenditure on industrial goods compares unfavorably, for
example, to the 37% marginal budget share spent on industrial goods by farm households in
Zambia, as estimated by Delgado et al. . So the question becomes, why are
consumption linkages relatively weak in Ethiopia? Each of the reasons suggested above
plays a role. The work by Delgado et al. provides a further clue to the importance of poverty
in limiting consumption linkages: for example, they find that marginal budget shares for
tradables rise from 24%, in the bottom decile of per capita expenditure, to nearly 40% in the
top decile of per capita expenditure. Thus it is reasonable to suggest that low levels of
income create relatively low marginal propensities to consume out of industry. As we have
argued the observed consumption linkages are not exogenously determined, but given
current conditions—in particular, the poor transportation infrastructure and relatively
isolated urban markets—the evidence suggests that agriculture in Ethiopia cannot at current
levels of income be expected to play as full a role in stimulating other sectors as it would in
4. Conclusion: agriculture in relation to other drivers of growth
Ethiopia has been following a particular vision on the role of growth in smallholder
agriculture, as reflected in ADLI: agriculture should be seen as the sector from which growth
should emanate, and growth in agriculture will lead to industrial growth, by offering labour,
inputs and demand for its products. In this paper, we have applied a ‘dual economy’
perspective to understanding the role of smallholder agriculture in Ethiopian growth,
reflecting on the foundations of the ADLI strategy. This framework starts from two premises.
First, agricultural innovation is not an exogenous process, but rather must be sustained by
virtue of its interaction with other sectors. Second, the contribution of agriculture to growth26
depends on the extent of surplus labour, of open markets for agricultural goods, and of
linkages (particularly consumption) between sectors.
Our discussion has led to a number of key insights relevant for understanding the scope and
constraints for growth in Ethiopia. First, the premise of endogenous agricultural innovation
requires us to consider sources of incentives for agricultural growth. Given government
budgetary limitations, significant agricultural productivity gains at an economy‐wide scale
will require high and sustained economic incentives. It requires high return technologies to
be available and to offer high economic returns, especially through the earlier stages of
adoption. Such incentives can be brought about in one of two ways. One approach is to
stimulate growth in the industry and service sectors, creating high and sustained domestic
demand for agricultural produce.
Another way to sustain incentives could be via international markets. High and, over the
long term, rising international food prices suggest that the latter may provide a source of
stimulus for domestic innovation, giving strong incentives to boost output when imports are
still required, while offering high export prices when harvests are successful. However,
Ethiopia’s scope to fully exploit the possibilities from international markets is limited by
being a landlocked economy, with high transport and handling costs to the borders. The
result is that for a wide range of prices, cereals are behaving as non‐tradable commodity,
with domestic demand and supply governing prices. Therefore, to fully take advantage of
the incentives offered by high international prices efforts should be made to reduce
transportation and other marketing costs in order to reduce import parity prices and
increase export parity prices, allowing farmer to receive higher shares of world prices, while
providing downward pressure on consumer prices.
Secondly, the evidence presented suggests that a classical model with largely closed markets
for cereals remains appropriate to the Ethiopian economy. Limited responsiveness in wages
to changes in other sectors is consistent with the existence of surplus labour in agriculture. A
positive marginal propensity to consume agricultural goods in rural households means that
rural‐urban migration can bring about sharp increases in food prices, even if the marginal
product of labour in agriculture is low. The available evidence suggests that urban
consumers’ unwillingness to substitute between consumption of agricultural and other
goods contributes further to high food prices – filtering through to high costs of labour – in
growing industrial areas.16 Given the high costs of import, not least in the current climate of
high international food prices, the total output of the agricultural sector remains then a
constraint to the growth of domestically oriented industry (through its effect on returns to
capital) and the competitiveness of export‐oriented industry.
Any means of boosting agricultural output remains therefore important during the process
of economic transformation with labour moving out of agriculture to urban areas. This
suggests that technologies enhancing land productivity have an important role to play in
maintaining low urban costs of living and facilitating growth in other sectors. This also
suggests a potential contribution of large‐scale, commercial farming to the growth of other
sectors through the supply of cereals. Either way, the impact on the marginal value product
of labour in smallholder agriculture would be limited, offering limited scope for real wage
and earnings growth in the rural sector.
Provided import parity prices can be kept sufficiently low (via better integration with
international markets and reduced transportation and marketing costs), an alternative
approach would rely more heavily on food imports. If feasible, this would allow smallholder
agriculture to focus on high‐value and exportable crops, utilizing economies of scale both
due to agglomeration (perhaps coordinated through outgrower schemes) and through the
marketing chain. Such technological changes would increase the marginal product of labour,
and so would accelerate the commercialization of agriculture, from which point further
technological improvements would provide real wage growth. While this may not be on the
cards immediately, it offers a route to allow agriculture to graduate from a focus on food
output to a diversified agriculture offering real wage growth and therefore poverty
reduction in the rural sector.
Thirdly, modest consumption linkages suggest that agricultural innovation will not be
sufficient to drive growth in industry. A growing consensus that agglomeration externalities
and other scale economies play an important role in industrial development suggests that it
will be difficult for industries that serve only the limited domestic market to compete with
foreign firms. More positively, the increasing frequency of trade ‘in tasks’ suggests that it
may not be necessary for Ethiopian agriculture to be oriented toward the supply of raw
materials for industry, nor for all industrial energies to focus on the processing of domestic
Finally, the above analysis largely takes the extent of these ‘linkages’ as given. However, the
most effective policies to stimulate growth may be those that strengthen domestic and
international linkages. Returning to the evaluation of sector‐specific government
expenditures, Mogues et al  find that the biggest gains in consumption are correlated
with investment in transportation infrastructure. Similarly, using micro data Dercon et al.
 find statistically significant and economically meaningful impacts of road access on
consumption and poverty outcomes: they find that households with access to a road capable
of carrying trucks and busses throughout the year increases consumption growth by 16%, for
example. Coupled with the prospect of increasing returns to scale in marketing activities –
considered in Dercon and Hill  – this suggests that government investments to
strengthen rural‐urban linkages should play a central role in growth strategy. Reductions in
the costs of transportation, for example, have the dual benefit of lowering urban food prices
– thereby stimulating industrial growth – and raising farmgate prices – thereby incentivizing
further growth in agricultural productivity. Investment in such linkages provides a means of28
stimulating agricultural innovation, as well as a means of maximizing the growth effects of
such innovation across the economy as a whole.
 ADENEW, B. A retrospective study on the performance of Ethiopian agriculture.
Report prepared for study on role of Agricultural Growth in Ethiopia, 2009.
 ALLEN, R. C. The growth of labor productivity in early modern English agriculture.
Explorations in Economic History 25, 2 (April 1988), 117–146.
 ALLEN, R. C. Labor productivity and farm size in English agriculture before
mechanization: Reply to Clark. Explorations in Economic History 28, 4 (October 1991),
 ALLEN, R. C. Ethiopian agriculture in a world‐historical context. Background note
prepared for study on role of Agricultural Growth in Ethiopia, 2008.
 ALLEN, R. C. The Agricultural Revolution. Forthcoming.
 BANDIERA, O., AND RASUL, I. Social networks and technology adoption in northern
mozambique. Economic Journal 116, 514 (October 2006), 869–902.
 BAUTISTA, R. M. Rapid agricultural growth is not enough: The philippines, 1965–1980.
In Agriculture on the Road to Industrialization, J. W. Mellor, Ed. Johns Hopkins
University Press, Baltimore, 1995, ch. 4, pp. 113–149.
 COLLIER, P., AND VENABLES, A. Rethinking trade preferences: How Africa can diversify its
exports. In The World Economy: Global Trade Policy 2007, D. Greenway, Ed. Wiley,
Hoboken, New Jersey, 2007.
 DE DATTA, S. K., TAURO, A. C., AND BALAOING, S. N. Effect of plant type and nitrogen level
on the growth characteristics and grain yield of indica rice in the tropics. Agronomics
Journal 60 (1968), 643–647.
 DEININGER, K., AND JIN, S. Land rental markets in the process of rural structural
transformation: Productivity and equity impacts from china. Unpublished, World
Bank, October 2006.
 DELGADO, C. L., HOPKINS, J., KELLY, V. A., HAZELL, P., MCKENNA, A. A., GRUHN, P., HOJJATI, B.,
SIL, J., AND COURBOIS, C. Agricultural growth linkages in sub‐saharan africa. IFPRI
Research Report 107, 1998.
 DERCON, S. Income risk, coping strategies, and safety nets. The World Bank Research
Observer 17, 2 (Fall 2002), 141–166.
 DERCON, S., GILLIGAN, D. O., HODDINOTT, J., AND WOLDEHANNA, T. The impact of roads and
agricultural extension on consumption growth and poverty in fifteen ethiopian
villages:. CSAE Working Paper WPS/2007‐01, January 2007.
 DIAO, X., FEKADU, B., HAGGBLADE, S., TAFFESSE, A. S., WAMISHO, K., AND YU, B. Agricultural
growth linkages in Ethiopia: Estimates using fixed and flexible price models. IFPRI
Discussion Paper No. 00695, March 2007.
 ESTUDILLO, J. P., AND OTSUKA, K. Green revolution, human capital, and off‐farm
employment: Changing sources of income among farm households in central luzon,
1966–1994. Economic Development and Cultural Change 47, 3 (April 1999), 497–523.
 ESWARAN, M., AND KOTWAL, A. A theory of real wage growth in LDCs. Journal of
Development Economics 42 (1993), 243–269.
 FIELDS, G. S. Rural‐urban migration, urban unemployment and underemployment and
job search activity in ldc’s. Journal of Development Economics 2, 2 (1975), 165–188.
 GOVERNMENT OF ETHIOPIA. An economic development strategy for Ethiopia: A
comprehensive guidance & a development strategy for the future. Ministry of
Planning and Economic Development, Addis Ababa, September 1993.
 GOVERNMENT OF ETHIOPIA. A plan for accelerated and sustained development to end
poverty (pasdep). Ministry of Finance and Economic Development, September 2006.
 GRILICHES, Z. Hybrid corn: an exploration in the economics of technological change.
Econometrica 25, 4 (October 1957), 501–522.
 GROSSMAN, G. M., AND ROSSI‐HANSBERG, E. Trading tasks: A simple theory of offshoring.
NBER Working Paper No. W12721, December 2006.
 HAYAMI, Y. Globalization and rural poverty: A perspective from a social observatory in
the philippines. UNU‐WIDER Research Paper 2006/44, 2006.
 HAYAMI, Y. Globalization and rural poverty: A perspective from a social observatory in
the philippines. UNU‐Wider working paper no. RP2006/44, 2006.
 HILL, R. V., DERCON, S., AND ZEITLIN, A. Growth constraints and scenarios for ethipian
smallholder agriculture. Report prepared for study on role of Agricultural Growth in
 HIRSCHMAN, A. The Strategy of Economic Development. Yale University Press, 1958.
 HODDINOT, J., DERCON, S., AND KRISHNAN, P. Networks and informal mutual support in
15 ethiopian villages. Working paper, Oxford University, July 2005.
 JORGENSON, D. W. The development of a dual economy. Economic Journal 71, 282
(June 1961), 309–334.
 JORGENSON, D. W. The role of agriculture in economic development: Classical versus
neoclassical mmodels of growth. In Subsistence Agriculture and Economic
Development, J. Clifton R. Wharton, Ed. Aldine, 1969.
 KANBUR, R., AND MCINTOSH, J. Dual economies. In The New Palgrave: A Dictionary of
Economics, J. Eatwell, M. Milgate, and P. K. Newman, Eds. The MacMillan Press, Ltd.,
 LEWIS, W. A. Economic development with unlimited supplies of labour. Manchester
School 28, 2 (May 1954), 139–191.
 LOENING, J., RIJKERS, B., AND SÖDERBOM, M. Nonfarm microenterprise performance and
the investment climate: Evidence from rural ethiopia. World Bank, Policy Research
Working Paper 4577, March 2008.
 MAO, Y.‐K., AND SCHIVE, C. Agricultural and industrial development in taiwan. In
Agriculture on the Road to Industrialization, J. W. Mellor, Ed. Johns Hopkins
University Press, Baltimore, 1995, ch. 2, pp. 23–66.
 MELLOR, J. W. Introduction. In Agriculture on the Road to Industrialization, J. W.
Mellor, Ed. Johns Hopkins University Press, Baltimore, 1995, pp. 1–22.
 RASHID, S., ASSEFA, M., AND AYELE, G. Distortions to agricultural incentives in ethiopia.
International Food Policy Research Institute, Agricultural Distortions Working Paper
43, December 2007.
 RIJKERS, B., SÖDERBOM, M., AND TEAL, F. Rural non‐farm enterprises in Ethiopia:
Challenges and prospects. Background note prepared for study on role of
Agricultural Growth in Ethiopia, 2008.
 SANDEFUR, J., AND SIDDIQI, B. Global trade in tasks. Background note for the 2008
UNIDO report, May 2008.
 SEN, A. K. Peasants and dualism with or without surplus labor. The Journal of Political
Economy 74, 5 (October 1966), 425–450.
 SEYOUM TAFFESSE, A. Decomposition of growth in cereal production in ethiopia.
Background note prepared for study on role of Agricultural Growth in Ethiopia,
 SWAMINATHAN, M. S. An evergreen revolution. Crop Science 46, 5 (2006), 2293–2303.
 TIMMER, C. P. The turnip, the new husbandry, and the English agricultural revolution.
Quarterly Journal of Economics 83, 3 (August 1969), 375–395.
 VINES, D., AND ZEITLIN, A. Dual economies. In The New Palgrave: A Dictionary of
Economics, 2nd ed., S. N. Durlauf and L. E. Blume, Eds. Macmillan, London. Also
available online: http://users.ox.ac.uk/$$exet1357/, forthcoming.
 WORLD BANK. World Development Report 2008: Agriculture for Development. The
International Bank for Reconstruction and Development, Washington, D.C., 2007.
 MOGUES, T., AYELE, G., AND PAULOS, Z. The bang for the birr: Public expenditures and
rural welfare in ethiopia. IFPRI Research Report no. 160, 2008.
 MUNSHI, K. Social learning in a heterogeneous population: technology diffusion in the
indian green revolution. Journal of Development Economics 73, 1 (February 2004),
 NEGASSA, A., AND MYERS, R. J. Estimating policy effects on spatial market efficiency: An
extension to the parity bounds model. American Journal of Agricultural Economics
89, 2 (May 2007), 338–352.
 NICITA, A., AND OLARREAGA, M. Trade and production: 1976–2004. World Bank
Economic Review 21, 1 (2007), 165–171.
 OTSUKA, K. Strategy for agricultural growth in Ethiopia. Background note prepared for
study on role of Agricultural Growth in Ethiopia, 2008.
 OTSUKA, K., AND KALIRAJAN, K. An exploration in to the possibliity of a Green Revolution
in Sub‐Saharan Africa: An introduction. Journal of Agricultural and Development
Economics 2, 1 (2005), 1–6.
 OTSUKA, K., AND KALIRAJAN, K. P. Rice green revolution in Asia and its transferability to
Africa: An introduction. Developing Economies 44, 2 (2006), 107–122.
 OTSUKA, K., AND YAMANO, T. The possibility of a green revolution in Sub‐Saharan
Africa: Evidence from Kenya. Electronic Journal of Agricultural and Development
Economics 2, 1 (2005), 7–19.
 PARK, A. Agricultural development in China: Lessons for Ethiopia. Background note
prepared for study on role of Agricultural Growth in Ethiopia, 2008.
 RANIS, G. Is dualism worth revisiting? Yale University, Economic Growth Center,
Center Discussion Paper No. 870, September 2003.
 RANIS, G., AND FEI, J. C. A theory of economic development. The American Economic
Review 51, 4 (September 1961), 533–565.