SlideShare a Scribd company logo
1 of 16
Download to read offline
Full Terms & Conditions of access and use can be found at
http://www.tandfonline.com/action/journalInformation?journalCode=raee20
Download by: [Research & Evidence Division] Date: 24 October 2016, At: 00:15
The Journal of Agricultural Education and Extension
Competence for Rural Innovation and Transformation
ISSN: 1389-224X (Print) 1750-8622 (Online) Journal homepage: http://www.tandfonline.com/loi/raee20
The perceived impact of agricultural advice in
Ethiopia
Alexander Hamilton & John Hudson
To cite this article: Alexander Hamilton & John Hudson (2016): The perceived impact of
agricultural advice in Ethiopia, The Journal of Agricultural Education and Extension, DOI:
10.1080/1389224X.2016.1245151
To link to this article: http://dx.doi.org/10.1080/1389224X.2016.1245151
Published online: 23 Oct 2016.
Submit your article to this journal
View related articles
View Crossmark data
The perceived impact of agricultural advice in Ethiopia
Alexander Hamiltona
and John Hudsonb
a
Department for International Development, Sudan, BFPO 5312, Ruislip, UK; b
Department of Economics,
University of Bath, Bath, UK
ABSTRACT
Purpose: We examine the impact of advice given by extension
agents to Ethiopian farmers, as perceived by the farmers
themselves. Design/methodology/approach: Using survey data from
2014, we analyze the perceived impact of advice on farmers’
incomes and crop yields. We use a bootstrapped instrumental
variable (IV) estimator and the conditional mixed process
estimator. Theoretical implications: The impact of advice will
depend upon its relevance and whether and how efficiently it is
implemented by the farmer. This in part depends upon the
farmer’s ability and on the impact of fully implemented advice on
output, which will vary from farm to farm. Findings: There is a
positive perceived impact of most advice on both crop yields and
income. However, some advice works better in drought-affected
areas and other in non-drought-affected areas. Fertilizers have
more impact on crop yields than income, possibly reflecting cost
factors. There is evidence that the farmers’ ability to implement
the advice increases with their level of education and that advice
is being tailored to the needs of the individual. Practical
implications: Advice has a positive impact on both crop yields and
income. However, not all advice is equally effective and
effectiveness varies according to farmer and farm characteristics.
There is little evidence of credit advice having a positive impact.
Originality/value: The paper is one of only a few to analyze
farmers’ perceptions of advice impact and, as far as we are aware,
is the first to analyze how advice effectiveness varies according to
farmer and farm characteristics.
ARTICLE HISTORY
Received 24 August 2015
Accepted 3 October 2016
KEYWORDS
Extension agents; farmers’
income; crop yields; fertilizer
advice; land management
advice; credit advice
Introduction
A core part of the Ethiopian government’s investment in agriculture is the public agricul-
tural extension system, with a substantial increase in the number of people graduating as
extension or development agents (DAs) (Davis et al. 2010). The Government has also
established farmer training centers, with two hectares of demonstration fields (Lefort
2012), in every local administrative area (there are 18,000 nationwide) with three exten-
sion agents at every training center. This effort could lead to Ethiopia having the
world’s highest ratio of extension agents to farmers1
and is reflected in the proportion
of fields that use extension services rising from just over 5% in 2008 to 12% in 2011
(Khan et al. 2014).
© 2016 Wageningen University
CONTACT John Hudson j.r.hudson@bath.ac.uk Department of Economics, University of Bath, Bath, BA2 7AY, UK
JOURNAL OF AGRICULTURAL EDUCATION AND EXTENSION, 2016
http://dx.doi.org/10.1080/1389224X.2016.1245151
There is evidence of an impact on agricultural efficiency. Dercon et al. (2009) showed
that one visit from a DA raised production growth by 7% and reduced poverty by 10%.
The posting of agricultural extension agents in local communities has improved their
attentiveness to farmers’ needs and constraints, and enhanced the working relationship
between them (Cohen and Lemma 2011). However, others have tended to be more scep-
tical about the impact of extension agents (Lefort 2012; Pender, Place, and Ehui 2006).
Davis et al. (2010) also note that agricultural productivity remains low, inputs are
scarce and expensive, and market and credit access are extremely limited and according
to Buchy and Basaznew (2005), women-focused extension is also limited.
In general, Ethiopia has tended to be viewed unfavorably in terms of agricultural pro-
ductivity (Dercon and Christiaensen 2011; Spielman et al. 2010). However in recent years,
certainly since 2008, there has been a substantial, even dramatic, improvement in cereal
yields. This has been accompanied by a 12.5% expansion in land under cereal production
between 2003 and 2012; hence it is not the case that more marginal land has disappeared
from the picture, raising the average productivity of what is left. Apart from the extensive
changes to the extension system, other changes impacting on Ethiopian agriculture
include substantial transport improvements which between 2000 and 2013 tripled the
length of all-weather surface roads, and a rapid increase in the urban population of
some 3.7 million (Bachewe et al. 2015). Together these have increased the market for com-
mercial crops and partly as a consequence output–input price ratios have increased sub-
stantially. There has also been increased access to credit, which saw the number of active
borrowers rise from about half a million in 2003 to approximately 3.5 million in 2014.
In this paper, we will be focusing on the potential role of the extension system in explain-
ing this recent success. Specifically, we analyze perceptions of the difference that the advice
made to the recipients of that advice with respect to (i) crop yields and (ii) income. Self-
perceptions data have the advantage that the focus is on the difference made by the exten-
sion service advice, whereas actual output and income can change for numerous reasons
unconnected with that advice. To include all the factors that might affect yields and
income is difficult and such data are also likely to be subject to measurement errors. We
assume that the farmers in evaluating the impact of the advice take account of all these
other factors. In addition, because the dependent variable relates to the impact of advice,
we can examine how that impact varies with variables such as the farmer’s age, education
and water resources. Other studies have tended to focus on the impact of variables such
as education on productivity, but not on the effectiveness of advice.
The main research question is whether extension advice benefits farmers both in terms
of their crop yields and income. We expect this to be the case, but the literature is ambig-
uous on the issue. A secondary question is what types of advice work best and in what
context. There is relatively little literature on this. However, there is a literature which
suggests that different types of technology, for example, fertilizers, work better in some
conditions rather than others (Dercon and Christiaensen 2011; Kassie et al. 2010). This
is of relevance, but advice on implementing a technology is different to the implemen-
tation of that technology. In addition part of this research relates to more than a
decade ago and, as already noted, in this time the extension system has substantially
changed, technology has changed and Ethiopia has changed. Hence, there is a need for
new research in this area.
2 A. HAMILTON AND J. HUDSON
We will be analyzing the impact of advice on (i) agricultural practices, (ii) land manage-
ment, (iii) fertilizers, (iv) marketing, (v) access to credit facilities and (vi) animal husban-
dry practices. We find that extension agent advice does lead to a perceived positive impact
on both crop yields and income. However, the effectiveness of the advice depends upon
the level of education of the farmer. It also differs between drought-affected and non-
drought-affected lands. The paper proceeds as follows. In the next section we discuss
the relevant literature. We then discuss, from a theoretical perspective, how advice
might impact on farmers and other methodological issues, and we also present the
data. The penultimate section presents the results and finally we conclude the paper.
Background
The literature on the extension program has been somewhat ambivalent in terms of its
impact. Extension services generally have positive impacts on nutrition and poverty
reduction (Dercon et al. 2009). However, their success has been said to be constrained
by weaknesses elsewhere in the system. Hence EEA/EEPRI (2006) argue that distribution
channels and institutions are flawed, the formal seed system has weaknesses, and there is a
lack of markets, both for inputs and outputs. Agents transfer knowledge to farmers, with
relatively little knowledge flow in the reverse direction (Buchy and Basaznew 2005), which
can lead to the knowledge not being tailored to the farmer’s needs. The literature argues
that the extension system has focused on the distribution of standard packages to farmers,
including seeds and commercial fertilizer, credit, soil and water conservation, livestock
and training. Efforts to promote other aspects of sustainable land management have con-
centrated on soil erosion without consideration of the underlying socioeconomic reasons
for low soil productivity (Kassie et al. 2010). As a consequence, advice has been given
which has been unprofitable, risky or irrelevant given the farmer’s resource constraints
(Pender, Place, and Ehui 2006). Lefort (2012) cites research which concludes the extension
program was not much use to farmers. However, Spielman et al. (2010) also note that a
series of reforms have been made to redress these weaknesses.
Berhanu and Poulton (2014) argue that the extension system is used to promote gov-
ernment control, with extension workers facilitating scrutiny and control of activities.
Extension workers engage in non-extension activities, such as administration, credit
repayment and tax collection (Kelemework and Kassa 2006). Apart from diverting their
attention from extension services, this can also strain relationships with local farmers.
In addition, it is claimed that in their allocation of seeds, fertilizers and credit, extension
workers prioritize farmers loyal to the governing coalition. It is also important to stress
that extension agents are not always transmitting knowledge and advice which can
immediately add to the farmer’s productivity. They may also be concerned with societal
impact. For example, in some cases the advice relates to environmental factors (Abegaz
and Wims 2015) for which the farmer may perceive little personal benefit, although
this is not to say that such benefit is absent.
More recent evidence is a little more positive, although still emphasizing that practices
could be better. Elias et al. (2013) found that participation in the extension program
increased productivity by about 20%. Other factors which influenced productivity
included age, male head of household and plot characteristics. Despite this, crop yields
were below the targets set by the extension program. Khan et al. (2014) conclude that
THE JOURNAL OF AGRICULTURAL EDUCATION AND EXTENSION 3
woreda-level spending on agricultural extension workers is associated with higher yields
for major crops and increases the probability that farmers will improve their farming
techniques.
There is relatively little evidence on the relative impact of different forms of advice, but
some related work has been done on the different constraints facing, and the different
technologies used by, farmers. One of the reasons Dercon and Christiaensen (2011)
gave for the poor performance of Ethiopian agriculture was lack of fertilizer use. Lack
of knowledge and skills in adopting modern inputs was only a very minor constraint in
1999, which would suggest only a limited role for extension agent advice in this respect.
Nor was lack of credit deemed a major factor. Fertilized plots were characterized by
greater yields than non-fertilized plots, although not in periods of extreme droughts
and floods. Thus, Dercon and Christiaensen argued that fertilizer use is a high return,
but high risk technology. Kassie et al. (2010) find evidence of a strong impact of land man-
agement practices on agricultural productivity in the low agricultural potential areas. In
the high agricultural potential region, however, fertilizers have a very significant and posi-
tive impact on crop productivity, whereas land management practices have no significant
impact. Fertilizers may be less profitable in such areas due to a lack of soil moisture.
Hence, their analysis raises the important point that the impact of different forms of
advice and increasing knowledge may not be the same in all areas, but vary according
to local conditions.
Methods
We will be analyzing the impact of advice on both crop yields and inccome. For the
moment, we focus on the impact on crop yields (Y), although the analysis for income
follows a similar path. We assume Y to be a function of resources, which are in turn a func-
tion of knowledge:
Yit = Aig(Sit, Lit,Kit, Fit, Wit), (1)
where i denotes the ith farmer and t the time period. Ai denotes overall efficiency with
which the different factors of production are used, that is, it is total factor productivity
(TFP) defined at the level of the individual farmer. In our analysis, we assume that this
is the vehicle by which extension agent advice impacts on output. g(.) can be regarded
as the basic output of the farm independent of the characteristics and expertise of the
farmer. The production function is composed of land (S), labor (L), capital (K), fertilizer
(F) and water (W). The impact of extension agent advice to farmer i (Ei) on output is then
given by
∂Yi
∂Ai
∂Ai
∂Ei
Ei =
6
j=1
∂Yi
∂Ai
∂Ai
∂Eij
Eij = g(Sit, Lit,Kit, Fit, Wit)
6
j=1
∂A
∂Ej
Eij. (2)
This is the combined impact of the six specific types of advice in our analysis.2
Eij is a con-
tinuous measure of the advice given to individual i on advice type j, although we only have
a discrete measure of this, which takes a value of one if advice was given. It has a lower
bound of zero, which is operative when advice of this type is not given. The marginal
impact of any one piece of advice is composed of ∂Ai/∂Eij, the impact of this advice on
4 A. HAMILTON AND J. HUDSON
TFP, and secondly, the impact of TFP on output, which is from (1) just g(.). ∂Ai/∂Eij will
depend upon the characteristics of the farm (T). It may also depend upon the character-
istics of the farmer, with more knowledgeable farmers more able to implement the advice
efficiently, although at the same time more knowledgeable farmers may be less likely to
seek advice. In this case, we link knowledge to education (Ed) and age, the latter
through learning by doing. Thus,
∂Ai
∂Eij
= fj(Edi, Agei, Ti, Wi). (3)
The dependent variable is coded 1 if the extension agent advice was perceived as making
no difference, 2 if it made some difference and 3 if it made a lot of difference. The response
lies in the kth category if:
ak−1 , g(.)
6
j=1
fj(.)Eij , ak; k = 1, . . . , m. (4)
In our analysis m = 3. Note that α0 = −∞ and αm = +∞, and hence we estimate just a1 and
a2. Define Zi,k = 1 if g(.) 6
j=1 fj(.)Eij is in the kth category, and otherwise Zi,k = 0. Lin-
earizing g(.) 6
j=1 fj(.)) we can estimate both the coefficients and the dividing points (αk)
between the different categories by ordered probit. The independent variables will include
farm characteristics, individual characteristics and dummy variables operative if a particu-
lar type of advice was given. This suggests a number of hypotheses which we will be
testing:
H1: Extension agent advice impacts positively on both crop yields and income.
H2: The degree of the impact for different types of advice will depend upon the characteristics
of the farmer and the characteristics of their farm.
H3: The degree of impact of this advice will be greater for more educated individuals and also
for older individuals, although we can expect these individuals to be less likely to receive
advice.
H1 reflects our first research question, on whether extension advice benefits farmers in
terms of both income and crop yields. The other two hypotheses relate to our second
research question as to what types of advice work best and under which conditions.
Our theoretical analysis has helped inform both of our research questions. The advice
will make a difference if it is relevant advice and if the farmer, given their own and
their farm’s characteristics, will implement it efficiently. The analysis also emphasizes
that different types of advice will have different impacts, which are linked to differences
in ∂Ai/∂Eij.
Data
The data were obtained from the Woreda and City Benchmarking Survey (WCBS) col-
lected in 2014 using a multi-stage stratified sampling approach based on the remoteness
and food security levels of households. Within each region the sub-sample size was deter-
mined by population (based on census data). Data were collected on 326 kebele in 48
THE JOURNAL OF AGRICULTURAL EDUCATION AND EXTENSION 5
woreda covering the whole of the country.3
In total 7429 individuals were interviewed.
This survey is focused on rural areas. The variables we use are defined in Table 1. Infor-
mation on the characteristics of the plot relates to whether the individual grew crops and
had animals. Most people sampled (60.8%) raised both animals and grew crops, a substan-
tial proportion just grew crops (31.3%) and an even smaller proportion just had animals
(7.9%).
The proportions receiving one, two, three, four, five and six types of advice as listed in
the appendix were 27%, 16%, 17%, 14%, 5% and 13%, respectively. Other advice is feasible,
which is why 8% received none of the types of advice specified. Table 2 shows the
summary data as it varies across individual characteristics. It is noticeable that the
young tend to be more likely recipients of advice, possibly reflecting that they will have
learnt less ‘by doing’. It is also noticeable how the highly educated are more likely to
have received advice on marketing and credit. The final two columns relate to the
impact this advice, in general, has had on crop yields and the individual’s income.4
The
responses ranged from 1 (none) to 3 (a lot). Thus, we assume that output and income
cannot fall as a result of the advice received. The responses to both questions were
fairly enthusiastic, although very slightly more so for crop yields than income. The
biggest gainers in both respects tend to be the better educated. An important question
relates to the role of the advice apparently not included for example, mechanization;
crop protection measures etc.? To an extent they may be subsumed within one of the
other categories, particularly agricultural practices, which as we note from Table 2 was
the most frequently cited form of advice. Any advice which still falls outside these cat-
egories would be picked up by the constant term in the regressions.
Regression results
In Table 3, we present the results relating to the impact on yields.5
Column 1 shows that
advice received on animal husbandry, marketing and land management were all signifi-
cant at the 1% level of significance and fertilizers at the 5% level. Advice on agricultural
Table 1. Data definitions.
Socioeconomic, demographic variables
Age Age in years
Education Coded from 1 (no schooling) to 24 (degree) and 25 above degree
Male Coded 1 ifa male
Family size Number of people currently living in the individual’s household
Plot characteristics
Grows crops Coded 1 if the individual grows crops, otherwise 0
Rears animals Coded 1 if the individual rears animals, otherwise 0
Drought Coded 1 if the individual suffers from regular periods of drought in the sense of a shortage of
drinking water, otherwise 0
Received advice on (coded 1 for yes and 0 no)
Agricultural practices; land management; fertilizer; marketing; credit facilities; animal husbandry
Impact
Crops improve/Income
improves
The difference the above advice has made to the crop yield/income
ranging from 1 (none) to 3 (a lot)
Kebele based variables (average of responses of others in the individual’s locality). Regional dummy variables relate to Tigray,
Amhara, Oromiya, SNNP, Binshangul Gumuz, Afar, Somali and Gambela. Because of the small numbers of people in our
sample who are in Gambela we join this region together with SNNP, the two being adjacent in the south west part of the
country. We also join the second smallest region, Afar, with Tigray, who are adjacent on the northern edge of Ethiopia
6 A. HAMILTON AND J. HUDSON
practices was not significant. However being in receipt of advice on credit was significantly
negative. This does not imply that output was lowered as a result of the advice. Rather, it
implies that being in receipt of advice on credit increases the probability of extension agent
advice having no impact. However, this does alert us to a potential problem of endogene-
ity. To the extent that the farmer is the one seeking this advice, rather than being proffered
it by the extension agent or some other person, then it could signal that the individual is in
financial problems. At the very least it reflects an interest by the farmer in gaining access to
credit. The negative sign in the regression may be picking this up. Another aspect of endo-
geneity is that it is possible that the individual selected for the advice is in some way more
able to use it.
Because of this possibility, in the second regression we instrument the advice variables
in a two-stage process. Firstly, we regress the advice variables on all the exogenous vari-
ables present in column 1 plus a series of variables relating to the extent others in the indi-
vidual’s kebele received different forms of advice. That is, taking credit as an example, for
each individual we calculate the average number of people in their kebele, excluding them-
selves, who received advice on credit. These effectively act as instruments. The second
stage of the estimation takes the predicted values from these regressions and re-estimates
the relationships shown in column 1. The results are shown in column 3.2. The main
change is the insignificance of advice on marketing and also on credit. A Hausman test
indicated that the two sets of coefficients were significantly different and hence the
need for an instrumental variable (IV) approach. This two-stage estimation technique is
similar to that employed by Adams, Almeida, and Ferreira (2009). The literature tends
to bootstrap the standard errors (Clarke and Windmeijer 2012) and this has been done.
On each of the 100 iterations of the bootstrap, both stages of the two-stage estimation tech-
nique were estimated. The t statistics reported are based on these standard errors.
In addition to the above, we also estimated the equations using IVs in the context of a
conditional mixed process (cmp) model (Roodman 2011), that is estimating a system
where the different equations can have different kinds of dependent variables. We
assume joint normality of the error terms of the different equations. It is a full system tech-
nique, which takes account of potential correlation between the different error terms in the
equations. There are potentially eight equations in the system, one for each of the advice
variables and two more for the impact variables, that is, the impact on crops and income.
However, this makes considerable demands on the data and we had to simplify the
equations in order to obtain estimates. Firstly, we estimated each of the policy impact
Table 2. Summary data relating to individual characteristics.
Animal
husbandry Credit Marketing Fertilizer
Land
management
Agricultural
practices
Crops
improve
Income
improves
All 0.62 0.291 0.329 0.679 0.524 0.817 2.36 2.3
Young <30 0.631 0.321 0.329 0.731 0.561 0.847 2.37 2.29
Older ≥30 0.612 0.264 0.326 0.634 0.488 0.787 2.35 2.29
Male 0.621 0.278 0.317 0.665 0.525 0.815 2.36 2.3
Highly
educated
0.585 0.342 0.447 0.707 0.523 0.845 2.51 2.41
No education 0.622 0.292 0.313 0.678 0.501 0.804 2.32
2.28
Notes: The final two columns relate to the average response which varied from 1 (none) to 3 (a lot) to the difference the
support has made. All other columns relate to the proportion receiving advice in the different headings.
THE JOURNAL OF AGRICULTURAL EDUCATION AND EXTENSION 7
Table 3. Regression results: impact on crop yields.
Sample Probit: Full 3.1 IV 2 stage Full 3.2 Probit: Full 3.3 Drought 3.4 No drought 3.5 CMP: Full 3.6 Drought 3.7 No drought 3.8
Extension agent advice
Animal husbandry 0.255**
(5.35)
0.4696*
(2.41)
0.4695*
(2.41)
0.9789**
(4.07)
0.1662
(0.52)
0.6085**
(5.27)
0.6273**
(4.44)
0.648**
(3.22)
Credit −0.2344**
(4.77)
0.0069
(0.03)
−0.0284
(0.12)
−0.2431
(0.95)
0.9285*
(2.11)
−0.3592*
(2.42)
−0.1748
(1.03)
−0.2966*
(2.18)
Marketing 0.4172**
(8.76)
0.3156
(1.72)
0.4122*
(2.38)
−0.0737
(0.35)
0.9992**
(2.91)
0.5683**
(4.45)
−0.1576
(0.93)
1.452**
(13.77)
Fertilizers 0.0884*
(1.98)
0.8863**
(3.68)
1.006**
(4.24)
0.8349*
(2.46)
0.7405*
(2.01)
0.8513**
(7.69)
0.9966**
(8.28)
0.7314**
(4.29)
LM – land management 0.2874**
(6.47)
0.6833**
(3.09)
AG – agricultural practices 0.0756
(1.30)
0.1408
(0.61)
LM + AG 0.8374**
(2.69)
0.3536
(0.73)
2.378**
(4.74)
1.134**
(5.82)
0.3289
(0.74)
1.222**
(4.67)
Individual characteristics
Log age 0.2891**
(4.52)
0.1094
(1.62)
0.102
(1.51)
−0.0723
(0.83)
0.1991
(1.79)
0.0518
(0.78)
−0.0664
(0.75)
0.1341
(1.33)
Education 0.0264**
(5.26)
0.0244**
(4.05)
0.0249**
(4.16)
0.0192*
(2.34)
0.0236**
(2.91)
0.0276**
(5.61)
0.0262**
(3.86)
0.0122
(1.67)
Male −0.0538
(1.52)
Log family size −0.0785
(1.78)
Farm characteristics
Crops 1.316**
(12.81)
0.6006**
(3.43)
0.4652**
(3.04)
0.058
(0.26)
0.3388
(1.39)
−0.2539
(1.60)
−0.3369
(1.18)
0.279
(1.14)
Animals −0.0251
(0.51)
−0.3715**
(2.91)
−0.3685**
(2.83)
−0.3783*
(2.35)
−0.4832*
(2.50)
−0.1909*
(2.16)
−0.2073
(1.94)
−0.1815
(1.22)
Constant −1.562**
(5.20)
−1.232**
(2.97)
−2.250**
(4.99)
Estimated cutoff points
Cutoff point 1 ˆa1 0.2798
(1.06)
−0.7927**
(2.66)
−0.7646*
(2.56)
−0.9719*
(2.26)
−0.3325
(0.69)
Cutoff point 2 ˆa2 2.573**
(9.65)
1.477**
(4.91)
1.503**
(5.00)
1.570**
(3.65)
1.801**
(3.71)
Observations 5192 5185 5185 2928 2257 5192 2932 2260
Log likelihood −3733 −3768 −3770 −2010 −1534 −12516 −7489 −4612
X2
1228 1089 1089 380.9 850.8 355571 159143 6384
Notes: Equation 3.1 estimated by ordered probit, 3.2–3.5 by a two stage instrumental variable ordered probit with bootstrapped standard errors, 3.6–3.8 by a conditional mixed processor estimator
with a binary dependent variable. (.) denotes t statistics and **/* denotes significance at the 1% and 5% levels. Regional dummy variables included in all regressions.
8A.HAMILTONANDJ.HUDSON
variables separately. Secondly, we used binomial probit for the impact equations rather
than ordered probit, with the variable differentiated between a lot of impact (coded 1)
and some or no impact. Finally, we combined two of the advice variables together, that
is land management and agricultural practices. These potentially relate to all farmers
and also to farming per se rather than other aspects of the business and have a reasonably
high correlation. The cmp estimator is complex, but the interpretation of the coefficients
and t statistics is as with techniques such as OLS.
The results are shown in column 3.6 and are similar to previously. The main difference
is the significance, at the 1% level, of the advice variable relating to both agricultural prac-
tices and land management and also marketing advice. In column 3.3, we replicate these
using the two stage approach used in 3.2. Comparing 3.3 and 3.6, the only significant
difference relates to the negative coefficient on the credit advice variable. Using both tech-
niques provides a robustness check on the findings and the cmp results largely confirm
those of the two stage IV approach.6
Taking the equations as a whole, family size is
never significant, nor gender and were dropped after the first equation.7
More educated
people and older people tend to have benefited more from the advice than others, although
the latter only in 3.1. However, in the cmp regression the variable relating to crop growers
is no longer positively significant.
The literature has suggested that the impact of advice may vary according to the con-
ditions facing the individual, for example soil moisture. We do not have in the database a
measure of this nor rainfall in the kebele, but we do have a variable which asked the indi-
vidual whether they were usually subject to water shortages for drinking at some time in
the year. Slightly over 51% responded that they were subject to such shortages. We now
split the sample into those who were and were not subject to such water shortages.
Columns 3.4 and 3.5 relate to those who were and were not subject to water shortages.
The positive impact of marketing advice is limited to the latter as is that of land management
and agricultural practices. However, the negative impact of credit advice is evident only for
those in non-drought areas. Finally, we note that animal husbandry advice is only significant
for those in areas of drought and advice on fertilizers is significant in both areas. These
equations were estimated using the two stage approach. Columns 3.7 and 3.8 show the
results of using the cmp estimator. The main difference is the positive significance of
animal husbandry advice in both areas. In Table 4, we look at the results pertaining to
income. The results are very similar. This was to be expected, but it was always possible
that because of expenses incurred in implementing the advice, a positive impact on yield
would not translate into a positive impact on income. The main difference to Table 3 is
the significantly negative coefficient on credit advice in both areas, implying that receivers
of such advice are significantly less likely to have perceived a beneficial impact on income.
In addition, those with either crops or animals were also less likely to perceive benefits.
Because the regression is limited to those who received advice, the possibility exists for
sample selection bias. When we tested for this in the two-stage IV regressions the evidence
was mixed. The inverse Mills ratio is insignificant in the income regression, but significant
in the crop yield regression, although not in drought-affected areas. The significance of the
coefficients reported in Table 3 for the full sample for crop yields did not change and nor
did those in the non-drought regression. The main difference in both these regressions was
an increase in the size of the coefficient on animal husbandry advice. The sample selection
equation included a kebele based variable reflecting the number in the kebele, other than
THE JOURNAL OF AGRICULTURAL EDUCATION AND EXTENSION 9
Table 4. Regression results: impact on income.
Sample Probit: Full 4.1 IV 2 stage Full 4.2 Probit: Full 4.3 Drought 4.4 No drought 4.5 CMP:Full 4.6 Drought 4.7 No drought 4.8
Extension agent advice
Animal husbandry 0.326**
(7.20)
0.8695**
(5.52)
0.773**
(4.82)
1.380**
(6.33)
0.4807
(1.62)
0.6428**
(5.19)
0.8796**
(6.24)
0.5656**
(2.70)
Credit −0.2863**
(6.27)
−0.0429
(0.19)
−0.0815
(0.35)
−0.5072
(1.76)
0.4657
(1.16)
−0.612**
(4.10)
−0.6696**
(3.36)
−0.4433**
(2.85)
Marketing 0.4026**
(8.83)
0.2882
(1.63)
0.4343*
(2.47)
−0.0009
(0.00)
1.153**
(3.73)
0.6562**
(4.91)
−0.103
(0.53)
1.466**
(12.21)
Fertilizers 0.1489**
(3.30)
0.149
(0.59)
0.583*
(2.40)
0.5567
(1.65)
0.3363
(0.82)
0.7453**
(5.43)
0.901**
(6.00)
0.7655**
(3.49)
Land management 0.2292**
(5.12)
0.6268**
(2.61)
Agricultural practices −0.0486
(0.85)
0.2364
(1.08)
LM + AG 0.739*
(2.23)
−0.0341
(0.07)
2.039**
(3.55)
1.332**
(5.64)
0.0223
(0.04)
1.164**
(2.96)
Individual characteristics
Log age 0.1423*
(2.24)
0.0612
(0.92)
0.0787
(1.17)
−0.1311
(1.33)
0.1717
(1.62)
−0.0833
(1.25)
−0.1737
(1.86)
−0.078
(0.76)
Education 0.0177**
(3.78)
0.017**
(3.33)
0.0179**
(3.47)
0.0147*
(2.00)
0.0109
(1.40)
0.0153**
(3.14)
0.0173*
(2.55)
−0.0046
(0.63)
Male −0.0065
(0.19)
Log family size 0.0081
(0.19)
Farm characteristics
Crops 0.2137*
(2.22)
−0.2432
(1.63)
−0.2617*
(1.96)
−0.0601
(0.25)
−0.8143**
(3.88)
−0.7315**
(4.45)
−0.3371
(0.98)
−0.3603
(1.33)
Animals −0.1338**
(2.81)
−0.5482**
(5.07)
−0.6039**
(5.42)
−0.4797**
(3.21)
−0.8772**
(4.73)
−0.3407**
(3.67)
−0.1669
(1.49)
−0.5536**
(3.74)
Constant −0.939**
(3.14)
−1.829**
(3.39)
−0.6396
(1.36)
Estimated cutoff points
Cutoff point 1 ˆa1 −2.423**
(4.16)
−2.500**
(8.96)
−1.043**
(3.64)
−0.8468
(1.87)
−1.210**
(2.84)
Cutoff point 2 ˆa2 −0.3297
(0.57)
−0.4381
(1.57)
1.003**
(3.51)
1.586**
(3.52)
0.6052
(1.40)
Observations 5195 5188 5188 2928 2260 5190 2930 2260
Log likelihood −4096 −4154 −4177 −2095 −1843 −12556 −7433 −4688
X2
697.6 590.5 566.6 301.6 535.1 111195 3517 64356
Notes: Equation 4.1 estimated by ordered probit, 4.2–4.5 by a two-stage instrumental variable ordered probit with bootstrapped standard errors, 4.6–4.8 by a conditional mixed processor estimator
with a binary dependent variable. (.) denotes t statistics and **/* denotes significance at the 1% and 5% levels. Regional dummy variables included in all regressions.
10A.HAMILTONANDJ.HUDSON
the respondent, receiving advice. Apart from this, older and more educated people were
less likely to receive advice. Men were more likely to receive advice, as were those in
large families and those with animals or crops. This is potentially consistent with those
who report a bias in extension advice towards men (Buchy and Basaznew 2005). A variable
differentiating drought from non-drought areas was insignificant.
In Table 5, we show the probabilities of the respondent finding that the advice made ‘a
lot of difference’ under different scenarios. These are obtained from the two stage instru-
mental ordered probit regressions. They are based on a 45-year-old individual with crops,
but no animals. The regional variables are averaged according to sample population. The
first element in the first column shows the probability for someone with the above charac-
teristics who receives none of the specific types of advice listed and has no schooling. The
second column relates to someone with an education corresponding to ‘grade 8’, and this
raises the probability to 0.0627. In the second row, we have the probabilities for someone
of the above characteristics who received advice on animal husbandry. These are consider-
able higher, reflecting the effective nature of this advice. However, the most effective forms
of advice are on fertilizers and the combined advice on land management and agricultural
practices. The next two columns relate solely to farmers in drought areas. The most impor-
tant forms of advice are on animal husbandry and fertilizers, with land management and
agricultural practices being third. This is in sharp contrast to the non-drought areas, where
the most effective advice is on agricultural planning and land management. The results for
income are also shown. The main difference is the increased impact of marketing advice.
Some of these impacts look quite small, although it should be borne in mind that they do
relate to the probability of having made ‘a lot of difference’ and also that multiple forms of
advice are often given which substantially increases this probability.
Conclusions and policy implications
With respect to the main research questions, we find that extension agent advice impacts
positively on both crop yields and income as reflected by the farmers’ own perceptions.
Table 5. Probabilities of advice have a ‘lot of difference’ under different scenarios.
Full sample Drought areas Non-drought areas
Education None Grade 8 None Grade 8 None Grade 8
Crop yield
None 0.040 0.063 0.059 0.082 0.0036 0.0067
Animal husbandry 0.099 0.144 0.280 0.341 0.0059 0.0105
Marketing 0.089 0.131 0.051 0.072 0.0458 0.0702
Credit 0.037 0.059 0.036 0.051 0.0394 0.0612
Fertilizer 0.226 0.299 0.233 0.289 0.0259 0.0415
Planning and land man. 0.179 0.243 0.113 0.150 0.3790 0.4620
Income
None 0.059 0.080 0.087 0.110 0.0146 0.0186
Animal husbandry 0.214 0.264 0.508 0.561 0.0445 0.0545
Marketing 0.129 0.166 0.088 0.110 0.1520 0.1760
Credit 0.050 0.069 0.031 0.041 0.0431 0.0529
Fertilizer 0.163 0.206 0.211 0.251 0.0325 0.0403
Planning and land man. 0.204 0.253 0.082 0.104 0.4430 0.4820
Notes: These probabilities are based on regressions 3.3–3.5 and 4.3–4.5. They relate to a 45-year-old individual with crops,
but no animals in a typical region. They show the probability of a single type of advice making ‘a lot of difference’ to crop
yield and income.
THE JOURNAL OF AGRICULTURAL EDUCATION AND EXTENSION 11
This confirms the first hypothesis and confirms the more favorable findings of Elias et al.
(2013) and Khan et al. (2014), rather than, frequently older, research such as Pender, Place,
and Ehui (2006) and Lefort (2012). We also find such advice to have a varying impact
according to both the farmer’s characteristics, such as education, and the characteristics
of the farm. The pattern of significance suggests that agricultural advice yields its best
results when targeted at those with the ability to use it, that is, the better educated.
However, it is also suggestive perhaps that the advice given to less well educated people
needs to be different, made simpler or given in more detail. This helps answer the
second of our research questions and partially supports the third hypothesis, although
we find only limited evidence of impact being related to age.
There are also significant differences in the impact of advice between different areas.
This again relates to the second research question and provides confirmation of our
second hypothesis. In drought-affected areas, advice on animal husbandry and fertilizers
is most effective. In non-drought areas, advice on marketing and land management and
agricultural practices is best. The impact of marketing on crop yields is plausibly an indir-
ect one whereby farmers respond to increased prices and a greater ability to sell output by
increased effort. Our findings are consistent with our theoretical analysis which suggested
that advice had the greatest potential in areas and conditions most conducive to applying
the advice. This differential impact of advice in different areas is consistent with the results
of Kassie et al. (2010). However, they find that advice on land management practices
works best in low agricultural potential areas which is slightly at odds with our findings.
In addition, they found advice on fertilizers to have a very significant and positive impact
on crop productivity in high potential areas, whereas we found this advice worked well in
both types of areas. Why the differences with Kassie et al.? Firstly, there is the time period.
Their data relate to 1998 and 2001. Ours is more recent. Technology moves on and what
might have been the case over a decade ago may no longer be the case today. Secondly,
their analysis relates to usage of fertilizer, minimum tillage, etc. Our data relate to exten-
sion agent advice on these technologies. The advice may feasibly be to use less fertilizer, or
to use it in a different manner. Thirdly, some technologies may take several years before
they have an impact. Hence Schmidt and Tadesse (2014) conclude that that sustainable
land and water management infrastructure in the Ethiopian highlands has a positive
impact on the value of production only seven years after construction. This might be
too long a time frame for our results to be picking up. Finally, there may be a difference
between their split of high and low productivity areas and ours of drought and non-
drought areas.
In addition, more educated individuals and older people were significantly less likely to
receive advice. The kind of advice received also differed between drought and non-drought
areas and also varied with the size of the family, education and gender, suggesting that
advice is targeted rather than given randomly. Thus, this does not support the observation
made by Kassie et al. (2010) that the advice showed little variation across different
environments nor responded to household-specific factors. This again confirms part of
the third hypothesis. Thus in terms of the three hypotheses we set out to test, all three
have been largely substantiated.
Expanding further on our second research question, some advice has been found to
be more effective than others. In virtually, all the regressions, advice on credit appears
to have the least positive impact on both yields and income, especially the latter. This
12 A. HAMILTON AND J. HUDSON
may be linked to credit being heavily under the control of the government and possibly
being used for political purposes (Berhanu and Poulton 2014). Advice on credit is also
almost always given in tandem with other advice and the combined effect, as can be
seen from the coefficients in Tables 3 and 4, is generally positive, at least for yields.
Nonetheless, it does raise questions about the recent emphasis on credit (Bachewe
et al. 2015).
We did not have enough data to investigate whether different types of advice work
better in tandem. But a variable equal to the number of types of advice given, although
negative, gave only weak evidence for declining returns with respect to the amount of
advice given. Hence, advice appears best if given on several dimensions with the impact
being largely cumulative. This is consistent with Teklewold et al.’s (2013) conclusion
that different types of sustainable land practices work best when adopted in combination
rather than isolation.
There is an important policy issue in that extension system advice can be unsuccessful
either because the problem lies with the ‘extension message’ or in the way the message is
transmitted. Too often the latter tends to get the blame for lack of impact while the
problem may be with the ‘advice’ or message. In a sense that is the more serious
problem in indicating that the advice is flawed to begin with, whereas retraining can alle-
viate problems with the messenger. Our analysis may be the start of a process of determin-
ing where the problem lies, if indeed there is a problem. In this context, our results
tentatively show that with much of the advice being effective in at least some contexts
the messenger is at least partially exonerated. This also links in to the current debate on
the relative roles of R&D and technology transfer (Anandajayasekeram 2011). Our
results point to the importance of the latter, in the specific context of extension agents.
But often new knowledge comes from R&D and the roles of research centers remain of
potential importance and here perhaps the linkages between agricultural research
centers and extension agents could be improved (Abegaz and Wims 2015; Teshome, de
Graaff, and Kassie 2015). Similarly, the use of regression analysis to quantify the impact
of extension agent advice has provided results which may be potentially useful to improv-
ing and informing extension agents’ advice in part by internalizing the results of this type
of analysis within the formal education system.
Notes
1. http://www.farmingfirst.org/wordpress/wp-content/uploads/2012/06/Global-Forum-for-
Rural-Advisory-Services_Fact-Sheet-on-Extension-Services.pdf.
2. There are types of advice given in addition to the six specified, but for the purpose of the
theory we focus on the impact of these six types of advice.
3. The kebele is the lowest administrative tier in Ethiopia’s federal structure, below the woreda.
4. If the question had been on what had happened to yields and income, then just focusing on
those who received advice would be problematic. But, we cannot ask a similar question of
those who did not receive advice as the question we are analysing pertains to the impact
the advice had on yields and income. Obviously, this question cannot be asked of those
who did not receive advice.
5. The constant term has been absorbed into the cut-off points for the ordered probit
regression.
6. The coefficients cannot be readily compared as the one relates to ordered probit and the other
to probit.
THE JOURNAL OF AGRICULTURAL EDUCATION AND EXTENSION 13
7. They were however significant at times in the equations relating to policy advice, rather than
impact and have been retained in those. For example, land management advice tended to
increase with family size, which may reflect the size of the holding, and credit advice declines
with family size. There was weaker evidence of credit and marketing advice declining for
men.
Acknowledgements
We acknowledge the valuable comments of referees and the editor.
Disclosure statement
No potential conflict of interest was reported by the authors.
Funding
This work was supported by the Department for International Development (DFID).
Notes on contributors
Dr Alexander Hamilton is a political economist, and evaluation specialist with significant
academic and field experience in fragile states. He has numerous publications and research
experience in the fields of corruption, impact evaluations, public financial management,
economic policy, and econometrics. He also has field experience from work conducted
in Ethiopia, Senegal, Sudan, and Yemen. He has an MPA in Public and Economic
Policy form the London School of Economics and a DPhil (PhD) from the University
of Oxford.
Professor John Hudson studied at the Universities of London and Warwick. He is a pro-
fessor of economics at the University of Bath. He has published more than 100 journal
papers in leading journals in all areas of economics and the wider social sciences, but in
particular development economics.
References
Abegaz, D. M., and P. Wims. 2015. “Extension Agents’ Awareness of Climate Change in Ethiopia.”
The Journal of Agricultural Education and Extension 21 (5): 479–495.
Adams, R., H. Almeida, and D. Ferreira. 2009. “Understanding the Relationship Between Founder –
CEOs and Firm Performance.” Journal of Empirical Finance 16 (1): 136–150.
Anandajayasekeram, P. 2011. “The Role of Agricultural R&D within the Agricultural Innovation
Systems Framework.” In Report Prepared for the ASTI/IFPRI-FARA Conference.
Bachewe, F. N., G. Berhane, B. Minten, and A. S. Taffesse. 2015. “Agricultural Growth in Ethiopia
(2004–2014): Evidence and Drivers.” ESSP Working Paper 81. Addis Ababa: International Food
Policy Research Institute/Ethiopia Strategy Support Program (ESSP).
Berhanu, K., and C. Poulton. 2014. “The Political Economy of Agricultural Extension in Ethiopia:
Economic Growth and Political Control.” Development Policy Review 32 (S2): s197–s213.
Buchy, M., and F. Basaznew. 2005. “Gender-blind Organizations Deliver Gender-biased Services:
The Case of Awasa Bureau of Agriculture in Southern Ethiopia.” Gender, Technology and
Development 9 (2): 235–251.
14 A. HAMILTON AND J. HUDSON
Clarke, P. S., and F. Windmeijer. 2012. “Instrumental Variable Estimators for Binary Outcomes”.
Journal of the American Statistical Association 107 (500): 1638–1652.
Cohen, M. J., and M. Lemma. 2011. Agricultural Extension Services and Gender Equality. Discussion
paper. Washington, DC: International Food Policy Research Institute.
Davis, K., B. Swanson, D. Amudavi, D. A. Mekonnen, A. Flohrs, J. Riese, and E. Zerfu. 2010. In-
depth Assessment of the Public Agricultural Extension System of Ethiopia and
Recommendations for Improvement. Discussion paper. Washington, DC: International Food
Policy Research Institute.
Dercon, S., and L. Christiaensen. 2011. “Consumption Risk, Technology Adoption and Poverty
Traps: Evidence from Ethiopia.” Journal of Development Economics 96 (2): 159–173.
Dercon, S., D. O. Gilligan, J. Hoddinott, and T. Woldehanna. 2009. “The Impact of Agricultural
Extension and Roads on Poverty and Consumption Growth in Fifteen Ethiopian Villages.”
American Journal of Agricultural Economics 91 (4): 1007–1021.
EEA/EEPRI. 2006. Evaluation of the Ethiopian Agricultural Extension with Particular Emphasis on
the Participatory Demonstration and Training Extension System (PADETES). Addis Ababa:
Ethiopian Economic Association/Ethiopian Economic Policy Research Institute.
Elias, A., M. Nohmi, K. Yasunobu, and A. Ishida. 2013. “Effect of Agricultural Extension Program
on Smallholders’ Farm Productivity: Evidence from Three Peasant Associations in the Highlands
of Ethiopia.” Journal of Agricultural Science 5 (8): 163–181.
Kassie, M., P. Zikhali, J. Pender, and G. Köhlin. 2010. “The Economics of Sustainable Land
Management Practices in the Ethiopian Highlands.” Journal of Agricultural Economics 61 (3):
605–627.
Kelemework, F., and H. Kassa. 2006. Assessment of the Current Extension System of Ethiopia: A
Closer Look at Planning and Implementation. Addis Ababa: Ethiopian Economic Association/
Ethiopian Economic Policy Research Institute.
Khan, Q. M., J. P. Faguet, C. Gaukler, and W. Mekasha. 2014. Improving Basic Services for the
Bottom Forty Percent: Lessons from Ethiopia. Washington, DC: The World Bank.
Lefort, R. 2012. “Free Market Economy, ‘Developmental State’ and Party-state Hegemony in
Ethiopia: The Case of the ‘Model Farmers.’” Journal of Modern African Studies 50 (4): 681–706.
Pender, J., F. Place, and S. Ehui, eds. 2006. Strategies for Sustainable Land Management in the East
Africa Highlands. Washington, DC: International Food Policy Research Institute.
Roodman, D. 2011. “Estimating Fully Observed Recursive Mixed-process Models with CMP.” Stata
Journal 11 (2): 159–206.
Schmidt, E., and F. Tadesse. 2014. “Sustainable Agriculture in the Blue Nile Basin: Land and
Watershed Management Practices in Ethiopia.” Environment and Development Economics 19
(5): 648–667.
Spielman, D. J., D. Byerlee, D. Alemu, and D. Kelemework. 2010. “Policies to Promote Cereal
Intensification in Ethiopia: The Search for Appropriate Public and Private Roles.” Food Policy
35 (3): 185–194.
Teklewold, H., M. Kassie, B. Shiferaw, and G. Köhlin 2013. “Cropping System Diversification,
Conservation Tillage and Modern Seed Adoption in Ethiopia: Impacts on Household Income,
Agrochemical Use and Demand for Labor.” Ecological Economics 93: 85–93.
Teshome, A., J. de Graaff, and M. Kassie 2015. “Household-level Determinants of Soil and Water
Conservation Adoption Phases: Evidence from North-Western Ethiopian Highlands.”
Environmental Management 57 (3): 620–636.
THE JOURNAL OF AGRICULTURAL EDUCATION AND EXTENSION 15

More Related Content

What's hot

The Neglected Dimension of Global Security: A Framework to Counter Infectious...
The Neglected Dimension of Global Security: A Framework to Counter Infectious...The Neglected Dimension of Global Security: A Framework to Counter Infectious...
The Neglected Dimension of Global Security: A Framework to Counter Infectious...The Rockefeller Foundation
 
The Augmented Sustainable Livelihood Framework (aSLF) -EJ Mensah
The Augmented Sustainable Livelihood Framework (aSLF) -EJ MensahThe Augmented Sustainable Livelihood Framework (aSLF) -EJ Mensah
The Augmented Sustainable Livelihood Framework (aSLF) -EJ MensahEmmanuel Joseph Mensah
 
Political Economy of Vietnam's response to Avian Influenza by Tuong Vu
Political Economy of Vietnam's response to Avian Influenza by Tuong VuPolitical Economy of Vietnam's response to Avian Influenza by Tuong Vu
Political Economy of Vietnam's response to Avian Influenza by Tuong VuSTEPS Centre
 
P1833 Sustainable Rural Livelihoods Ids Paper72
P1833 Sustainable Rural Livelihoods Ids Paper72P1833 Sustainable Rural Livelihoods Ids Paper72
P1833 Sustainable Rural Livelihoods Ids Paper72guest833725e7
 
Vulnerable Groups and Communities in The Context of Adaptation and Developmen...
Vulnerable Groups and Communities in The Context of Adaptation and Developmen...Vulnerable Groups and Communities in The Context of Adaptation and Developmen...
Vulnerable Groups and Communities in The Context of Adaptation and Developmen...Tariq A. Deen
 
An introduction to the sustainable livelihoods framework
An introduction to the sustainable livelihoods frameworkAn introduction to the sustainable livelihoods framework
An introduction to the sustainable livelihoods frameworkafrica-rising
 
A mircofinancing-framework-report
A mircofinancing-framework-reportA mircofinancing-framework-report
A mircofinancing-framework-reportUNU-MERIT
 
767600 jrn0wbro00box374387b00public0
767600 jrn0wbro00box374387b00public0767600 jrn0wbro00box374387b00public0
767600 jrn0wbro00box374387b00public0Arvind Rathod
 
A Guide to the Livelihoods Framework
A Guide to the Livelihoods FrameworkA Guide to the Livelihoods Framework
A Guide to the Livelihoods FrameworkOlivier Serrat
 
Towards improving service delivery in local authorities. A case of Chegutu Mu...
Towards improving service delivery in local authorities. A case of Chegutu Mu...Towards improving service delivery in local authorities. A case of Chegutu Mu...
Towards improving service delivery in local authorities. A case of Chegutu Mu...iosrjce
 
Syringe access landscape, 2010 for USCA
Syringe access landscape, 2010 for USCASyringe access landscape, 2010 for USCA
Syringe access landscape, 2010 for USCADaniel Raymond
 

What's hot (20)

The Neglected Dimension of Global Security: A Framework to Counter Infectious...
The Neglected Dimension of Global Security: A Framework to Counter Infectious...The Neglected Dimension of Global Security: A Framework to Counter Infectious...
The Neglected Dimension of Global Security: A Framework to Counter Infectious...
 
Velasquez poster
Velasquez posterVelasquez poster
Velasquez poster
 
The Augmented Sustainable Livelihood Framework (aSLF) -EJ Mensah
The Augmented Sustainable Livelihood Framework (aSLF) -EJ MensahThe Augmented Sustainable Livelihood Framework (aSLF) -EJ Mensah
The Augmented Sustainable Livelihood Framework (aSLF) -EJ Mensah
 
Health Vulnerabilities of Informal Workers
Health Vulnerabilities of Informal WorkersHealth Vulnerabilities of Informal Workers
Health Vulnerabilities of Informal Workers
 
Alan Nicol Keynote Speech
Alan Nicol Keynote SpeechAlan Nicol Keynote Speech
Alan Nicol Keynote Speech
 
Comments on "Is Corruption Good for your Health?"
Comments on "Is Corruption Good for your Health?"Comments on "Is Corruption Good for your Health?"
Comments on "Is Corruption Good for your Health?"
 
Political Economy of Vietnam's response to Avian Influenza by Tuong Vu
Political Economy of Vietnam's response to Avian Influenza by Tuong VuPolitical Economy of Vietnam's response to Avian Influenza by Tuong Vu
Political Economy of Vietnam's response to Avian Influenza by Tuong Vu
 
UHLIVELIHOODS
UHLIVELIHOODSUHLIVELIHOODS
UHLIVELIHOODS
 
P1833 Sustainable Rural Livelihoods Ids Paper72
P1833 Sustainable Rural Livelihoods Ids Paper72P1833 Sustainable Rural Livelihoods Ids Paper72
P1833 Sustainable Rural Livelihoods Ids Paper72
 
Vulnerable Groups and Communities in The Context of Adaptation and Developmen...
Vulnerable Groups and Communities in The Context of Adaptation and Developmen...Vulnerable Groups and Communities in The Context of Adaptation and Developmen...
Vulnerable Groups and Communities in The Context of Adaptation and Developmen...
 
An introduction to the sustainable livelihoods framework
An introduction to the sustainable livelihoods frameworkAn introduction to the sustainable livelihoods framework
An introduction to the sustainable livelihoods framework
 
Livelihood mapping
Livelihood mappingLivelihood mapping
Livelihood mapping
 
A mircofinancing-framework-report
A mircofinancing-framework-reportA mircofinancing-framework-report
A mircofinancing-framework-report
 
767600 jrn0wbro00box374387b00public0
767600 jrn0wbro00box374387b00public0767600 jrn0wbro00box374387b00public0
767600 jrn0wbro00box374387b00public0
 
A Guide to the Livelihoods Framework
A Guide to the Livelihoods FrameworkA Guide to the Livelihoods Framework
A Guide to the Livelihoods Framework
 
Towards improving service delivery in local authorities. A case of Chegutu Mu...
Towards improving service delivery in local authorities. A case of Chegutu Mu...Towards improving service delivery in local authorities. A case of Chegutu Mu...
Towards improving service delivery in local authorities. A case of Chegutu Mu...
 
Help Wanted: Creating Equal Opportunities for Minorities
Help Wanted: Creating Equal Opportunities for MinoritiesHelp Wanted: Creating Equal Opportunities for Minorities
Help Wanted: Creating Equal Opportunities for Minorities
 
Syringe access landscape, 2010 for USCA
Syringe access landscape, 2010 for USCASyringe access landscape, 2010 for USCA
Syringe access landscape, 2010 for USCA
 
Sas overview logo
Sas overview logoSas overview logo
Sas overview logo
 
Livelihood concept
Livelihood conceptLivelihood concept
Livelihood concept
 

Viewers also liked (20)

Business financials
Business financialsBusiness financials
Business financials
 
SamsatOLslide
SamsatOLslideSamsatOLslide
SamsatOLslide
 
Globpr2
Globpr2Globpr2
Globpr2
 
Suunto ambit2 s_user_guide_en
Suunto ambit2 s_user_guide_enSuunto ambit2 s_user_guide_en
Suunto ambit2 s_user_guide_en
 
Seguridad en implementaciones de voz sobre ip paper
Seguridad en implementaciones de voz sobre ip   paperSeguridad en implementaciones de voz sobre ip   paper
Seguridad en implementaciones de voz sobre ip paper
 
N stf08 002-3
N   stf08 002-3N   stf08 002-3
N stf08 002-3
 
Emotional Intelligence
Emotional IntelligenceEmotional Intelligence
Emotional Intelligence
 
Catalogo de elementos culturales 2
Catalogo de elementos culturales 2Catalogo de elementos culturales 2
Catalogo de elementos culturales 2
 
Evolución a lo largo del tiempo de Windows y Linux
Evolución a lo largo del tiempo de Windows y LinuxEvolución a lo largo del tiempo de Windows y Linux
Evolución a lo largo del tiempo de Windows y Linux
 
Acupuncture techniques
Acupuncture  techniquesAcupuncture  techniques
Acupuncture techniques
 
Νέος Αναπτυξιακός Νόμος 4399/2016
Νέος Αναπτυξιακός Νόμος 4399/2016Νέος Αναπτυξιακός Νόμος 4399/2016
Νέος Αναπτυξιακός Νόμος 4399/2016
 
Antonio Zarragosa's CV
Antonio Zarragosa's CVAntonio Zarragosa's CV
Antonio Zarragosa's CV
 
ESPÍRITAS NO EVANGELHO
ESPÍRITAS NO EVANGELHOESPÍRITAS NO EVANGELHO
ESPÍRITAS NO EVANGELHO
 
RESUMEN DE CLASE
RESUMEN DE CLASERESUMEN DE CLASE
RESUMEN DE CLASE
 
Ornamente de paste
Ornamente de pasteOrnamente de paste
Ornamente de paste
 
njkmnkl
njkmnklnjkmnkl
njkmnkl
 
Recuperacion final
Recuperacion finalRecuperacion final
Recuperacion final
 
Aranjamente masa de paste
Aranjamente masa de pasteAranjamente masa de paste
Aranjamente masa de paste
 
Biosíntesis de grasas
Biosíntesis de grasasBiosíntesis de grasas
Biosíntesis de grasas
 
Paleta
PaletaPaleta
Paleta
 

Similar to The perceived impact of agricultural advice in Ethiopia

Environmental sustainability and financial development in Africa; does instit...
Environmental sustainability and financial development in Africa; does instit...Environmental sustainability and financial development in Africa; does instit...
Environmental sustainability and financial development in Africa; does instit...matthew ntow-gyamfi
 
Raising Returns, Managing Risk: A Randomized Experiment on Combining Input Su...
Raising Returns, Managing Risk: A Randomized Experiment on Combining Input Su...Raising Returns, Managing Risk: A Randomized Experiment on Combining Input Su...
Raising Returns, Managing Risk: A Randomized Experiment on Combining Input Su...Stockholm Institute of Transition Economics
 
What Constrains Smallholder Farmers Decisions to Participate and use Agricult...
What Constrains Smallholder Farmers Decisions to Participate and use Agricult...What Constrains Smallholder Farmers Decisions to Participate and use Agricult...
What Constrains Smallholder Farmers Decisions to Participate and use Agricult...ijtsrd
 
An Overview of the Challenges of Decentralized Agricultural Extension Practic...
An Overview of the Challenges of Decentralized Agricultural Extension Practic...An Overview of the Challenges of Decentralized Agricultural Extension Practic...
An Overview of the Challenges of Decentralized Agricultural Extension Practic...Premier Publishers
 
Leisa Perch: Reconciling participation and benefits-sharing - policy implicat...
Leisa Perch: Reconciling participation and benefits-sharing - policy implicat...Leisa Perch: Reconciling participation and benefits-sharing - policy implicat...
Leisa Perch: Reconciling participation and benefits-sharing - policy implicat...AfricaAdapt
 
Reconciling Participation And Benefits Sharing 1
Reconciling Participation And Benefits Sharing 1Reconciling Participation And Benefits Sharing 1
Reconciling Participation And Benefits Sharing 1Ln Perch
 
RUNNING Head IMPACTS ON FOOD SYSTEMS. .docx
RUNNING Head IMPACTS ON FOOD SYSTEMS.                            .docxRUNNING Head IMPACTS ON FOOD SYSTEMS.                            .docx
RUNNING Head IMPACTS ON FOOD SYSTEMS. .docxwlynn1
 
Ethiopia Health Sector Financing Reform/HFG: End-of-Project Report
Ethiopia Health Sector Financing Reform/HFG: End-of-Project ReportEthiopia Health Sector Financing Reform/HFG: End-of-Project Report
Ethiopia Health Sector Financing Reform/HFG: End-of-Project ReportHFG Project
 
Intersections between Poverty, Environment and Inclusive Growth: A Global Per...
Intersections between Poverty, Environment and Inclusive Growth: A Global Per...Intersections between Poverty, Environment and Inclusive Growth: A Global Per...
Intersections between Poverty, Environment and Inclusive Growth: A Global Per...UNDP Policy Centre
 
China’s economy中国经济httpworldmap.harvard.educhinamap.docx
China’s economy中国经济httpworldmap.harvard.educhinamap.docxChina’s economy中国经济httpworldmap.harvard.educhinamap.docx
China’s economy中国经济httpworldmap.harvard.educhinamap.docxspoonerneddy
 
China’s economy中国经济httpworldmap.harvard.educhinamap.docx
China’s economy中国经济httpworldmap.harvard.educhinamap.docxChina’s economy中国经济httpworldmap.harvard.educhinamap.docx
China’s economy中国经济httpworldmap.harvard.educhinamap.docxmccormicknadine86
 
The adverse impact of ecosystem degradation and poor governance on marginaliz...
The adverse impact of ecosystem degradation and poor governance on marginaliz...The adverse impact of ecosystem degradation and poor governance on marginaliz...
The adverse impact of ecosystem degradation and poor governance on marginaliz...NAAR Journal
 
Investing in Agriculture and Social Protection for Halving Hunger and Poverty
Investing in Agriculture and Social Protection for Halving Hunger and PovertyInvesting in Agriculture and Social Protection for Halving Hunger and Poverty
Investing in Agriculture and Social Protection for Halving Hunger and PovertyShenggen Fan
 
Influence of risk taking propensity among kenyan community health workers
Influence of risk taking propensity among kenyan community health workersInfluence of risk taking propensity among kenyan community health workers
Influence of risk taking propensity among kenyan community health workersfredrickaila
 
HFG Rapid Assessment of TB Payment and PFM Systems in Cambodia: Lessons Learn...
HFG Rapid Assessment of TB Payment and PFM Systems in Cambodia: Lessons Learn...HFG Rapid Assessment of TB Payment and PFM Systems in Cambodia: Lessons Learn...
HFG Rapid Assessment of TB Payment and PFM Systems in Cambodia: Lessons Learn...HFG Project
 
Poverty alleviation and the environmental governance
Poverty alleviation and the environmental governancePoverty alleviation and the environmental governance
Poverty alleviation and the environmental governancePari Doll
 
Factors Influencing Adoption of Improved Agricultural Technologies (IATs) amo...
Factors Influencing Adoption of Improved Agricultural Technologies (IATs) amo...Factors Influencing Adoption of Improved Agricultural Technologies (IATs) amo...
Factors Influencing Adoption of Improved Agricultural Technologies (IATs) amo...Premier Publishers
 

Similar to The perceived impact of agricultural advice in Ethiopia (20)

Environmental sustainability and financial development in Africa; does instit...
Environmental sustainability and financial development in Africa; does instit...Environmental sustainability and financial development in Africa; does instit...
Environmental sustainability and financial development in Africa; does instit...
 
Raising Returns, Managing Risk: A Randomized Experiment on Combining Input Su...
Raising Returns, Managing Risk: A Randomized Experiment on Combining Input Su...Raising Returns, Managing Risk: A Randomized Experiment on Combining Input Su...
Raising Returns, Managing Risk: A Randomized Experiment on Combining Input Su...
 
What Constrains Smallholder Farmers Decisions to Participate and use Agricult...
What Constrains Smallholder Farmers Decisions to Participate and use Agricult...What Constrains Smallholder Farmers Decisions to Participate and use Agricult...
What Constrains Smallholder Farmers Decisions to Participate and use Agricult...
 
An Overview of the Challenges of Decentralized Agricultural Extension Practic...
An Overview of the Challenges of Decentralized Agricultural Extension Practic...An Overview of the Challenges of Decentralized Agricultural Extension Practic...
An Overview of the Challenges of Decentralized Agricultural Extension Practic...
 
Operational challenges in africa
Operational challenges in africa Operational challenges in africa
Operational challenges in africa
 
Leisa Perch: Reconciling participation and benefits-sharing - policy implicat...
Leisa Perch: Reconciling participation and benefits-sharing - policy implicat...Leisa Perch: Reconciling participation and benefits-sharing - policy implicat...
Leisa Perch: Reconciling participation and benefits-sharing - policy implicat...
 
Reconciling Participation And Benefits Sharing 1
Reconciling Participation And Benefits Sharing 1Reconciling Participation And Benefits Sharing 1
Reconciling Participation And Benefits Sharing 1
 
RUNNING Head IMPACTS ON FOOD SYSTEMS. .docx
RUNNING Head IMPACTS ON FOOD SYSTEMS.                            .docxRUNNING Head IMPACTS ON FOOD SYSTEMS.                            .docx
RUNNING Head IMPACTS ON FOOD SYSTEMS. .docx
 
Ethiopia Health Sector Financing Reform/HFG: End-of-Project Report
Ethiopia Health Sector Financing Reform/HFG: End-of-Project ReportEthiopia Health Sector Financing Reform/HFG: End-of-Project Report
Ethiopia Health Sector Financing Reform/HFG: End-of-Project Report
 
Intersections between Poverty, Environment and Inclusive Growth: A Global Per...
Intersections between Poverty, Environment and Inclusive Growth: A Global Per...Intersections between Poverty, Environment and Inclusive Growth: A Global Per...
Intersections between Poverty, Environment and Inclusive Growth: A Global Per...
 
China’s economy中国经济httpworldmap.harvard.educhinamap.docx
China’s economy中国经济httpworldmap.harvard.educhinamap.docxChina’s economy中国经济httpworldmap.harvard.educhinamap.docx
China’s economy中国经济httpworldmap.harvard.educhinamap.docx
 
China’s economy中国经济httpworldmap.harvard.educhinamap.docx
China’s economy中国经济httpworldmap.harvard.educhinamap.docxChina’s economy中国经济httpworldmap.harvard.educhinamap.docx
China’s economy中国经济httpworldmap.harvard.educhinamap.docx
 
APR Workshop 2010- S&S Cooperation-Public Services and Agricultural Developme...
APR Workshop 2010- S&S Cooperation-Public Services and Agricultural Developme...APR Workshop 2010- S&S Cooperation-Public Services and Agricultural Developme...
APR Workshop 2010- S&S Cooperation-Public Services and Agricultural Developme...
 
The adverse impact of ecosystem degradation and poor governance on marginaliz...
The adverse impact of ecosystem degradation and poor governance on marginaliz...The adverse impact of ecosystem degradation and poor governance on marginaliz...
The adverse impact of ecosystem degradation and poor governance on marginaliz...
 
Investing in Agriculture and Social Protection for Halving Hunger and Poverty
Investing in Agriculture and Social Protection for Halving Hunger and PovertyInvesting in Agriculture and Social Protection for Halving Hunger and Poverty
Investing in Agriculture and Social Protection for Halving Hunger and Poverty
 
Influence of risk taking propensity among kenyan community health workers
Influence of risk taking propensity among kenyan community health workersInfluence of risk taking propensity among kenyan community health workers
Influence of risk taking propensity among kenyan community health workers
 
HFG Rapid Assessment of TB Payment and PFM Systems in Cambodia: Lessons Learn...
HFG Rapid Assessment of TB Payment and PFM Systems in Cambodia: Lessons Learn...HFG Rapid Assessment of TB Payment and PFM Systems in Cambodia: Lessons Learn...
HFG Rapid Assessment of TB Payment and PFM Systems in Cambodia: Lessons Learn...
 
INNOVATIVE FINANCING AND INVESTMENT IN AGRICULTURE: AFRICA’S EXPERIENCE
INNOVATIVE FINANCING AND INVESTMENT IN AGRICULTURE: AFRICA’S EXPERIENCEINNOVATIVE FINANCING AND INVESTMENT IN AGRICULTURE: AFRICA’S EXPERIENCE
INNOVATIVE FINANCING AND INVESTMENT IN AGRICULTURE: AFRICA’S EXPERIENCE
 
Poverty alleviation and the environmental governance
Poverty alleviation and the environmental governancePoverty alleviation and the environmental governance
Poverty alleviation and the environmental governance
 
Factors Influencing Adoption of Improved Agricultural Technologies (IATs) amo...
Factors Influencing Adoption of Improved Agricultural Technologies (IATs) amo...Factors Influencing Adoption of Improved Agricultural Technologies (IATs) amo...
Factors Influencing Adoption of Improved Agricultural Technologies (IATs) amo...
 

More from Alexander Hamilton, PhD

More from Alexander Hamilton, PhD (6)

Evidence Base Programme
Evidence Base ProgrammeEvidence Base Programme
Evidence Base Programme
 
improving-tax-systems
improving-tax-systemsimproving-tax-systems
improving-tax-systems
 
1-s2.0-S1877584515000519-main
1-s2.0-S1877584515000519-main1-s2.0-S1877584515000519-main
1-s2.0-S1877584515000519-main
 
61506_Capstone_Report_DFID_FINAL_Quantifying_Governance__Indicators-4
61506_Capstone_Report_DFID_FINAL_Quantifying_Governance__Indicators-461506_Capstone_Report_DFID_FINAL_Quantifying_Governance__Indicators-4
61506_Capstone_Report_DFID_FINAL_Quantifying_Governance__Indicators-4
 
comparing-corruption-ethiopia-sudan-4
comparing-corruption-ethiopia-sudan-4comparing-corruption-ethiopia-sudan-4
comparing-corruption-ethiopia-sudan-4
 
31-14
31-1431-14
31-14
 

The perceived impact of agricultural advice in Ethiopia

  • 1. Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=raee20 Download by: [Research & Evidence Division] Date: 24 October 2016, At: 00:15 The Journal of Agricultural Education and Extension Competence for Rural Innovation and Transformation ISSN: 1389-224X (Print) 1750-8622 (Online) Journal homepage: http://www.tandfonline.com/loi/raee20 The perceived impact of agricultural advice in Ethiopia Alexander Hamilton & John Hudson To cite this article: Alexander Hamilton & John Hudson (2016): The perceived impact of agricultural advice in Ethiopia, The Journal of Agricultural Education and Extension, DOI: 10.1080/1389224X.2016.1245151 To link to this article: http://dx.doi.org/10.1080/1389224X.2016.1245151 Published online: 23 Oct 2016. Submit your article to this journal View related articles View Crossmark data
  • 2. The perceived impact of agricultural advice in Ethiopia Alexander Hamiltona and John Hudsonb a Department for International Development, Sudan, BFPO 5312, Ruislip, UK; b Department of Economics, University of Bath, Bath, UK ABSTRACT Purpose: We examine the impact of advice given by extension agents to Ethiopian farmers, as perceived by the farmers themselves. Design/methodology/approach: Using survey data from 2014, we analyze the perceived impact of advice on farmers’ incomes and crop yields. We use a bootstrapped instrumental variable (IV) estimator and the conditional mixed process estimator. Theoretical implications: The impact of advice will depend upon its relevance and whether and how efficiently it is implemented by the farmer. This in part depends upon the farmer’s ability and on the impact of fully implemented advice on output, which will vary from farm to farm. Findings: There is a positive perceived impact of most advice on both crop yields and income. However, some advice works better in drought-affected areas and other in non-drought-affected areas. Fertilizers have more impact on crop yields than income, possibly reflecting cost factors. There is evidence that the farmers’ ability to implement the advice increases with their level of education and that advice is being tailored to the needs of the individual. Practical implications: Advice has a positive impact on both crop yields and income. However, not all advice is equally effective and effectiveness varies according to farmer and farm characteristics. There is little evidence of credit advice having a positive impact. Originality/value: The paper is one of only a few to analyze farmers’ perceptions of advice impact and, as far as we are aware, is the first to analyze how advice effectiveness varies according to farmer and farm characteristics. ARTICLE HISTORY Received 24 August 2015 Accepted 3 October 2016 KEYWORDS Extension agents; farmers’ income; crop yields; fertilizer advice; land management advice; credit advice Introduction A core part of the Ethiopian government’s investment in agriculture is the public agricul- tural extension system, with a substantial increase in the number of people graduating as extension or development agents (DAs) (Davis et al. 2010). The Government has also established farmer training centers, with two hectares of demonstration fields (Lefort 2012), in every local administrative area (there are 18,000 nationwide) with three exten- sion agents at every training center. This effort could lead to Ethiopia having the world’s highest ratio of extension agents to farmers1 and is reflected in the proportion of fields that use extension services rising from just over 5% in 2008 to 12% in 2011 (Khan et al. 2014). © 2016 Wageningen University CONTACT John Hudson j.r.hudson@bath.ac.uk Department of Economics, University of Bath, Bath, BA2 7AY, UK JOURNAL OF AGRICULTURAL EDUCATION AND EXTENSION, 2016 http://dx.doi.org/10.1080/1389224X.2016.1245151
  • 3. There is evidence of an impact on agricultural efficiency. Dercon et al. (2009) showed that one visit from a DA raised production growth by 7% and reduced poverty by 10%. The posting of agricultural extension agents in local communities has improved their attentiveness to farmers’ needs and constraints, and enhanced the working relationship between them (Cohen and Lemma 2011). However, others have tended to be more scep- tical about the impact of extension agents (Lefort 2012; Pender, Place, and Ehui 2006). Davis et al. (2010) also note that agricultural productivity remains low, inputs are scarce and expensive, and market and credit access are extremely limited and according to Buchy and Basaznew (2005), women-focused extension is also limited. In general, Ethiopia has tended to be viewed unfavorably in terms of agricultural pro- ductivity (Dercon and Christiaensen 2011; Spielman et al. 2010). However in recent years, certainly since 2008, there has been a substantial, even dramatic, improvement in cereal yields. This has been accompanied by a 12.5% expansion in land under cereal production between 2003 and 2012; hence it is not the case that more marginal land has disappeared from the picture, raising the average productivity of what is left. Apart from the extensive changes to the extension system, other changes impacting on Ethiopian agriculture include substantial transport improvements which between 2000 and 2013 tripled the length of all-weather surface roads, and a rapid increase in the urban population of some 3.7 million (Bachewe et al. 2015). Together these have increased the market for com- mercial crops and partly as a consequence output–input price ratios have increased sub- stantially. There has also been increased access to credit, which saw the number of active borrowers rise from about half a million in 2003 to approximately 3.5 million in 2014. In this paper, we will be focusing on the potential role of the extension system in explain- ing this recent success. Specifically, we analyze perceptions of the difference that the advice made to the recipients of that advice with respect to (i) crop yields and (ii) income. Self- perceptions data have the advantage that the focus is on the difference made by the exten- sion service advice, whereas actual output and income can change for numerous reasons unconnected with that advice. To include all the factors that might affect yields and income is difficult and such data are also likely to be subject to measurement errors. We assume that the farmers in evaluating the impact of the advice take account of all these other factors. In addition, because the dependent variable relates to the impact of advice, we can examine how that impact varies with variables such as the farmer’s age, education and water resources. Other studies have tended to focus on the impact of variables such as education on productivity, but not on the effectiveness of advice. The main research question is whether extension advice benefits farmers both in terms of their crop yields and income. We expect this to be the case, but the literature is ambig- uous on the issue. A secondary question is what types of advice work best and in what context. There is relatively little literature on this. However, there is a literature which suggests that different types of technology, for example, fertilizers, work better in some conditions rather than others (Dercon and Christiaensen 2011; Kassie et al. 2010). This is of relevance, but advice on implementing a technology is different to the implemen- tation of that technology. In addition part of this research relates to more than a decade ago and, as already noted, in this time the extension system has substantially changed, technology has changed and Ethiopia has changed. Hence, there is a need for new research in this area. 2 A. HAMILTON AND J. HUDSON
  • 4. We will be analyzing the impact of advice on (i) agricultural practices, (ii) land manage- ment, (iii) fertilizers, (iv) marketing, (v) access to credit facilities and (vi) animal husban- dry practices. We find that extension agent advice does lead to a perceived positive impact on both crop yields and income. However, the effectiveness of the advice depends upon the level of education of the farmer. It also differs between drought-affected and non- drought-affected lands. The paper proceeds as follows. In the next section we discuss the relevant literature. We then discuss, from a theoretical perspective, how advice might impact on farmers and other methodological issues, and we also present the data. The penultimate section presents the results and finally we conclude the paper. Background The literature on the extension program has been somewhat ambivalent in terms of its impact. Extension services generally have positive impacts on nutrition and poverty reduction (Dercon et al. 2009). However, their success has been said to be constrained by weaknesses elsewhere in the system. Hence EEA/EEPRI (2006) argue that distribution channels and institutions are flawed, the formal seed system has weaknesses, and there is a lack of markets, both for inputs and outputs. Agents transfer knowledge to farmers, with relatively little knowledge flow in the reverse direction (Buchy and Basaznew 2005), which can lead to the knowledge not being tailored to the farmer’s needs. The literature argues that the extension system has focused on the distribution of standard packages to farmers, including seeds and commercial fertilizer, credit, soil and water conservation, livestock and training. Efforts to promote other aspects of sustainable land management have con- centrated on soil erosion without consideration of the underlying socioeconomic reasons for low soil productivity (Kassie et al. 2010). As a consequence, advice has been given which has been unprofitable, risky or irrelevant given the farmer’s resource constraints (Pender, Place, and Ehui 2006). Lefort (2012) cites research which concludes the extension program was not much use to farmers. However, Spielman et al. (2010) also note that a series of reforms have been made to redress these weaknesses. Berhanu and Poulton (2014) argue that the extension system is used to promote gov- ernment control, with extension workers facilitating scrutiny and control of activities. Extension workers engage in non-extension activities, such as administration, credit repayment and tax collection (Kelemework and Kassa 2006). Apart from diverting their attention from extension services, this can also strain relationships with local farmers. In addition, it is claimed that in their allocation of seeds, fertilizers and credit, extension workers prioritize farmers loyal to the governing coalition. It is also important to stress that extension agents are not always transmitting knowledge and advice which can immediately add to the farmer’s productivity. They may also be concerned with societal impact. For example, in some cases the advice relates to environmental factors (Abegaz and Wims 2015) for which the farmer may perceive little personal benefit, although this is not to say that such benefit is absent. More recent evidence is a little more positive, although still emphasizing that practices could be better. Elias et al. (2013) found that participation in the extension program increased productivity by about 20%. Other factors which influenced productivity included age, male head of household and plot characteristics. Despite this, crop yields were below the targets set by the extension program. Khan et al. (2014) conclude that THE JOURNAL OF AGRICULTURAL EDUCATION AND EXTENSION 3
  • 5. woreda-level spending on agricultural extension workers is associated with higher yields for major crops and increases the probability that farmers will improve their farming techniques. There is relatively little evidence on the relative impact of different forms of advice, but some related work has been done on the different constraints facing, and the different technologies used by, farmers. One of the reasons Dercon and Christiaensen (2011) gave for the poor performance of Ethiopian agriculture was lack of fertilizer use. Lack of knowledge and skills in adopting modern inputs was only a very minor constraint in 1999, which would suggest only a limited role for extension agent advice in this respect. Nor was lack of credit deemed a major factor. Fertilized plots were characterized by greater yields than non-fertilized plots, although not in periods of extreme droughts and floods. Thus, Dercon and Christiaensen argued that fertilizer use is a high return, but high risk technology. Kassie et al. (2010) find evidence of a strong impact of land man- agement practices on agricultural productivity in the low agricultural potential areas. In the high agricultural potential region, however, fertilizers have a very significant and posi- tive impact on crop productivity, whereas land management practices have no significant impact. Fertilizers may be less profitable in such areas due to a lack of soil moisture. Hence, their analysis raises the important point that the impact of different forms of advice and increasing knowledge may not be the same in all areas, but vary according to local conditions. Methods We will be analyzing the impact of advice on both crop yields and inccome. For the moment, we focus on the impact on crop yields (Y), although the analysis for income follows a similar path. We assume Y to be a function of resources, which are in turn a func- tion of knowledge: Yit = Aig(Sit, Lit,Kit, Fit, Wit), (1) where i denotes the ith farmer and t the time period. Ai denotes overall efficiency with which the different factors of production are used, that is, it is total factor productivity (TFP) defined at the level of the individual farmer. In our analysis, we assume that this is the vehicle by which extension agent advice impacts on output. g(.) can be regarded as the basic output of the farm independent of the characteristics and expertise of the farmer. The production function is composed of land (S), labor (L), capital (K), fertilizer (F) and water (W). The impact of extension agent advice to farmer i (Ei) on output is then given by ∂Yi ∂Ai ∂Ai ∂Ei Ei = 6 j=1 ∂Yi ∂Ai ∂Ai ∂Eij Eij = g(Sit, Lit,Kit, Fit, Wit) 6 j=1 ∂A ∂Ej Eij. (2) This is the combined impact of the six specific types of advice in our analysis.2 Eij is a con- tinuous measure of the advice given to individual i on advice type j, although we only have a discrete measure of this, which takes a value of one if advice was given. It has a lower bound of zero, which is operative when advice of this type is not given. The marginal impact of any one piece of advice is composed of ∂Ai/∂Eij, the impact of this advice on 4 A. HAMILTON AND J. HUDSON
  • 6. TFP, and secondly, the impact of TFP on output, which is from (1) just g(.). ∂Ai/∂Eij will depend upon the characteristics of the farm (T). It may also depend upon the character- istics of the farmer, with more knowledgeable farmers more able to implement the advice efficiently, although at the same time more knowledgeable farmers may be less likely to seek advice. In this case, we link knowledge to education (Ed) and age, the latter through learning by doing. Thus, ∂Ai ∂Eij = fj(Edi, Agei, Ti, Wi). (3) The dependent variable is coded 1 if the extension agent advice was perceived as making no difference, 2 if it made some difference and 3 if it made a lot of difference. The response lies in the kth category if: ak−1 , g(.) 6 j=1 fj(.)Eij , ak; k = 1, . . . , m. (4) In our analysis m = 3. Note that α0 = −∞ and αm = +∞, and hence we estimate just a1 and a2. Define Zi,k = 1 if g(.) 6 j=1 fj(.)Eij is in the kth category, and otherwise Zi,k = 0. Lin- earizing g(.) 6 j=1 fj(.)) we can estimate both the coefficients and the dividing points (αk) between the different categories by ordered probit. The independent variables will include farm characteristics, individual characteristics and dummy variables operative if a particu- lar type of advice was given. This suggests a number of hypotheses which we will be testing: H1: Extension agent advice impacts positively on both crop yields and income. H2: The degree of the impact for different types of advice will depend upon the characteristics of the farmer and the characteristics of their farm. H3: The degree of impact of this advice will be greater for more educated individuals and also for older individuals, although we can expect these individuals to be less likely to receive advice. H1 reflects our first research question, on whether extension advice benefits farmers in terms of both income and crop yields. The other two hypotheses relate to our second research question as to what types of advice work best and under which conditions. Our theoretical analysis has helped inform both of our research questions. The advice will make a difference if it is relevant advice and if the farmer, given their own and their farm’s characteristics, will implement it efficiently. The analysis also emphasizes that different types of advice will have different impacts, which are linked to differences in ∂Ai/∂Eij. Data The data were obtained from the Woreda and City Benchmarking Survey (WCBS) col- lected in 2014 using a multi-stage stratified sampling approach based on the remoteness and food security levels of households. Within each region the sub-sample size was deter- mined by population (based on census data). Data were collected on 326 kebele in 48 THE JOURNAL OF AGRICULTURAL EDUCATION AND EXTENSION 5
  • 7. woreda covering the whole of the country.3 In total 7429 individuals were interviewed. This survey is focused on rural areas. The variables we use are defined in Table 1. Infor- mation on the characteristics of the plot relates to whether the individual grew crops and had animals. Most people sampled (60.8%) raised both animals and grew crops, a substan- tial proportion just grew crops (31.3%) and an even smaller proportion just had animals (7.9%). The proportions receiving one, two, three, four, five and six types of advice as listed in the appendix were 27%, 16%, 17%, 14%, 5% and 13%, respectively. Other advice is feasible, which is why 8% received none of the types of advice specified. Table 2 shows the summary data as it varies across individual characteristics. It is noticeable that the young tend to be more likely recipients of advice, possibly reflecting that they will have learnt less ‘by doing’. It is also noticeable how the highly educated are more likely to have received advice on marketing and credit. The final two columns relate to the impact this advice, in general, has had on crop yields and the individual’s income.4 The responses ranged from 1 (none) to 3 (a lot). Thus, we assume that output and income cannot fall as a result of the advice received. The responses to both questions were fairly enthusiastic, although very slightly more so for crop yields than income. The biggest gainers in both respects tend to be the better educated. An important question relates to the role of the advice apparently not included for example, mechanization; crop protection measures etc.? To an extent they may be subsumed within one of the other categories, particularly agricultural practices, which as we note from Table 2 was the most frequently cited form of advice. Any advice which still falls outside these cat- egories would be picked up by the constant term in the regressions. Regression results In Table 3, we present the results relating to the impact on yields.5 Column 1 shows that advice received on animal husbandry, marketing and land management were all signifi- cant at the 1% level of significance and fertilizers at the 5% level. Advice on agricultural Table 1. Data definitions. Socioeconomic, demographic variables Age Age in years Education Coded from 1 (no schooling) to 24 (degree) and 25 above degree Male Coded 1 ifa male Family size Number of people currently living in the individual’s household Plot characteristics Grows crops Coded 1 if the individual grows crops, otherwise 0 Rears animals Coded 1 if the individual rears animals, otherwise 0 Drought Coded 1 if the individual suffers from regular periods of drought in the sense of a shortage of drinking water, otherwise 0 Received advice on (coded 1 for yes and 0 no) Agricultural practices; land management; fertilizer; marketing; credit facilities; animal husbandry Impact Crops improve/Income improves The difference the above advice has made to the crop yield/income ranging from 1 (none) to 3 (a lot) Kebele based variables (average of responses of others in the individual’s locality). Regional dummy variables relate to Tigray, Amhara, Oromiya, SNNP, Binshangul Gumuz, Afar, Somali and Gambela. Because of the small numbers of people in our sample who are in Gambela we join this region together with SNNP, the two being adjacent in the south west part of the country. We also join the second smallest region, Afar, with Tigray, who are adjacent on the northern edge of Ethiopia 6 A. HAMILTON AND J. HUDSON
  • 8. practices was not significant. However being in receipt of advice on credit was significantly negative. This does not imply that output was lowered as a result of the advice. Rather, it implies that being in receipt of advice on credit increases the probability of extension agent advice having no impact. However, this does alert us to a potential problem of endogene- ity. To the extent that the farmer is the one seeking this advice, rather than being proffered it by the extension agent or some other person, then it could signal that the individual is in financial problems. At the very least it reflects an interest by the farmer in gaining access to credit. The negative sign in the regression may be picking this up. Another aspect of endo- geneity is that it is possible that the individual selected for the advice is in some way more able to use it. Because of this possibility, in the second regression we instrument the advice variables in a two-stage process. Firstly, we regress the advice variables on all the exogenous vari- ables present in column 1 plus a series of variables relating to the extent others in the indi- vidual’s kebele received different forms of advice. That is, taking credit as an example, for each individual we calculate the average number of people in their kebele, excluding them- selves, who received advice on credit. These effectively act as instruments. The second stage of the estimation takes the predicted values from these regressions and re-estimates the relationships shown in column 1. The results are shown in column 3.2. The main change is the insignificance of advice on marketing and also on credit. A Hausman test indicated that the two sets of coefficients were significantly different and hence the need for an instrumental variable (IV) approach. This two-stage estimation technique is similar to that employed by Adams, Almeida, and Ferreira (2009). The literature tends to bootstrap the standard errors (Clarke and Windmeijer 2012) and this has been done. On each of the 100 iterations of the bootstrap, both stages of the two-stage estimation tech- nique were estimated. The t statistics reported are based on these standard errors. In addition to the above, we also estimated the equations using IVs in the context of a conditional mixed process (cmp) model (Roodman 2011), that is estimating a system where the different equations can have different kinds of dependent variables. We assume joint normality of the error terms of the different equations. It is a full system tech- nique, which takes account of potential correlation between the different error terms in the equations. There are potentially eight equations in the system, one for each of the advice variables and two more for the impact variables, that is, the impact on crops and income. However, this makes considerable demands on the data and we had to simplify the equations in order to obtain estimates. Firstly, we estimated each of the policy impact Table 2. Summary data relating to individual characteristics. Animal husbandry Credit Marketing Fertilizer Land management Agricultural practices Crops improve Income improves All 0.62 0.291 0.329 0.679 0.524 0.817 2.36 2.3 Young <30 0.631 0.321 0.329 0.731 0.561 0.847 2.37 2.29 Older ≥30 0.612 0.264 0.326 0.634 0.488 0.787 2.35 2.29 Male 0.621 0.278 0.317 0.665 0.525 0.815 2.36 2.3 Highly educated 0.585 0.342 0.447 0.707 0.523 0.845 2.51 2.41 No education 0.622 0.292 0.313 0.678 0.501 0.804 2.32 2.28 Notes: The final two columns relate to the average response which varied from 1 (none) to 3 (a lot) to the difference the support has made. All other columns relate to the proportion receiving advice in the different headings. THE JOURNAL OF AGRICULTURAL EDUCATION AND EXTENSION 7
  • 9. Table 3. Regression results: impact on crop yields. Sample Probit: Full 3.1 IV 2 stage Full 3.2 Probit: Full 3.3 Drought 3.4 No drought 3.5 CMP: Full 3.6 Drought 3.7 No drought 3.8 Extension agent advice Animal husbandry 0.255** (5.35) 0.4696* (2.41) 0.4695* (2.41) 0.9789** (4.07) 0.1662 (0.52) 0.6085** (5.27) 0.6273** (4.44) 0.648** (3.22) Credit −0.2344** (4.77) 0.0069 (0.03) −0.0284 (0.12) −0.2431 (0.95) 0.9285* (2.11) −0.3592* (2.42) −0.1748 (1.03) −0.2966* (2.18) Marketing 0.4172** (8.76) 0.3156 (1.72) 0.4122* (2.38) −0.0737 (0.35) 0.9992** (2.91) 0.5683** (4.45) −0.1576 (0.93) 1.452** (13.77) Fertilizers 0.0884* (1.98) 0.8863** (3.68) 1.006** (4.24) 0.8349* (2.46) 0.7405* (2.01) 0.8513** (7.69) 0.9966** (8.28) 0.7314** (4.29) LM – land management 0.2874** (6.47) 0.6833** (3.09) AG – agricultural practices 0.0756 (1.30) 0.1408 (0.61) LM + AG 0.8374** (2.69) 0.3536 (0.73) 2.378** (4.74) 1.134** (5.82) 0.3289 (0.74) 1.222** (4.67) Individual characteristics Log age 0.2891** (4.52) 0.1094 (1.62) 0.102 (1.51) −0.0723 (0.83) 0.1991 (1.79) 0.0518 (0.78) −0.0664 (0.75) 0.1341 (1.33) Education 0.0264** (5.26) 0.0244** (4.05) 0.0249** (4.16) 0.0192* (2.34) 0.0236** (2.91) 0.0276** (5.61) 0.0262** (3.86) 0.0122 (1.67) Male −0.0538 (1.52) Log family size −0.0785 (1.78) Farm characteristics Crops 1.316** (12.81) 0.6006** (3.43) 0.4652** (3.04) 0.058 (0.26) 0.3388 (1.39) −0.2539 (1.60) −0.3369 (1.18) 0.279 (1.14) Animals −0.0251 (0.51) −0.3715** (2.91) −0.3685** (2.83) −0.3783* (2.35) −0.4832* (2.50) −0.1909* (2.16) −0.2073 (1.94) −0.1815 (1.22) Constant −1.562** (5.20) −1.232** (2.97) −2.250** (4.99) Estimated cutoff points Cutoff point 1 ˆa1 0.2798 (1.06) −0.7927** (2.66) −0.7646* (2.56) −0.9719* (2.26) −0.3325 (0.69) Cutoff point 2 ˆa2 2.573** (9.65) 1.477** (4.91) 1.503** (5.00) 1.570** (3.65) 1.801** (3.71) Observations 5192 5185 5185 2928 2257 5192 2932 2260 Log likelihood −3733 −3768 −3770 −2010 −1534 −12516 −7489 −4612 X2 1228 1089 1089 380.9 850.8 355571 159143 6384 Notes: Equation 3.1 estimated by ordered probit, 3.2–3.5 by a two stage instrumental variable ordered probit with bootstrapped standard errors, 3.6–3.8 by a conditional mixed processor estimator with a binary dependent variable. (.) denotes t statistics and **/* denotes significance at the 1% and 5% levels. Regional dummy variables included in all regressions. 8A.HAMILTONANDJ.HUDSON
  • 10. variables separately. Secondly, we used binomial probit for the impact equations rather than ordered probit, with the variable differentiated between a lot of impact (coded 1) and some or no impact. Finally, we combined two of the advice variables together, that is land management and agricultural practices. These potentially relate to all farmers and also to farming per se rather than other aspects of the business and have a reasonably high correlation. The cmp estimator is complex, but the interpretation of the coefficients and t statistics is as with techniques such as OLS. The results are shown in column 3.6 and are similar to previously. The main difference is the significance, at the 1% level, of the advice variable relating to both agricultural prac- tices and land management and also marketing advice. In column 3.3, we replicate these using the two stage approach used in 3.2. Comparing 3.3 and 3.6, the only significant difference relates to the negative coefficient on the credit advice variable. Using both tech- niques provides a robustness check on the findings and the cmp results largely confirm those of the two stage IV approach.6 Taking the equations as a whole, family size is never significant, nor gender and were dropped after the first equation.7 More educated people and older people tend to have benefited more from the advice than others, although the latter only in 3.1. However, in the cmp regression the variable relating to crop growers is no longer positively significant. The literature has suggested that the impact of advice may vary according to the con- ditions facing the individual, for example soil moisture. We do not have in the database a measure of this nor rainfall in the kebele, but we do have a variable which asked the indi- vidual whether they were usually subject to water shortages for drinking at some time in the year. Slightly over 51% responded that they were subject to such shortages. We now split the sample into those who were and were not subject to such water shortages. Columns 3.4 and 3.5 relate to those who were and were not subject to water shortages. The positive impact of marketing advice is limited to the latter as is that of land management and agricultural practices. However, the negative impact of credit advice is evident only for those in non-drought areas. Finally, we note that animal husbandry advice is only significant for those in areas of drought and advice on fertilizers is significant in both areas. These equations were estimated using the two stage approach. Columns 3.7 and 3.8 show the results of using the cmp estimator. The main difference is the positive significance of animal husbandry advice in both areas. In Table 4, we look at the results pertaining to income. The results are very similar. This was to be expected, but it was always possible that because of expenses incurred in implementing the advice, a positive impact on yield would not translate into a positive impact on income. The main difference to Table 3 is the significantly negative coefficient on credit advice in both areas, implying that receivers of such advice are significantly less likely to have perceived a beneficial impact on income. In addition, those with either crops or animals were also less likely to perceive benefits. Because the regression is limited to those who received advice, the possibility exists for sample selection bias. When we tested for this in the two-stage IV regressions the evidence was mixed. The inverse Mills ratio is insignificant in the income regression, but significant in the crop yield regression, although not in drought-affected areas. The significance of the coefficients reported in Table 3 for the full sample for crop yields did not change and nor did those in the non-drought regression. The main difference in both these regressions was an increase in the size of the coefficient on animal husbandry advice. The sample selection equation included a kebele based variable reflecting the number in the kebele, other than THE JOURNAL OF AGRICULTURAL EDUCATION AND EXTENSION 9
  • 11. Table 4. Regression results: impact on income. Sample Probit: Full 4.1 IV 2 stage Full 4.2 Probit: Full 4.3 Drought 4.4 No drought 4.5 CMP:Full 4.6 Drought 4.7 No drought 4.8 Extension agent advice Animal husbandry 0.326** (7.20) 0.8695** (5.52) 0.773** (4.82) 1.380** (6.33) 0.4807 (1.62) 0.6428** (5.19) 0.8796** (6.24) 0.5656** (2.70) Credit −0.2863** (6.27) −0.0429 (0.19) −0.0815 (0.35) −0.5072 (1.76) 0.4657 (1.16) −0.612** (4.10) −0.6696** (3.36) −0.4433** (2.85) Marketing 0.4026** (8.83) 0.2882 (1.63) 0.4343* (2.47) −0.0009 (0.00) 1.153** (3.73) 0.6562** (4.91) −0.103 (0.53) 1.466** (12.21) Fertilizers 0.1489** (3.30) 0.149 (0.59) 0.583* (2.40) 0.5567 (1.65) 0.3363 (0.82) 0.7453** (5.43) 0.901** (6.00) 0.7655** (3.49) Land management 0.2292** (5.12) 0.6268** (2.61) Agricultural practices −0.0486 (0.85) 0.2364 (1.08) LM + AG 0.739* (2.23) −0.0341 (0.07) 2.039** (3.55) 1.332** (5.64) 0.0223 (0.04) 1.164** (2.96) Individual characteristics Log age 0.1423* (2.24) 0.0612 (0.92) 0.0787 (1.17) −0.1311 (1.33) 0.1717 (1.62) −0.0833 (1.25) −0.1737 (1.86) −0.078 (0.76) Education 0.0177** (3.78) 0.017** (3.33) 0.0179** (3.47) 0.0147* (2.00) 0.0109 (1.40) 0.0153** (3.14) 0.0173* (2.55) −0.0046 (0.63) Male −0.0065 (0.19) Log family size 0.0081 (0.19) Farm characteristics Crops 0.2137* (2.22) −0.2432 (1.63) −0.2617* (1.96) −0.0601 (0.25) −0.8143** (3.88) −0.7315** (4.45) −0.3371 (0.98) −0.3603 (1.33) Animals −0.1338** (2.81) −0.5482** (5.07) −0.6039** (5.42) −0.4797** (3.21) −0.8772** (4.73) −0.3407** (3.67) −0.1669 (1.49) −0.5536** (3.74) Constant −0.939** (3.14) −1.829** (3.39) −0.6396 (1.36) Estimated cutoff points Cutoff point 1 ˆa1 −2.423** (4.16) −2.500** (8.96) −1.043** (3.64) −0.8468 (1.87) −1.210** (2.84) Cutoff point 2 ˆa2 −0.3297 (0.57) −0.4381 (1.57) 1.003** (3.51) 1.586** (3.52) 0.6052 (1.40) Observations 5195 5188 5188 2928 2260 5190 2930 2260 Log likelihood −4096 −4154 −4177 −2095 −1843 −12556 −7433 −4688 X2 697.6 590.5 566.6 301.6 535.1 111195 3517 64356 Notes: Equation 4.1 estimated by ordered probit, 4.2–4.5 by a two-stage instrumental variable ordered probit with bootstrapped standard errors, 4.6–4.8 by a conditional mixed processor estimator with a binary dependent variable. (.) denotes t statistics and **/* denotes significance at the 1% and 5% levels. Regional dummy variables included in all regressions. 10A.HAMILTONANDJ.HUDSON
  • 12. the respondent, receiving advice. Apart from this, older and more educated people were less likely to receive advice. Men were more likely to receive advice, as were those in large families and those with animals or crops. This is potentially consistent with those who report a bias in extension advice towards men (Buchy and Basaznew 2005). A variable differentiating drought from non-drought areas was insignificant. In Table 5, we show the probabilities of the respondent finding that the advice made ‘a lot of difference’ under different scenarios. These are obtained from the two stage instru- mental ordered probit regressions. They are based on a 45-year-old individual with crops, but no animals. The regional variables are averaged according to sample population. The first element in the first column shows the probability for someone with the above charac- teristics who receives none of the specific types of advice listed and has no schooling. The second column relates to someone with an education corresponding to ‘grade 8’, and this raises the probability to 0.0627. In the second row, we have the probabilities for someone of the above characteristics who received advice on animal husbandry. These are consider- able higher, reflecting the effective nature of this advice. However, the most effective forms of advice are on fertilizers and the combined advice on land management and agricultural practices. The next two columns relate solely to farmers in drought areas. The most impor- tant forms of advice are on animal husbandry and fertilizers, with land management and agricultural practices being third. This is in sharp contrast to the non-drought areas, where the most effective advice is on agricultural planning and land management. The results for income are also shown. The main difference is the increased impact of marketing advice. Some of these impacts look quite small, although it should be borne in mind that they do relate to the probability of having made ‘a lot of difference’ and also that multiple forms of advice are often given which substantially increases this probability. Conclusions and policy implications With respect to the main research questions, we find that extension agent advice impacts positively on both crop yields and income as reflected by the farmers’ own perceptions. Table 5. Probabilities of advice have a ‘lot of difference’ under different scenarios. Full sample Drought areas Non-drought areas Education None Grade 8 None Grade 8 None Grade 8 Crop yield None 0.040 0.063 0.059 0.082 0.0036 0.0067 Animal husbandry 0.099 0.144 0.280 0.341 0.0059 0.0105 Marketing 0.089 0.131 0.051 0.072 0.0458 0.0702 Credit 0.037 0.059 0.036 0.051 0.0394 0.0612 Fertilizer 0.226 0.299 0.233 0.289 0.0259 0.0415 Planning and land man. 0.179 0.243 0.113 0.150 0.3790 0.4620 Income None 0.059 0.080 0.087 0.110 0.0146 0.0186 Animal husbandry 0.214 0.264 0.508 0.561 0.0445 0.0545 Marketing 0.129 0.166 0.088 0.110 0.1520 0.1760 Credit 0.050 0.069 0.031 0.041 0.0431 0.0529 Fertilizer 0.163 0.206 0.211 0.251 0.0325 0.0403 Planning and land man. 0.204 0.253 0.082 0.104 0.4430 0.4820 Notes: These probabilities are based on regressions 3.3–3.5 and 4.3–4.5. They relate to a 45-year-old individual with crops, but no animals in a typical region. They show the probability of a single type of advice making ‘a lot of difference’ to crop yield and income. THE JOURNAL OF AGRICULTURAL EDUCATION AND EXTENSION 11
  • 13. This confirms the first hypothesis and confirms the more favorable findings of Elias et al. (2013) and Khan et al. (2014), rather than, frequently older, research such as Pender, Place, and Ehui (2006) and Lefort (2012). We also find such advice to have a varying impact according to both the farmer’s characteristics, such as education, and the characteristics of the farm. The pattern of significance suggests that agricultural advice yields its best results when targeted at those with the ability to use it, that is, the better educated. However, it is also suggestive perhaps that the advice given to less well educated people needs to be different, made simpler or given in more detail. This helps answer the second of our research questions and partially supports the third hypothesis, although we find only limited evidence of impact being related to age. There are also significant differences in the impact of advice between different areas. This again relates to the second research question and provides confirmation of our second hypothesis. In drought-affected areas, advice on animal husbandry and fertilizers is most effective. In non-drought areas, advice on marketing and land management and agricultural practices is best. The impact of marketing on crop yields is plausibly an indir- ect one whereby farmers respond to increased prices and a greater ability to sell output by increased effort. Our findings are consistent with our theoretical analysis which suggested that advice had the greatest potential in areas and conditions most conducive to applying the advice. This differential impact of advice in different areas is consistent with the results of Kassie et al. (2010). However, they find that advice on land management practices works best in low agricultural potential areas which is slightly at odds with our findings. In addition, they found advice on fertilizers to have a very significant and positive impact on crop productivity in high potential areas, whereas we found this advice worked well in both types of areas. Why the differences with Kassie et al.? Firstly, there is the time period. Their data relate to 1998 and 2001. Ours is more recent. Technology moves on and what might have been the case over a decade ago may no longer be the case today. Secondly, their analysis relates to usage of fertilizer, minimum tillage, etc. Our data relate to exten- sion agent advice on these technologies. The advice may feasibly be to use less fertilizer, or to use it in a different manner. Thirdly, some technologies may take several years before they have an impact. Hence Schmidt and Tadesse (2014) conclude that that sustainable land and water management infrastructure in the Ethiopian highlands has a positive impact on the value of production only seven years after construction. This might be too long a time frame for our results to be picking up. Finally, there may be a difference between their split of high and low productivity areas and ours of drought and non- drought areas. In addition, more educated individuals and older people were significantly less likely to receive advice. The kind of advice received also differed between drought and non-drought areas and also varied with the size of the family, education and gender, suggesting that advice is targeted rather than given randomly. Thus, this does not support the observation made by Kassie et al. (2010) that the advice showed little variation across different environments nor responded to household-specific factors. This again confirms part of the third hypothesis. Thus in terms of the three hypotheses we set out to test, all three have been largely substantiated. Expanding further on our second research question, some advice has been found to be more effective than others. In virtually, all the regressions, advice on credit appears to have the least positive impact on both yields and income, especially the latter. This 12 A. HAMILTON AND J. HUDSON
  • 14. may be linked to credit being heavily under the control of the government and possibly being used for political purposes (Berhanu and Poulton 2014). Advice on credit is also almost always given in tandem with other advice and the combined effect, as can be seen from the coefficients in Tables 3 and 4, is generally positive, at least for yields. Nonetheless, it does raise questions about the recent emphasis on credit (Bachewe et al. 2015). We did not have enough data to investigate whether different types of advice work better in tandem. But a variable equal to the number of types of advice given, although negative, gave only weak evidence for declining returns with respect to the amount of advice given. Hence, advice appears best if given on several dimensions with the impact being largely cumulative. This is consistent with Teklewold et al.’s (2013) conclusion that different types of sustainable land practices work best when adopted in combination rather than isolation. There is an important policy issue in that extension system advice can be unsuccessful either because the problem lies with the ‘extension message’ or in the way the message is transmitted. Too often the latter tends to get the blame for lack of impact while the problem may be with the ‘advice’ or message. In a sense that is the more serious problem in indicating that the advice is flawed to begin with, whereas retraining can alle- viate problems with the messenger. Our analysis may be the start of a process of determin- ing where the problem lies, if indeed there is a problem. In this context, our results tentatively show that with much of the advice being effective in at least some contexts the messenger is at least partially exonerated. This also links in to the current debate on the relative roles of R&D and technology transfer (Anandajayasekeram 2011). Our results point to the importance of the latter, in the specific context of extension agents. But often new knowledge comes from R&D and the roles of research centers remain of potential importance and here perhaps the linkages between agricultural research centers and extension agents could be improved (Abegaz and Wims 2015; Teshome, de Graaff, and Kassie 2015). Similarly, the use of regression analysis to quantify the impact of extension agent advice has provided results which may be potentially useful to improv- ing and informing extension agents’ advice in part by internalizing the results of this type of analysis within the formal education system. Notes 1. http://www.farmingfirst.org/wordpress/wp-content/uploads/2012/06/Global-Forum-for- Rural-Advisory-Services_Fact-Sheet-on-Extension-Services.pdf. 2. There are types of advice given in addition to the six specified, but for the purpose of the theory we focus on the impact of these six types of advice. 3. The kebele is the lowest administrative tier in Ethiopia’s federal structure, below the woreda. 4. If the question had been on what had happened to yields and income, then just focusing on those who received advice would be problematic. But, we cannot ask a similar question of those who did not receive advice as the question we are analysing pertains to the impact the advice had on yields and income. Obviously, this question cannot be asked of those who did not receive advice. 5. The constant term has been absorbed into the cut-off points for the ordered probit regression. 6. The coefficients cannot be readily compared as the one relates to ordered probit and the other to probit. THE JOURNAL OF AGRICULTURAL EDUCATION AND EXTENSION 13
  • 15. 7. They were however significant at times in the equations relating to policy advice, rather than impact and have been retained in those. For example, land management advice tended to increase with family size, which may reflect the size of the holding, and credit advice declines with family size. There was weaker evidence of credit and marketing advice declining for men. Acknowledgements We acknowledge the valuable comments of referees and the editor. Disclosure statement No potential conflict of interest was reported by the authors. Funding This work was supported by the Department for International Development (DFID). Notes on contributors Dr Alexander Hamilton is a political economist, and evaluation specialist with significant academic and field experience in fragile states. He has numerous publications and research experience in the fields of corruption, impact evaluations, public financial management, economic policy, and econometrics. He also has field experience from work conducted in Ethiopia, Senegal, Sudan, and Yemen. He has an MPA in Public and Economic Policy form the London School of Economics and a DPhil (PhD) from the University of Oxford. Professor John Hudson studied at the Universities of London and Warwick. He is a pro- fessor of economics at the University of Bath. He has published more than 100 journal papers in leading journals in all areas of economics and the wider social sciences, but in particular development economics. References Abegaz, D. M., and P. Wims. 2015. “Extension Agents’ Awareness of Climate Change in Ethiopia.” The Journal of Agricultural Education and Extension 21 (5): 479–495. Adams, R., H. Almeida, and D. Ferreira. 2009. “Understanding the Relationship Between Founder – CEOs and Firm Performance.” Journal of Empirical Finance 16 (1): 136–150. Anandajayasekeram, P. 2011. “The Role of Agricultural R&D within the Agricultural Innovation Systems Framework.” In Report Prepared for the ASTI/IFPRI-FARA Conference. Bachewe, F. N., G. Berhane, B. Minten, and A. S. Taffesse. 2015. “Agricultural Growth in Ethiopia (2004–2014): Evidence and Drivers.” ESSP Working Paper 81. Addis Ababa: International Food Policy Research Institute/Ethiopia Strategy Support Program (ESSP). Berhanu, K., and C. Poulton. 2014. “The Political Economy of Agricultural Extension in Ethiopia: Economic Growth and Political Control.” Development Policy Review 32 (S2): s197–s213. Buchy, M., and F. Basaznew. 2005. “Gender-blind Organizations Deliver Gender-biased Services: The Case of Awasa Bureau of Agriculture in Southern Ethiopia.” Gender, Technology and Development 9 (2): 235–251. 14 A. HAMILTON AND J. HUDSON
  • 16. Clarke, P. S., and F. Windmeijer. 2012. “Instrumental Variable Estimators for Binary Outcomes”. Journal of the American Statistical Association 107 (500): 1638–1652. Cohen, M. J., and M. Lemma. 2011. Agricultural Extension Services and Gender Equality. Discussion paper. Washington, DC: International Food Policy Research Institute. Davis, K., B. Swanson, D. Amudavi, D. A. Mekonnen, A. Flohrs, J. Riese, and E. Zerfu. 2010. In- depth Assessment of the Public Agricultural Extension System of Ethiopia and Recommendations for Improvement. Discussion paper. Washington, DC: International Food Policy Research Institute. Dercon, S., and L. Christiaensen. 2011. “Consumption Risk, Technology Adoption and Poverty Traps: Evidence from Ethiopia.” Journal of Development Economics 96 (2): 159–173. Dercon, S., D. O. Gilligan, J. Hoddinott, and T. Woldehanna. 2009. “The Impact of Agricultural Extension and Roads on Poverty and Consumption Growth in Fifteen Ethiopian Villages.” American Journal of Agricultural Economics 91 (4): 1007–1021. EEA/EEPRI. 2006. Evaluation of the Ethiopian Agricultural Extension with Particular Emphasis on the Participatory Demonstration and Training Extension System (PADETES). Addis Ababa: Ethiopian Economic Association/Ethiopian Economic Policy Research Institute. Elias, A., M. Nohmi, K. Yasunobu, and A. Ishida. 2013. “Effect of Agricultural Extension Program on Smallholders’ Farm Productivity: Evidence from Three Peasant Associations in the Highlands of Ethiopia.” Journal of Agricultural Science 5 (8): 163–181. Kassie, M., P. Zikhali, J. Pender, and G. Köhlin. 2010. “The Economics of Sustainable Land Management Practices in the Ethiopian Highlands.” Journal of Agricultural Economics 61 (3): 605–627. Kelemework, F., and H. Kassa. 2006. Assessment of the Current Extension System of Ethiopia: A Closer Look at Planning and Implementation. Addis Ababa: Ethiopian Economic Association/ Ethiopian Economic Policy Research Institute. Khan, Q. M., J. P. Faguet, C. Gaukler, and W. Mekasha. 2014. Improving Basic Services for the Bottom Forty Percent: Lessons from Ethiopia. Washington, DC: The World Bank. Lefort, R. 2012. “Free Market Economy, ‘Developmental State’ and Party-state Hegemony in Ethiopia: The Case of the ‘Model Farmers.’” Journal of Modern African Studies 50 (4): 681–706. Pender, J., F. Place, and S. Ehui, eds. 2006. Strategies for Sustainable Land Management in the East Africa Highlands. Washington, DC: International Food Policy Research Institute. Roodman, D. 2011. “Estimating Fully Observed Recursive Mixed-process Models with CMP.” Stata Journal 11 (2): 159–206. Schmidt, E., and F. Tadesse. 2014. “Sustainable Agriculture in the Blue Nile Basin: Land and Watershed Management Practices in Ethiopia.” Environment and Development Economics 19 (5): 648–667. Spielman, D. J., D. Byerlee, D. Alemu, and D. Kelemework. 2010. “Policies to Promote Cereal Intensification in Ethiopia: The Search for Appropriate Public and Private Roles.” Food Policy 35 (3): 185–194. Teklewold, H., M. Kassie, B. Shiferaw, and G. Köhlin 2013. “Cropping System Diversification, Conservation Tillage and Modern Seed Adoption in Ethiopia: Impacts on Household Income, Agrochemical Use and Demand for Labor.” Ecological Economics 93: 85–93. Teshome, A., J. de Graaff, and M. Kassie 2015. “Household-level Determinants of Soil and Water Conservation Adoption Phases: Evidence from North-Western Ethiopian Highlands.” Environmental Management 57 (3): 620–636. THE JOURNAL OF AGRICULTURAL EDUCATION AND EXTENSION 15