Demand, Supply and Willingness-to-Pay

for Extension Services in an

Emerging-Market Setting
Garth John Holloway and Simeon K. Ehui

Although it may be wholly inappropriate to highlands (Nicholson). Previous work with

generalize, the most important resource avail- the data (Holloway et al.) suggests that
able to a subsistence household is the total extension visitation is a potentially important
amount of time that its members have avail- catalyst for market expansion. Consequently,
able to spend in productive enterprises. In a number of important questions arise con-
this context, services that minimize the time cerning the actual impacts of extension on
that it takes to perform productive activi- participating and non-participating house-
ties are valuable to the household. Conse- holds; the amount that extension-requesting
quently the household is willing to relinquish households would be willing to pay for the
quantities of other resources in exchange for
service if it was privatized; the corresponding
quantities of the time-saving service. These
simple observations motivate a search for the demand schedule for extension services; and
values that subsistence households place on the requisite conditions for the existence of
time-saving services. This search is especially a private market for the service. These ques-
important when it is realized that extension tions are central to the development of mar-
services promote productivity, enhance the kets, to the issues raised in this session and
surplus-generating potential of the household are answered in this paper in the context of
and can, as a consequence, promote immer- our Ethiopian data.
sion into markets that are currently con-
strained by thinness and instability. In this
capacity, extension visitation has the poten-
Milk-Market Development in
tial to overcome one of the principal impedi-
ments to economic development, namely lack the Ethiopian Highlands
of density of market participation. In this arti-
cle, we consider this issue in the context of a Small-holder dairy production in the
rich data set on milk-market participation by Ethiopian highlands received an enormous
small-holder dairy producers in the Ethiopian catalyst when, in 1997, at two sites close to
Addis Ababa, the inauguration of two milk
cooperatives provided incentive for farmer
This paper was presented at the ASSA winter meetings (New
Orleans, LA, January 2001). Papers in these sessions are not sub- participation in milk-marketing. Small-holder
jected to the journal’s standard refereeing process. dairy producers in the highlands face a num-
Garth John Holloway is visiting scientist and and Simeon K.
Ehui is coordinator, Livestock Policy Analysis Program, Interna-
ber of barriers to marketing including poor
tional Livestock Research Institute. access to information, low levels of infras-
The authors are grateful to Chris Barrett, Chris Delgado, tructure and problems of transportation and
and Ariel Dinar for providing inspiration; Mohammed Ahmed,
Graeme Donovan, and Bill Thorpe for supportive conversations; perishability, leading to low levels of par-
Abebe Misgina and Yigezu Atnafe for information about exten- ticipation in markets otherwise constrained
sion services in the Ethiopian Highlands; and Charles Nichol-
son, who conceived the collection of the data while a postdoc-
by significant search costs. Along with the
toral fellow at the International Livestock Research Institute introduction of the milk cooperatives (“milk
under a grant from the Rockefeller Foundation. The authors are groups”) a production innovation with enor-
especially grateful to Adugna Mekonnen, Head, North Shewa
Zone Ministry of Agriculture; Fille Kedida, Head, Wouchale mous potential for improving productivity
Woreda Ministry of Agriculture; Getachew Kebede, Supervi- among small-holders is the crossbreeding
sor, Wouchale Woreda; and Asgedech Eshete, Etenesh Seme,
and Derrese Woyessa, Development Agents, Wouchale Woreda of exotic dairy-producing breeds with grade
for information offered in an October 16 interview concerning indigenous stock. Although susceptible to dis-
extension activities in Wouchale Woreda.
This article is dedicated to U.C. Davis Extension Specialist
ease and requiring more labor intensive man-
Bees Butler. agement practices than purely indigenous
Amer. J. Agr. Econ. 83(3) (August 2001): 764–768
Copyright 2001 American Agricultural Economics Association
Holloway and Ehui Willingness-to-Pay for Extension 765

herds, crossbred dairy-producing herds pro- decreasing in extension visitation. Transac-

duce upwards of twice their indigenous coun- tions costs may involve the cost of acquir-
terparts (Kiwuwa et al.). To varying degrees ing transport equipment, of walking (time)
this productivity gain has stemmed increases or, possibly, learning a new production tech-
in marketable surpluses and brought with it nique in order that the selling unit conforms
the potential for market immersion by for- to some market standard. Given estimates
merly subsistence households. In addition, of the impact of extension visitation on
the inauguration of cooperative selling has the transactions costs function, the task of
lowered significantly the remaining barrier retrieving estimates of willingness-to-pay is
to trading and has freed the time absorbed simplified considerably. The quantity that we
by producing transportable products from seek is the amount of income that the house-
fluid milk (butter, cheese, yogurt) to other hold is willing to forgo in order to have one
productive enterprises in the household. additional unit of the service rendered or, the
With these issues in mind, a comprehen- quantity ω that solves:
sive survey of some sixty-eight households
was undertaken in 1997 (Nicholson) with the (3) V (p̂ + δ m − ω q) = V (p̂ m q)
view to assessing the significance of factors
that impede or promote market participation. where V (·) denotes indirect utility, δ ≡
Each household was visited three times and c(z) = ∂c(z)/∂zx zx represents the change
at each visit sales of milk to the milk group in effective price afforded by the increase
in the preceding seven days were recorded. in extension visitation and q denotes other
Thus, some (68 × 3 × 7 =) 1428 observations factors affecting producing and consuming
are available for analysis. decisions. Taking a Taylor-series approxima-
tion at an arbitrary expansion point on both
sides of equation (3) an empirically observ-
able (approximation to) the willingness-to-
Transactions Costs Mitigation and
pay measure is
Willingness-To-Pay For
Extension Visitation (4) ω ≡ δVp /Vm
Despite potential complications, the deriva- Econometric interests focus on the extent
tion of household-production willingness-to- to which the right-hand-side of this expres-
pay estimates shares many similarities with its sion is observable and the location and scale
derivation in the more traditional consumer of its posterior distribution. The term in the
model. We focus most attention on the cash- numerator is the product of the extension-
income constraint, service change (which is observable) and the
marginal valuation of one additional unit of
(1) z ≤ p̂ys + ww + m extension (which, in general, is not observ-
able); the term in the denominator gives the
where z denotes a (numeraire) purchased marginal valuation of one additional unit of
good, p̂ denotes the household’s “effec- income (which, again and in general, is not
tive price,” w denotes the wage in off-farm observable). However, from a simple applica-
employment, w denotes labor allocation to tion of the envelope theorem, Vp ≡ λys and
off-farm employment and m denotes exoge- Vm ≡ λ, where λ denotes the Lagrange mul-
nous income. Per-unit transactions costs are tiplier corresponding to the cash-income con-
conceptualized as the difference between the straint and ys denotes the amount of the food
per-unit price received in the market, p, and product that is sold in equilibrium. Hence, the
a per-unit costs function c(z) so that right-hand-side of (4) is observable and pro-
vides the basis for the willingness-to-pay cal-
(2) p̂ = p − c(z)
culations that follow.
denotes the effective (net) return to the
household and z ≡ (z1 z2 zk ) denotes
a vector of potentially observable, household- Estimating Willingness-To-Pay From
specific characteristics, one of which we Market Participation Data
assume to be extension visitation. Implicit
in what follows, is the assumption that the In estimating the impact of extension on the
transactions costs function is monotonically transactions costs function we infer a direct
766 August 2001 Amer. J. Agr. Econ.

relationship between the household’s effec- Results

tive price and the effect of extension visi-
tation on this price. To this end, household Table 1 reports results. The variables in ques-
i’s decision about whether to participate can tion are “Distance” ≡ return time in minutes
be expressed in terms of (indirect) utility to walk to the milk group; “Education” ≡
differences years of formal schooling of the house-
hold head; “Crossbred” ≡ number of cross-
(5) G(p̂i mi qi ) ≡ V p (p̂i mi qi ) bred cows currently being milked; “Local” ≡
− V n (mi qi ) number of local-breed cows currently being
milked; “Extension” ≡ number of visits by
where V p (·) denotes utility from partic- an extension agent discussing production or
ipation, V n (·) denotes utility from non- marketing activities in the twelve months
participation, G(·) denotes the utility prior to the survey; “Ilu-Kura” ≡ dummy
difference and mi and qi are defined variable (= 1 if from the Ilu-Kura peasant
previously. Given the monotonicity of association); “Mirti” ≡ dummy variable (= 1
G(p̂i mi qi ) in p̂i we can reinterpret the if from the Mirti peasant association). Num-
entry condition (G(·) positive or negative) bers in parentheses are the ratios of poste-
in terms of a critical price, say p̂ic , such that rior means to their standard deviations in
entry occurs whenever the effective price, p̂i , the Gibbs sample. All of the covariates have
a significant impact on the price differences.
exceeds the critical price and not conversely.
The price difference is decreasing in the dis-
Given the dependence of G(p̂i mi qi ) on
tance that the household resides from the
mi and qi , it is natural to group these terms
milk group and is increasing in each of the
into a set of covariates, xi ≡ (mi qi ), and
other covariates. Particular interest lies in
consider the price rule as a transformation
the impacts of the extension-visitation covari-
of a standard-normal regression,
ate. The results suggest that for each unit
increase in extension visitation the effective
(6) σi (p̂i − p̂ic ) = σi xi
+ σi ui
price increases (the transactions-costs func-
tion is lowered) by 0.62 Ethiopian birr (EB).
where σi is an arbitrary, positive scalar; xi is
(Currently, $US 1.0 = EB 8.24). Hence, exten-
a k-vector of observations on the covariates;
sion visitation shows promise as a market-

is a k-vector of unknown coefficients; and
entry catalyst.
ui ∼ N (0 1) is a standard-normal random
Combining the probit regression results
variable. Normalizing on the scale parameter,
with the willingness-to-pay calculation in
and assuming independence across house-
holds, we are able to derive estimates of the
Table 1. Probit Regression Results
quantities we seek from the regression,
Distance −0 06
(7) y = x
+ u (−9 27)
Education 0 56
where y ≡ (y1 y2 yn ) denotes an (9 94)
n-vector of latent (unobserved) values cor- Crossbred 2 87
(14 78)
responding to the price differences; x ≡
Local 1 14
(x1 x2 xn ) , x1 ≡ (x11 x12 x1k ), (7 22)
x2 ≡ (x21 x22 x2k ) xn ≡ (xn1 , Extension 0 62
xn2 xnk ) are observations on the covari- (10 66)
ates; and u ∼ N (0n In ). We observe the Ilu-Kura −6 67
participation choice but neither y nor
. (−13 82)
However, inferences about the locations Mirti −11 31
and scales of the elements of
are eas- (−18 53)
ily obtained using a Markov-chain, Monte- Participants 168
Carlo algorithm (Albert and Chib). In short, Pos. Predict. 74
an alternative interpretation of the probit Neg. Predict. 94
regression leads to robust estimates of the Non-Participants 1260
impact of extension on the household’s effec- Pos. Predict. 24
tive price and its willingness to pay for exten- Neg. Predict. 1236
sion visitation.
Holloway and Ehui Willingness-to-Pay for Extension 767

equation (4), we can estimate the quantities is “normal demand” (November–December)

of interest in terms of the Ethiopian birr average visits amount to about 100 per week
that the households are willing to give up per agent. The extension department of the
for one additional extension visit. Figure 1 administrative unit in question has a total of
presents the estimates across the 168 partic- 29 Development Agents and in the produc-
ipating households. The point estimates are tion year 1998–99 (Ethiopian calendar year
virtually indistinguishable from their 95% 1991) incurred running costs amounting to
highest-posterior-density values (confidence EB 294,748 (Wouchale Woreda Extension
intervals) and the estimates appear credible Personnel). Dividing this amount by the total
in terms of the operations under considera- number of visits, which is ((180 × 20) + (18 ×
tion. Average sales in the sample are 3.9 l 20) + (100 × 8)) × 29 = 138 040 visits per year,
milk per day and the prices received are EB we arrive at a rough estimate of the marginal
1.00 and EB 1.25 per liter, respectively, at the cost of each visit—EB 2.14 per visit. Using
Ilu-Kura and Mirti milk groups. The maxi- the willingness to pay estimates in figure 1,
mum (standard deviation) willingness-to-pay it appears that some sixty-five “households”
estimate is 6.77 (0.50) EB and the minimum (observations in the sample) would be will-
(standard deviation) willingness-to-pay esti- ing to purchase extension services. Hence, net
mate is 0.62 (0.05) EB. These estimates pro- of fixed costs, and in the context of milk-
vide relatively precise information to which market development in the Ethiopian high-
supply considerations should be targeted. lands, the privatization of extension services
Complicating supply estimates are prob- is an intriguing possibility.
lems arising in the allocation of fixed costs
(educating field representatives, housing a
base of personnel and equipment, updat- Conclusions
ing and revising communication techniques)
and estimating the variable costs of visitation
(wages and variable expenditures incurred Extension visitation is a potent catalyst stim-
in travel). However, rough estimates about ulating entry into emerging milk-markets.
the per-unit costs can be made by divid- This article has analyzed willingness to pay
ing the annual extension budget of the local for extension visitation in the context of data
administrative unit by the number of vis- collected from two sites close to Addis Ababa
its made to each household. During the and has assessed the conditions necessary for
“peak season” for visitation (April–August) a private market to prevail. Although a pre-
extension agents typically visit with 180 cise response must hinge on the magnitude
farmers per week; during the “slack seasons” of any fixed costs, variables costs cannot pre-
(September–October and January–March) clude the possibility that privatization may be
average visits amount to 18 per week per currently possible. Future work should aim at
agent; and during the seasons in which there strengthening confidence in our estimates by
overcoming limitations of the analysis, includ-
ing the assumption of unit variance in the
pricing model, the assumption of indepen-
dence across time periods and the arbitrary
assignment of covariates to the transactions
costs function.

References

Albert, J., and S. Chib. “Bayesian Analysis of

Binary and Polychotomous Data.” J. Amer.
Stat. Assoc. 88(1993):669–79.
Holloway, G., C. Nicholson, C. Delgado, S. Ehui,
and S. Staal. “Agroindustrialization Through
Institutional Innovation: Transactions Costs,
Cooperatives and Milk-Market Development
Figure 1. Willingness-to-pay estimates for a in the East African Highlands.” Agric. Econ.
one unit increase in extension visitation 23(2000):279–88.
768 August 2001 Amer. J. Agr. Econ.

Kiwuwa, G.H., J.C.M. Trail, M.Y. Kurtu, G. Worku, Shewa and Arsi Regions of Ethiopia: Project
F.M. Anderson, and J. Durkin. Cross-Breed Description, Survey Methodology, and Collec-
Dairy Cattle Productivity in Arsi Region, tion Procedures.” Mimeograph, Livestock Pol-
Ethiopia. International Livestock Center for icy Analysis Project, International Livestock
Africa (ILCA) Research Report No. 11, Research Institute, Addis Ababa, 1997.
Addis Ababa: ILCA, 1983. Wouchale Woreda Extension Personnel. Personal
Nicholson, C.F. “The Impact of Milk Groups in the communication, October 14, 2000.