You are on page 1of 11

Ecological Economics 70 (2011) 1508–1518

Contents lists available at ScienceDirect

Ecological Economics
j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / e c o l e c o n

Analysis

Scope economies and technical efficiency of cocoa agroforesty systems in Ghana


Adeline Ofori-Bah, John Asafu-Adjaye ⁎
School of Economics, University of Queensland, Brisbane, Q4072, Australia

a r t i c l e i n f o a b s t r a c t

Article history: A number of studies have addressed issues relating to the physiological, environmental and economic values
Received 7 June 2010 of trees in cocoa farming systems. However, to date, little has been done to quantitatively examine the effect
Received in revised form 25 January 2011 of crop diversity on cocoa farming efficiency. This study therefore sets out to first investigate whether and to
Accepted 16 March 2011
what extent crop diversity (defined as the mixing of cocoa with other crop species on farmers' plots) affects
Available online 27 April 2011
productivity on cocoa farms. Secondly, it sought to establish whether there are economies of scope (i.e., cost
Keywords:
complementarities) from the sharing of farm inputs by crops on the same plots. Our results indicate that
Technical efficiency diversified (i.e., multi-crop) cocoa farms are more efficient than single (i.e., mono) crop farms. Furthermore,
Economies of scope our estimate for the economies of scope parameter indicates possibilities for cost complementarities between
Cocoa agroforestry production of cocoa and other crops on the same plot. We advocate further investigation on the issue of scope
Crop diversity economies to determine which crop combinations offer better cost complementarities and also meet
Multi-output distance function frontiers biodiversity conservation objectives.
Ghana © 2011 Elsevier B.V. All rights reserved.

1. Introduction producer. This has contributed to smuggling of Ghana's cocoa to that


country. The Ghana government has attempted to offset the low
Agriculture and related enterprises contributed an average of prices by providing subsided inputs and other services. In early
38.6% to Ghana's Gross Domestic Product (GDP) from 2004 to 2008. October 2010, it announced a 33% increase in the farm gate price
The sector provided employment for close to 51% of the country's (GNA, 2010).
economically active labor force and accounted for 37.9% of its export The increase in cocoa production has exacerbated deforestation
earnings in 2008. Cocoa production contributed 13% and 28.5%, and raised concerns about biodiversity conservation and the sustain-
respectively, of the country's agricultural GDP and total export ability of agricultural production (Asare, 2005). A number of re-
earnings in 2008 (ISSER, 2009). Hence, it can be said that, like many searchers (Amanor, 1996; Asare, 2005; Ruf and Zadi, 1998) have
rural areas of West and Central Africa, cocoa is the “engine of suggested that cocoa production can be sustained over the long term
economic growth” (Ruf and Zadi, 1998; Sonwa et al., 2009). The only through the adoption of cultivation methods that incorporate the
production of cocoa is predominantly a smallholder activity on an use of forest tree species in the rehabilitation and recycling of land.
average farm size of 2.2 ha and within a range of less than one to forty Cocoa agro-forests contribute to conservation strategies by providing
hectares (Nyanteng, 1995). Dependence on cocoa, both in the macro habitat and resources for flora and fauna and by maintaining
and rural household economies, generates remarkable interest and connectivity between forest areas (Schroth et al., 2004). The Ghana
support for the sector from policy makers and the public sector. It is government and associated national and international institutions are
one of the few agricultural commodity sectors in which smallholders thus promoting agroforestry technologies that facilitate rehabilitation
in Ghana can expect a dedicated and assured market for their output. of old cocoa farms and recycling of degraded lands in order to reduce
The upward trend in the producer price of cocoa over the period the problems impeding farm efficiency while still considering
between 2001 and 2010 together with government support for biodiversity and agricultural sustainability.
farmers in the sector and the subsequent profitability of farming the Although a number of studies (Alvim, 1977; Beer et al., 1998;
crop have provided incentives for farmers in Ghana to expand Duguma et al., 2001; Leakey, 1998; Rice and Greenberg, 2000; Ruf and
cultivated land, leading to the conversion of forest lands to cocoa Schroth, 2004; Scherer-Lorenzen et al., 2005a; Schroth, 1999; Schroth
farmland (Anglaaere, 2005; Teal et al., 2006). Although there have et al., 2004; Wessel and Gerritsma, 1994) have addressed various
been some market reforms in Ghana, farm gate prices for cocoa are issues relating to the ecological and economic values of other trees in
still relatively low compared to Côte d'Ivoire, the World's No. 1 cocoa farming systems, the issue of economies of scope that can arise
in cocoa agroforestry systems remains unexplored. Some studies (Teal
et al., 2006; Vigneri, 2007) have examined efficiency of cocoa
⁎ Corresponding author. Tel.: + 61 7 33656539. production in Ghana using farm yields (output per land size) as a
E-mail address: j.asafuadjaye@uq.edu.au (J. Asafu-Adjaye). proxy. However, farm yields provide only a partial measure of

0921-8009/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.ecolecon.2011.03.013
A. Ofori-Bah, J. Asafu-Adjaye / Ecological Economics 70 (2011) 1508–1518 1509

efficiency and inferences based on them could be misleading. Binam farms. While Sections 1 and 2 of Act 547 of Ghana's legislation titled
et al. (2008) used a more robust procedure to measure technical ‘The Timber Resource Management Act, 1997’ (Forestry Commission,
efficiency and productivity of cocoa farming in West and Central 2009) clearly disqualify farmers from harvesting timber from their
Africa. However, their analysis was based only on cocoa output and land, Section 4 gives them the authority to make a decision about
did not account for output of other crops on the farm that might also which third party body gets the right to take timber from their land.
contribute to total farm efficiency. The Act has therefore had a significant influence on farmers' decisions
In acknowledgement of the inherent technical interdependence of on whether or not to integrate other trees on their cocoa farms. This
crops and the inability to allocate inputs to crop components of issue was brought up a number of times during our survey. The
intercropping systems, Alene et al. (2006) compared the empirical prevailing trend is for farmers interested in growing other tree species
performances of single-output production and multi-output distance with cocoa to intercrop using those exotic and local tree species that
function frontiers for measuring efficiency in the system. Their generate significant and direct economic value. With this practice,
findings reiterated the observation made by a number of other some cocoa is still grown under the shade of other trees and farmers
researchers (Brummer et al., 2003; Coelli and Fleming, 2004; Coelli are able to earn additional income. Findings from studies by Bobo et al.
and Perelman, 1996; Morrison-Paul et al., 2000) that efficiency in (2006) and Frimpong et al. (2007) indicate that the levels of forest
multiple output production technologies needs to be evaluated using biodiversity are significantly reduced with the practice.
analytical methods that account for and model multiple outputs and A significant proportion of cocoa is still grown in full sun and in a
multiple inputs. The analytical framework adopted for this study is monoculture system. The phenomenon has been attributed to two
therefore motivated by these observations. main factors: a) the emergence of the new hybrid varieties developed
The study has two key objectives. Firstly, it examines the extent to to thrive without or with little shade; and b) the scarcity of forest trees
which crop diversity, defined as the mixing of cocoa with other crop to provide the required shade due to increasing pressure on land,
species on farmers' plots, affects farm technical efficiency. Secondly, deforestation and land degradation (Anglaaere, 2005).
it seeks to establish whether there exist cost complementarities (i.e., The beneficial effects of species mixtures that enhance crop
economies of scope) from the sharing of farm inputs by crops on the yield are of interest in this context. Proponents of crop biodiversity,
same plots within cocoa agroforestry systems in Ghana. The results and agroforestry in particular, typically base their argument on
have important policy implications in that evidence of enhanced ecological underpinnings in support of this production practice. There
technical efficiency and/or economies of scope would suggest that is a growing body of literature that supports the hypotheses that
there are prospects for “win–win” outcomes on cocoa farms in terms mixed crop-tree systems allow for better cycling of nutrients, erosion
of biodiversity conservation and improved production efficiency. The control (water and soil retention), carbon sequestration, a richer
paper makes two important contributions to the literature. Firstly, it biological diversity and a more complex interaction between
is the first to examine the possible economies of scope associated flora and fauna than mono-crop farming systems (Nair, 1993; Parrish
with crop diversity on cocoa farms. Secondly, it investigates the et al., 1998; Rolim and Chiarello, 2004; Scherer-Lorenzen et al.,
efficiency differential between multi-crop and mono-crop cocoa 2005b). Thus, inter-planting of trees and crops is often encouraged to
farms. To the best of our knowledge, no previous studies have promote effective mutualism between plant species, a practice
considered the effects of these issues on the productivity of cocoa perceived to compensate for external inputs (Isaac et al., 2007; Jose
farming systems. et al., 2004).
The remainder of this paper is organized as follows. The current Data from two-species mixtures has improved knowledge about
literature on the effects of mixing tree species is discussed in Section 2. the mixture effects. It is now known that the relationship between
Section 3 briefly describes the conceptual framework for estimating the species diversity and productivity varies and depends considerably on
production technology, technical efficiency and economies of scope on species mix, site, silviculture treatment and risks. It has also been
the sample farms. Section 4 presents the methodology employed which established that the crucial factors for productivity enhancement in
includes a description of the survey area, survey approach and the mixed cultures are dependent on the ecological niche of the species
estimation methods. Section 5 presents and discusses the empirical and their compatibility (Pretzsch, 2005). Kelty (1992) and Jones et al.
results, while Section 6 concludes with the summary and policy (2005) suggest that there is potential for increased productivity in a
implications. mixture of equally productive species because they are able to
complement each other in the spatial–temporal usage of space and
2. Effects of Tree Species Mixture on Efficiency in Cocoa Agroforestry other resources.
Systems Economists can expect that similarities in the atmospheric and
edaphic factors that influence the survival of crops coexisting on a plot
Cocoa cultivation in Ghana in the past has been based on the can generate commonalities in production technologies and resource
removal of natural forest trees and planting the cocoa crop using some use among the crops, thereby giving scope for sharing inputs (Kay,
residual shade (Anim-Kwapong, 2003). This traditional practice 1997). Both fixed and variable inputs are jointly applied to all crops
presented a good example of the agroforestry approach and exhibited because farmers cannot selectively apply inputs to crop components,
an excellent illustration of compatibility and complementarity of resulting in a more complete use of farmers' inputs and other
various species (Alvim and Nair, 1986). In addition it sustained the resources available to the crops. Opportunities can be offered in terms
multi-storey system typical of the humid tropics (ibid). The practice of economies of scope derived from common and fuller use of
has, however, given way to clear-felling and burning of the forest in indivisible input resources among crops produced on the same plot
order to farm cocoa. Presently, a typical cocoa farm starts as a food- (Kay, 1997). An implied reduction in cost per unit of output could in
tree intercrop during the first few years of its establishment (about 3 turn enhance crop production (Alene et al., 2006). It is also argued
to 4 years). Shade is provided in this case by inter-planting plantain, that the inputs required by mono-cropped cocoa are also likely to be
banana and food crops that provide early shade until coppice shoots expensive for the farmers of smaller holdings who produce the bulk of
mature and become the over-storey species. Ghana's cocoa and hence could lead to poor yields (Anglaaere, 2005).
As the farms become older, the forest seed bank allows some
native trees to regenerate. Though farmers acknowledge the benefi- 3. Conceptual Framework
cial effects of these native trees, the current national legislation and
policy on the ownership of and rights to harvest the species used for The procedures for predicting technical efficiency and estimating
timber does not motivate farmers to nurture those species on their economies of scope for a production unit require appropriate
1510 A. Ofori-Bah, J. Asafu-Adjaye / Ecological Economics 70 (2011) 1508–1518

modeling of the underlying production technology. Agroforestry Using the production technology depicted in Fig. 1, the input set
systems are essentially diversified economic entities and thus the L(y) is the area bounded from below by the isoquant, UU′. The value of
process of transforming their inputs into outputs is best represented the distance function at point A is the ratio ρ = 0A/0B which is the
by multiple output transformation functions. Early parametric inverse of the measure of technical efficiency 0B/0A. Hence, for any
frontier studies attempting to model multiple output scenarios number of outputs produced by a single firm, an input-oriented
were noted to have used either input requirement functions or a measure of technical efficiency is defined by the function:
distance function with an input or output orientation. The input
requirement function restricts the production technology to a single n o
I I
T E ðx; yÞ = min ρ : D ðρx; yÞ≥1 : ð3Þ
composite input and, more often than not, is based on the
assumption that inputs are used in fixed proportions (Coelli and
Perelman, 1996). Alternatively, the composite input is represented Distance functions are increasingly gaining popularity among
by an aggregate index of inputs or the single input assumed to be the researchers seeking to investigate efficiencies in smallholder systems
most important. This restriction is avoided when the distance where data on input costs for estimating cost functions are often
function is used. unavailable or unreliable, and for production technologies character-
The representation of multiple outputs and multiple inputs ized by multiple output technology.
production technology by a distance function presents a multi-
dimensional problem which incorporates all the two-dimensional 3.2. Economies of Scope
cases of output–output, input–output and input–input (Morrison-
Paul et al., 2000). The analytical framework of a multiple output, Agroforestry is a related diversification strategy (Kay, 1997) in
multiple input case is aptly depicted with a distance function, since it which a farmer extends existing capabilities, resources, or areas of
provides a functional characterization of the structure of production expertise to more than one crop cultivated on the same plot. It affords
technology (Kumbhakar and Lovell, 2000). Distance functions have the possibility for the simultaneous production of more than one crop
been used to represent production interactions in the multiple on a given farm to generate scope economies in contrast with their
outputs and inputs context by a number of researchers including production in isolation, i.e., with each crop on its own specialized
Färe et al. (1993), Lovell et al. (1994), Coelli and Perelman (1996), farm. Economies of scope are a relative cost advantage that result from
Morrison-Paul et al. (2000), Brummer et al. (2003), Coelli and Fleming spreading the cost of a set of resources and skills over more than one
(2004) and Alene et al. (2006). product or enterprise.
Economies of scope (EOS) are traditionally defined relative to a
3.1. Technical Efficiency multiple output cost function. Following Gravelle and Rees (2004),
EOS in a two product case at some input price vector p and output
Technical efficiency is the ability to minimize inputs used in vector y can be defined as:
the production of a given level of output. It is conceptualized
and measured relative to a production function. Within the past

two decades, Shephard's (1953, 1970) input and output distance C ðy1 ;y2 ;pÞ b C ðy1 ;0;pÞ + C ð0;y2 ;pÞ; 0≤ y ≤y ð4Þ
functions have gained increasing popularity for characterizing the
structure of production technology especially when multiple inputs
are used to produce multiple outputs. The input distance function where, y⁎ is the specified technically efficient output. Eq. (4) states
DI(x,y), which involves the scaling of an input vector, is defined on the that economies of scope arise when the minimized cost of producing
input set L(y) as: given quantities of two outputs together, C(y1,y2;p), is less than the
minimized cost of producing the same quantities of each output
separately. The reverse of the inequality defines diseconomies of
I
D ðx; yÞ = maxfρ : ðx = ρÞ∈LðyÞg; ð1Þ scope. This function is assumed to be differentiable (and hence
continuous) only for y N 0 so as to allow the possibility of over-all, or
product-specific, set-up costs (Panzar and Willig, 1977).
where the input set L(y) represents the set of all input vectors, x, A measure that summarizes the degree of EOS observed at some
which can produce the output vector, y (Coelli et al., 2005). Farrell's input price vector p and output vector y is defined in the two output case
(1957) input-oriented measure of technical efficiency (TEI) can be as follows:
defined in terms of the input-distance function as follows:

I 1 EOS = ½C ðy1 ; 0; pÞ + C ð0; y2 ; pÞ−C ðy1 ; y2 ; pÞ = C ðy1 ; y2 ; pÞ: ð5Þ


TE = : ð2Þ
DI ðx; yÞ

It measures the relative increase in cost that might result from


dividing the two outputs for production in two separate production
U A units. A positive EOS is an indication that splitting the two outputs for
x 2/y separate firms will increase total cost. A negative EOS on the other
B hand is an indication that separating the production of the two
outputs will decrease total cost. Hajarasht et al. (2008) emphasize the
P
fact that Eq. (5) cannot be verified directly when outputs appear in
D logarithmic or reciprocal forms in the cost function. In such cases the
C following sufficient condition depicting cost complementarities
which has been proposed by Baumol et al. (1988) can be used to
U´(L(y))
verify the existence of scope economies:

0 P x1/y
∂2 C ðy; pÞ 
≤ 0; i≠j 0≤y≤y : ð6Þ
Fig. 1. Representation of technical efficiencies. ∂y1 ∂y2
A. Ofori-Bah, J. Asafu-Adjaye / Ecological Economics 70 (2011) 1508–1518 1511

Fig. 2. Ghana's cocoa growing region.


Source: Cadbury (2010).

4. Methodology The close-ended questions were first pre-tested as open-ended


questions to ensure that answer options in the final version covered
4.1. Survey Area all possible options. The field survey was conducted in the period July
to December 2009. Wherever possible the output data were validated
The survey was carried out in the Ashanti, Western and Eastern from the farmers' passbooks, which were supplied to them by local
Regions of Ghana, out of the six Administrative Regions in the country licensed cocoa purchasing agents for sales record keeping purposes.
that grow cocoa (Fig. 2). The Western Region is a relatively new cocoa The sample was selected using a combination of probability and non-
producing area and currently produces 45% of Ghana's total annual probability methods. The purposive sampling approach was used to
output (ECAM Consultancy, 2006). It is the region that has undergone select the Administrative Regions to be included in the study. Following
the greatest conversion of virgin forest to cocoa farming in recent that, a two-stage random sampling procedure, stratified by whether the
years. farmer single cropped or multiple cropped his/her farm, was used to
The Eastern Region is the oldest cocoa-growing area and it has select cocoa farmers from four districts. The first stage involved a
witnessed the highest incidence of cocoa virus diseases. High pro- random selection of one district in the Ashanti and Eastern Regions and
portions of farmers in these two regions are migrants from other regions two in the Western Region. Weighting of the number of districts per
of the country. The Ashanti Region, which is located between the other region was based on the regional distribution of cocoa farming districts.
two, has significantly different land tenure arrangements with about Using a sampling frame of farmers provided by the Ministry of Food and
80% of farms being managed by members of the local ethnic group Agriculture, about 80 farmers were randomly selected at the second
(Takane, 2000). stage from each of the sample districts to form the study sample. This
procedure was adopted to account for the differences in production
4.2. Survey Design, Sample, and Questionnaire Design conditions across the regions and to avoid the possibility of selecting an
excessive number from any particular region. Another advantage of the
The data collection procedure involved a random survey of 340 procedure was the reduction in the distances between sample units and
cocoa farmers using the face-to-face interview technique. A struc- hence reduction in travel time and cost.
tured questionnaire with a combination of both open and close-ended
questions was used to collect the data. The interviews were 4.3. Stochastic Frontier Estimation
complemented with focus group discussions with cocoa farmers in
the study area. Prior to the main survey, the survey instrument was Stochastic frontier estimation provides a useful methodology for
pre-tested for the purpose of evaluation and refinement. predicting technical efficiency patterns among a cross section of
1512 A. Ofori-Bah, J. Asafu-Adjaye / Ecological Economics 70 (2011) 1508–1518

farms. Through a two-component error term in the estimation and CT4 are estimated to be 418, 797, 666 and 403 grams per tree,
process, frontier econometric techniques allow the separation of respectively, the estimated w's are as follows:
noise associated with measurement error from technical inefficiency
linked to the inability to attain best-practice technology. Several w1 = 418 = 796 = 0:53; w2 = 796 = 796 = 1:0
extensions and applications of the stochastic production frontier w3 = 666 = 796 = 0:84; w4 = 403 = 796 = 0:51:
function have been documented since it was first proposed indepen-
dently by Aigner et al. (1977) and Meeusen and van den Broeck The farm level distance function is approximated by a translog
(1977). This study uses one of such extensions suggested by Battese function as follows:
and Coelli (1995) for incorporating exogenous factors that might
4 2 1 4 4
influence technical efficiency in a one-step estimation of the I
lnDi ðx; yÞ = β0i + ∑ βn lnxni + ∑ αm lnymi + ∑ ∑ β lnx lnx
stochastic production frontier. n=1 m=1 2 n = 1 j = 1 nj ni ji
The multi-dimensional production technology of a cocoa agro- 1 2 2
+ ∑ ∑ αmk lnymi lnyki
forest farm is represented by the following transformational function: 2 m=1 k=1
4 2
T ðX; Y Þ = 0 ð7Þ + ∑ ∑ φnm lnxni lnymi ð9Þ
n=1 m=1

where X denotes the input vector and Y denotes the output vector. where ln denotes the natural logarithm of the corresponding
Alternative representation of this production technology is an input variables; and α, β and φ are unknown parameters to be estimated.
set, L(Y), which encompasses the set of all X vectors that can produce Eq. (9) cannot be estimated directly because the dependent variable is
Y. An input distance function identifying the least input that can be unobservable. In order to obtain an estimable equation, homogeneity
used to produce the given output vector is defined by Eq. (1). of the input distance function with respect to input quantities to
For this farm level model, the Y vector contains two variables degree 1, as well as symmetry restrictions due to Young's theorem are
while the X vector contains four variables. The output variables are imposed such that lnDIi (λx, y) = lnλ + lnDIi (x, y), λ = x− 1
1
and
dried cocoa beans (C) measured in kilograms and the market value βnj = βjn and αmk = αkm, for all n, j, m and k.
(i.e., the product of total output and the prevailing price at the farm Having normalized the input variables x with one of the input
gate in the district) of other crops (OC) intercropped with cocoa. variables, the following expression
Market value was estimated as the product of the existing market  
price and the farmers' total output (includes both sold and household I x2 x −1 I
lnDi ; …; N−1 ; y = lnx1 + lnDi ðx; yÞ
consumption). The prices were obtained from the Ministry of x1 x1
Agriculture and verified with the farmers. The major crops
commonly intercropped with cocoa at its fruiting ages are avocado is obtained by setting −lnDIi = vi − ui as proposed by Coelli and
(Persea americana), coconut (Cocos nucifera), citrus (Citrus sinensis), Perelman (1996), and hypothesizing that ui is influenced by a set of
cola (Cola nitida) and oil palm (Elaeis guineensis), as well as various exogenous variables described below, the multi-output–multi-input
native species of timber trees. stochastic input distance function representing the production
Land size (LS) is a fixed input variable measured in hectares. The technology can be specified as:
three variable inputs are: the weighted number of cocoa trees (WCT)      
WCTi TLi PIi
which accounts for the age of the trees; total labor (TL) measured in −lnLSi = β0 + β1 ln + β2 ln + β3 ln + α1 lnðCi Þ + α2 lnðOCi Þ
LSi LSi LSi
person days to include household labor, hired labor and communal/  2  2  2
1 WCTi 1 TLi 1 PIi
shared labor; and the market value of all intermediate inputs + β4 ln + β5 ln + β6 ln
2 LSi 2 LSi 2 LSi
purchased (PI) by the farmer and/or provided by externally funded        
WCTi TLi WCTi PIi
programs. Intermediate inputs included fertilizer, pesticide and + β7 ln × ln + β8 ln × ln
LSi LSi LSi LSi
fungicide, food for workers, and transport.    
TL PIi 1 2 1 2
A preliminary linear regression analysis of cocoa output involving + β9 ln × ln + α3 lnðCi Þ + α4 lnðOCi Þ
LSi LSi 2 2
a number of regressors revealed the significant influence of the age of  
WCTi
cocoa trees on the level of output. Cocoa trees start producing from + α5 lnðCi Þ × lnðOCi Þ + φ1 lnðCi Þ × ln
LSi
three to five years after planting, depending on the variety. They    
TLi PIi
reach full bearing capacity between 6 and 15 years with good + φ2 lnðCi Þ × ln + φ3 lnðCi Þ × ln
LSi LSi
agronomic practice but their output begins to decline gradually    
WCTi TLi
thereafter. One way to incorporate age effects into the model without + φ4 lnðOCi Þ × ln + φ5 lnðOCi Þ × ln
LSi LSi
affecting the degrees of freedom of the econometric estimation is to  
PIi
generate a new variable that captures the effect of both age and tree + φ6 lnðOCi Þ × ln + vi −ui
LSi
number (Gimbol et al., 1995; Hung et al., 1993). The variable is
defined as: and

WCTi = w1 CT1i + w2 CT2i + w3 CT3i + w4 CT4i ð8Þ ui = δ0 + δ1 STi + δ2 Divi + δ3 ECi + δ4 Agei + δ5 ELi + δ6 FTi + δ7 Gi ;
ð10Þ
where WCTi is the weighted number of cocoa trees on farmer i's plot;
CT1, CT2, CT3 and CT4 are the categories of 3–5 years, 6–15 years, 16– where ST=number of trees; Div=diversification on farm efficiency
25 years and over 25 years, respectively; and the wi's are the weights (defined below); EC=a dummy variable, equaling 1 if the respondent
to be estimated. Using data obtained from the sampled farms, the made contact with any form of formal extension services in the year, and
average weight of dried cocoa beans per tree is calculated for each age 0 if not; Age=farmer's age; EL=the farmer's education level; and
category. Since trees of ages 6–15 years are at their full production G=gender, 1=male, 0=female; FT=dummy variable with 1=full-
capacity they are assigned a weight of w2 = 1. The ratios of average time farmer and 0=part-time; and all other variables are as previously
dry-bean weight for each age category to the average weight for those defined.
at full production (i.e., CT2) give the w's assigned to the respective In the above equation, the vi's are random variables assumed to be
categories. Given that the average dry-bean weight for CT1, CT2, CT3 independently and identically distributed with a zero mean and a
A. Ofori-Bah, J. Asafu-Adjaye / Ecological Economics 70 (2011) 1508–1518 1513

variance of σ2v . The ui's are assumed to be non-negative random that, in this case, a higher level of formal education could have a
variables independently determined but with a distribution that is negative influence on technical inefficiency.
truncated at zero with a mean of μi and a variance of σ2u, where μi = ziδ, An initial examination of the sample data revealed a strong
zi is a p × 1 vector of farm specific exogenous variables hypothesized to correlation (r = 0.82) between farmers' age and years of farming
influence farm efficiency (Battese and Coelli, 1995). An assumption is experience. Years of farming experience was therefore dropped in
also made that vi and ui are uncorrelated. favor of farmers' age (Age) to avoid the problem of multicollinearity. The
ST, Div, EC, Age, EL, FT and G are assumed to be exogenous factors age distribution of the sampled farmers shows that about 70% are aged
related to the production environment. Their influence on the over 50 years. This finding is consistent with data reported by Vigneri
production technology is assumed to be indirect and they therefore (2007) which show that about 80% of cocoa farmers are well above their
do not enter the deterministic kernel of the distance frontier. Instead, youthful years and the average age is 49 years. Age has been shown to
they are treated as determinants of the one sided inefficiency error have an increasing effect on technical inefficiency partly because older
term of the stochastic input distance function. farmers are less likely to adopt new technologies (Battese and Coelli,
The first of these (ST) is the number of trees on the farm that 1992; Burki and Terrell, 1998; Illukpitiya, 2005). They are also relatively
provide shade for the benefit of cocoa production. This includes both less energetic compared to their younger counterparts.
fruit and native timber trees. There has been significant debate about The remaining variables represent fulltime farming (FT) and
the effect of shade on cocoa output. While some studies reported gender (G). Eighty percent of the sample farmers were men, which
significant agro-ecological benefits that reduce plant stress (Adu- is consistent with past studies (Teal et al., 2006). Because cocoa
Ampomah et al., 1998; Beer et al., 1998) others provide evidence to production is a male dominated livelihood in Ghana, lower technical
show its limiting effect on the crop's yield (Anglaaere, 2005; Frimpong inefficiency values are expected for male farmers.
et al., 2007). Present consensus on the topic suggests that there is a The translog function was estimated using the maximum likelihood
threshold level beyond which shade provided by these crops becomes facility in the FRONTIER Version 4.1 computer program (Coelli, 1996).
detrimental to cocoa production (Bentley et al., 2004). With increased The sample data was mean corrected prior to estimation to ensure that
interest in diversified cocoa systems using fruit trees, present cocoa each variable is deflated by its mean. Both inputs and outputs were
research in Ghana tends to focus on finding an appropriate balance of scaled to have their means equal to one and the logarithm of the means
shade, and on identifying compatible tree species. The inclusion of this equal to zero. This procedure ensures that the first partial derivatives of
variable in the estimation of technical efficiency is to test whether the input distance function will equal the estimated first order
current numbers of shade trees planted by farmers influence farm coefficients. This procedure also facilitates the estimation of the value
efficiency. for economies of scope, as is shown later.
In order to evaluate the influence of crop diversification on farm
efficiency, a variable (Div) was constructed using the reciprocal of the
Herfindahl Index. It is a measure adopted from Simpson's Diversity 4.4. Justification of the Model
Index in ecology where it is often used to quantify the biodiversity of a
habitat. The mathematical expression for calculating the index of The use of a distance function is appropriate in this context even
diversity for each observation is as follows: though scope economies between two or more outputs is a cost
concept which is traditionally evaluated using the cost function. The
1 estimation of a cost function is based on the premise of cost
Index of diversityðDivi Þ = ; q = 1; 2; …; Q ð11Þ minimization, a phenomenon that might not always be applicable to
∑Qq = 1 Yiq2
smallholder systems. In addition, cost data on labor inputs in
particular are often not reliable for smallholder farming systems.
where Q is the number of crop species and Yiq is the fraction of the Household labor, a major source of labor input in smallholder
farmer's income generated from crop q. D ranges between 1 and agriculture, is adequately captured in an input distance function
infinity with values farther away from one corresponding to highly because it does not require cost data. Recent studies (Coelli and
diverse farms and values near 1 corresponding to more homogeneous Fleming, 2004; Hajarasht et al., 2008; Villano et al., 2005) have
farms. In this study, we expect the level of diversity to have a positive exploited the duality between the cost function and the input distance
effect on farm specific technical efficiency. function – when shadow costs can be assumed to be minimized – to
With regard to extension contact (EC), it is usually hypothesized represent this cost concept by using input distance functions.
to have a positive effect on farm specific technical efficiency because Hajarasht et al. (2008) have developed a measure for economies of
information disseminated through the service should ideally im- scope in terms of the input distance function and they advocate its use
prove farming practices (Illukpitiya, 2005). However, various factors in cases where costs and input price data are not available.
such as the quality of the extension service, the willingness and The translog functional form was selected for the empirical
ability of the farmer to utilize the disseminated information, the representation of the distance function because it is relatively more
existence of other sources of information, inter alia, can confound the flexible and does not impose restrictions of rectangular hyperbola
response to extension contact. The farmer's level of education (EL) is isoquants as in the case of the Cobb–Douglas function (Morrison-
measured on an ascending scale between 0 and 4, with 0 representing Paul et al., 2000). Besides, it permits testing for the more restrictive
no formal education and 4 representing tertiary education. A general Cobb–Douglas function as a special case. The translog function is
expectation is that farmers with higher levels of formal education particularly apt for evaluating scope economies because it in-
would be more knowledgeable and apply improved methods of corporates the cross terms for both inputs and outputs. Other
production, and hence achieve a higher level of technical efficiency. important features of the translog function include the fact that it
However, results from some studies of smallholder agricultural allows the representation of substitution options. Lastly, propor-
production systems (Coelli and Fleming, 2004; Fleming and tional elasticity measures can be derived directly from this functional
Lummani, 2001) have shown that farmers with relatively higher specification.
levels of education have higher levels of technical inefficiency. A A similar model, but with a single output was also used to estimate
reason given for this observation is that farmers with higher levels of technical efficiency levels for farms engaged in cocoa production as a
formal education tend to have other income sources to which they mono-crop. This allows for some comparison of the distribution of
devote more attention at the expense of managing their farms. efficiency scores among multiple crop and mono-crop scenarios of
However, previous studies in Ghana (Teal et al., 2006) also suggest cocoa production systems.
1514 A. Ofori-Bah, J. Asafu-Adjaye / Ecological Economics 70 (2011) 1508–1518

4.5. Inefficiency Estimation Table 2


Maximum likelihood estimates of the input distance function.

The exogenous factors influencing efficiency were incorporated Variable Coefficient Estimate Standard error t-ratio
using a one-step process proposed by Battese and Coelli (1995). The
Constant β0 0.410 0.107 3.820⁎⁎⁎
input distance of a farm is measured by its deviation from the frontier ln (WCT/LS) β1 0.234 0.138 1.690⁎
and is predicted as: ln (TL/LS) β2 0.455 0.082 5.546⁎⁎⁎
ln (PI/LS) β3 0.143 0.056 2.568⁎⁎⁎
I ln C α1 − 0.448 0.073 − 6.152⁎⁎⁎
Di = Efexpðui Þjei g; ei = ui + vi ð12Þ
ln OC α2 − 0.146 0.049 − 2.957⁎⁎⁎
0.5 ln (WCT/LS)2 β4 − 0.134 0.457 − 0.293
where DIi varies between 1 and infinity and lnDIi = ui is expected to 0.5 ln (TL/LS)2 β5 0.263 0.142 1.971⁎⁎
take a value between 0 and infinity. The estimated coefficients of the 0.5 ln (PI/LS)2 β6 0.040 0.049 0.835
distance functions for the multiple output and single output cases and ln (WCT/LS) × ln(TL/LS) β7 0.210 0.189 1.112
ln (WCT/LS) × ln(PI/LS) β8 − 0.184 0.089 − 2.061⁎⁎
associated efficiency scores are presented and discussed in the next
ln (TL/LS) × ln(PI/LS) β9 − 0.013 0.046 − 0.278
section. 0.5 ln (C)2 α3 − 0.029 0.084 − 0.350
0.5 ln (OC)2 α4 − 0.006 0.034 − 0.187
4.6. Measurement of Scope Economies ln C × ln OC α5 − 0.019 0.033 − 0.558
ln C × ln (WCT/LS) φ1 0.135 0.163 0.829
ln C × ln (TL/LS) φ2 0.102 0.099 1.029
Following Hajarasht et al. (2008), the EOS between cocoa and ln C × ln (PI/LS) φ3 0.009 0.039 0.225
other crops found in cocoa agroforestry systems in Ghana are ln OC × ln (WCT/LS) φ4 0.123 0.113 1.082
expressed in terms of the multiple output distance function as: ln OC × ln (TL/LS) φ5 − 0.011 0.043 − 0.262
ln OC × ln (PI/LS) φ6 0.024 0.021 1.092
2 Sigma squared σ2 0.111 0.014 7.573⁎⁎⁎
1 ∂ C ðy; pÞ
EOS = σu/σv + σu γ 0.065 0.084 0.774
C ð y; pÞ ∂ym ∂yk
Notes:
ln is natural logarithm.
! ⁎⁎⁎Significant at the 1% level.
I I 2 I 2 I 2 I
∂D ∂D ∂ D ∂ D ∂ D ⁎⁎ Significant at the 5% level.
= − + … ⁎ Significant at the 10% level.
∂ym ∂yk ∂ym ∂yk ∂ym ∂x1 ∂ym ∂xn

0 1−1 0 2 I 1
∂2 DI ∂DI ∂DI ∂2 DI ∂DI ∂DI ∂ D
B + ⋯ + C B C 5.2. Estimates for the Multi-output Multi-input Distance Function
B ∂x21 ∂x1 ∂x1 ∂x1 ∂xn ∂x1 ∂xn C B ∂x1 ∂yk C
B C B C
B C B C
×B
B ⋮ ⋱ ⋮ C B
C B ⋮ C
C Table 2 reports the estimates, along with their corresponding
B C B C
B 2 I C B 2 I C
standard errors and t-ratios, for the multiple output translog
@ ∂ D ∂DI ∂DI 2 I
∂ D I
∂D ∂D I A @ ∂ D A
+ ⋯ + stochastic distance function. Estimates for the single output model
∂xn ∂x1 ∂xn ∂x1 ∂x2n ∂xn ∂xn ∂xn ∂yk can be found in Appendix A. It can be seen that all the first order
ð13Þ coefficients have standard errors that are small relative to their
means, indicating that the model is fairly well estimated. The
Weak economies of scope are said to exist if Eq. (13) yields a non- estimated function is also found to meet monotonicity and curvature
positive value. Because all inputs and outputs are mean corrected, the conditions at the sample means, thus satisfying the theoretical
estimated coefficients of the multiple-output translog input distance requirements.
function provide equivalent measures of derivatives of the function. The signs of the first order coefficients are as expected. The input
coefficients (βn's) are positive while the output coefficients (αm's) are
5. Empirical Results negative. The positive first order βn's indicate that, ceteris paribus,
additions to the level of an input tend to increase the distance from the
5.1. Descriptive Statistics stochastic production frontier (i.e., increase technical inefficiency). On
the other hand, the negative sign for the first order αm's indicate that
Table 1 provides a summary of the statistics for the variables used increases in output tend to decrease the distance to the frontier. The
to estimate the multiple-output multiple-input model. The average signs for the output variables are also consistent with their represen-
cocoa output per farm is close to the estimate of 1686 kg reported by tation of the slope for the production possibility frontier.
Vigneri (2007) but is higher than the national average recorded in the Four sets of hypotheses are tested in this analysis. First, t-tests are
last survey carried out in 2004 (Teal et al., 2006). This could be due to conducted to verify the statistical significance of the estimated input
a number of factors. First, there has been massive public investment in and output coefficients. Next, a test of the appropriateness of the
cocoa since that time. Second, our sample contains a significant translog model versus the simpler Cobb–Douglas case is undertaken.
number of farmers from the Western region where yields are known The third set of hypotheses involves a t-test and a likelihood ratio test
to be higher (Vigneri, 2007). for the significance of the technical inefficiency coefficient (γ), while

Table 1
Summary statistics for the farm inputs and outputs.

Variable Observation Mean Standard Deviation Minimum Maximum

Dried cocoa per farm (kg) 161 1628.7 1761.7 89 10262.6


Output of other crops (Gh¢) 161 901.1 1459.79 18 15,600
Farm size (ha) 161 3.98 3.58 0.44 21.45
Weighted cocoa trees (number) 161 4024.4 3583.29 180 22,530
Labor (person days) 161 537.6 294.94 86 1585
Purchased inputs (Gh¢) 161 317.4 359.09 2.7 1867
A. Ofori-Bah, J. Asafu-Adjaye / Ecological Economics 70 (2011) 1508–1518 1515

the fourth looks at the effects of the exogenous variables on technical Table 3
inefficiency using t-tests. Coefficients for the exogenous variables estimated with the multiple-output input
distance function.
The results in Table 2 show that apart from the coefficient for
weighted cocoa trees (WCT) which is significant at the 10% level, all Exogenous variable Coefficient Estimate Standard error t-ratio
the other first order coefficients for inputs and outputs are statistically Constant δ0 − 0.182 0.376 − 0.484
significant at either the 5% level or lower, indicating their substantial Shade trees δ1 − 0.0029 0.0015 − 1.933⁎
influences on the distance from the frontier. The effect of WCT on the Diversity Index δ2 − 0.527 0.143 − 0.370
Extension contact δ3 − 0.110 0.094 − 1.174
distance from the frontier is not particularly strong probably due to
Farmer's age δ4 0.011 0.004 2.360⁎⁎⁎
the variations in the mix of cocoa varieties present on the sample Fulltime farmers δ5 0.285 0.091 3.152⁎⁎⁎
farms. The estimated coefficients can be considered as elasticities Level of formal education δ5 − 0.008 0.068 − 0.119
because the estimation was done at the means of the sample data. Gender δ5 0.057 0.102 0.566
These distance elasticities represent the responsiveness of the Log likelihood function − 50.24
Likelihood-ratio test − 35.54
distance function to changes in the input and output variables. For
example, a 10% increase in each of weighted number of cocoa trees, Notes:
⁎⁎⁎Significant at the 1% level.
total labor (TL) and intermediate inputs (PI) is associated with 2.3%, ⁎⁎ Significant at the 5% level.
4.6% and 1.4% increases, respectively, in the average farm's distance ⁎ Significant at the 10% level.
from the frontier, ceteris paribus. These results reflect the importance
of labor and purchased inputs in cocoa production. In terms of the
outputs, a 10% increase in cocoa production decreases the average
farm's distance from the frontier by 4.5%. In comparison, a similar average of 0.47. Only 4% of mono-crop farms lie within the top decile.
increase in the output of other crops decreases the distance by only Therefore, opportunities to raise crop output without increasing
1.5%, ceteris paribus. inputs are higher for this group of cocoa farms.
We tested the functional form by imposing restrictions on the
translog model in which all the squared and interaction terms 5.3. Estimates for Scale Economies
were set to zero. That is, non-rejection of the null hypothesis
(H0:β4 = β5 = β6 = β7 = β8 = β9 = α3 = α4 = α5 = φ1 = The estimation results also provide some information about scale
φ2 = φ3 = φ4 = φ5 = φ6 = 0) would support the Cobb–Douglas form. economies. Given that the estimation is done at the sample means, the
We obtained a p-value of 0.027, implying that the translog functional inverse of the sum of the output coefficients gives a measure of radial
form is more appropriate for the data set. returns to scale elasticity (Hajarasht et al., 2008). The estimate, − 1/
The estimated coefficient for γ in the multiple output model (α1 + α2) = 1.68, indicates the existence of increasing returns to scale.
(0.065) is relatively small, implying that only a small proportion It implies that an increase in all inputs in the same proportion k will
(6.5%) of the variations in the composed error term (σ2 = σ2u + σ2V) is result in an increase in output of a proportion greater than k. This
due to the inefficiency component. However, for the mono-crop finding suggests that smallholders can improve their productivity by
model, the value of γ is as high as 0.99, indicating that the total scaling up their operation with equal proportions of inputs. We find
variance of the composed error term of the mono-crop model is due this result to be in contrast to that of Teal et al. (2006), which was
mostly to the asymmetric error term. Overall, these results suggest based on an observed negative relationship between cocoa farm size
that diversified (i.e. multiple cropped) cocoa farms are a lot more and yields (i.e., returns to land) in Ghana. However, farm yield is only
efficient than mono-crop cocoa farms. a partial measure of efficiency which does not explain scalar effects. It
The distribution of technical efficiency scores for the sample farms also does not account for the use of other critical inputs, especially
is presented in Fig. 3. The technical efficiency scores for multiple labor which we have shown to have a significant effect on
cropped cocoa farms range between 0.49 and 1.0, with an average of productivity in cocoa farming. Our findings reinforce the argument
0.86. Although the technical efficiency range may seem rather wide, that observations based on partial analysis do not provide adequate
97% of the scores are within the top half of the distribution and 42% information on scalar effects.
are within the top decile. This suggests that, for the majority of the
multiple crop cocoa farms, opportunities to expand output without a 5.4. Effect of the Exogenous Factors on Farm-specific Technical Inefficiency
substantial increase in input use or adoption of new technologies is
limited. The range of technical efficiency distribution is wider for In a t-test of the statistical significance of γ in the estimated
mono-crop cocoa farms, lying between 0.1 and 0.99 and with an multiple output model (i.e., H0: γ = 0), the null hypothesis could not
be rejected at any acceptable level of significance. However, a
likelihood-ratio test for the null hypothesis of no technical inefficiency
effects (H0: γ = δ0 = δ1 = δ2 = δ3 = δ4 = δ5 = δ6 = δ7 = 0) was over-
whelmingly rejected. Using Table 1 of Kodde and Palm (1986), our
test statistic of 35.5 was found to be greater than the χ2 critical value
of 20.97 at the 1% level of significance for nine restrictions.
The above result is evidence of the joint significance of γ and the
inefficiency parameters reported in Table 3. That is, shade trees,
extent of crop diversity, age, education, gender and fulltime farming
jointly explain the differences in inefficiency levels among cocoa
farmers pursuing agroforestry. Our model of inefficiency effects is
therefore appropriate for the dataset.1 The negative sign for the

1
The technical inefficiency estimates do not consider regional differences such as
soils, temperature, rainfall, pests, and so on, as well as technology and other factors not
explicitly measured in the model. Regional differences are unobserved in the model
and are captured by the error term (vi). We plan to collect another round of data,
which will give us a panel dataset thus allowing us to model these unobserved effects
Fig. 3. Distribution of technical efficiencies of sample cocoa farms. more explicitly.
1516 A. Ofori-Bah, J. Asafu-Adjaye / Ecological Economics 70 (2011) 1508–1518

coefficient on shade trees (δ1) suggests that farms with higher Another pertinent issue regarding multiple crop cocoa farming is
numbers of shade trees are relatively less inefficient. In other words, a that some crop combinations may compete with cocoa for resources
marginal increase in the number of shade trees has a small positive or play host to cocoa pests and hence reduce the productivity of the
but significant effect on technical efficiency. This finding is in contrast latter.2 For example, Coelli and Fleming (2004) found that some crop
to Binam et al.'s (2008) result for Ghana in which the level of shade combinations require overlapping labor and management demands
had a negative albeit insignificant effect on technical efficiency. The which can negatively affect productivity even when biological
difference between the two findings could be due to how shade was complementarities suggests positive productivity effects. There is
measured. Though it is not clear how the amount of shade was therefore a need for future research in this area to determine which
measured, Binam et al. (2008) applied a more robust measure, an crop combinations are technically interdependent enough to generate
index for shade ranging from 0 (no shade) to 3 (heavy shade) to scope economies to help achieve both economic and conservation
reflect shade intensity. On the other hand, we counted the number of goals in cocoa farming systems.
shade trees on the farmer's plot. Admittedly the intensity of shade
varies depending on tree species planted. But the significance of the
number of shade trees might suggest that the trees contribute other 6. Conclusions and Policy Implications
biodiversity services besides shade to enhance farm technical
efficiency. The twin objectives of this study were to investigate the extent to
As expected, the coefficient for extension contact is negative but it which crop diversity on cocoa farms affects their technical efficiency,
is not statistically significant. Our measure of extension contact did and whether the production practice offers opportunities for cost
not probe the nature or type of the extension services sought. complementarities. The results indicated that diversified (or multi-
However, various studies (Feder et al., 2004) have found extension crop) cocoa farms are more efficient than single crop (or mono-crop)
contact to have a limited impact on cocoa productivity in a number of cocoa farms. The results also showed that factors such as shade trees,
developing countries. For example, in Binam et al.'s (2008) study, extent of crop diversity, extension contact, age, gender and full-time
extension was found to be an insignificant determinant of technical farming jointly influence the level of technical efficiency on cocoa
efficiency for Ghana, Côte d'Ivoire and the pooled model. This suggests farms. However, in terms of individual factors, the significant
that the quality of extension services tends to be poor in these variables were found to be shade trees, age, and fulltime farming.
countries. Fulltime farmers and relatively older farmers were found to be
The results also show that although our Diversity Index has the relatively inefficient. The former effect cannot be explained based on
expected sign, it is also not statistically significant. It is worth noting evidence from the study, but the latter effect may reflect the fact that
that our sample farms are all diversified but to different degrees in older farmers are less receptive to adoption of new technology as
terms of the shares of species cropped. Hence, it is mostly an economic reported in previous studies.
measure that does not adequately reflect species distribution or Of particular interest is the finding that a marginal increase in the
density. The coefficients for fulltime farming and farmer's age have a number of shade trees has a small positive effect on technical
significantly positive relationship with the level of technical ineffi- efficiency. Our estimate for the economies of scope parameter
ciency, highlighting their important influence on technical inefficiency. indicates possibilities for cost complementarities between production
The finding that fulltime smallholder farms are relatively inefficient of cocoa and other crops on a given plot. Thus, the study provides
could reflect a combination of factors including the relatively high some economic evidence in support of the ecological notion that
use of labor inputs. The observation that older farmers are relatively complementarities in resource use by forest trees can enhance crop
inefficient is consistent with past efficiency studies mentioned productivity.
earlier. Older farmers, though more experienced, are also less likely Market niches for forestry and agroforestry products as well as
to undertake rehabilitation of their farms. Finally, education and ecosystem services in the West and Central Africa regions are rising
gender have the hypothesized signs but are not statistically both domestically and overseas (Sonwa et al., 2009). Therefore,
significant. efforts to encourage the cultivation of certified tree crops has the
potential to deliver “win–win” benefits by improving farmers'
incomes as well as helping to conserve forest biodiversity. However,
5.5. Evidence of Economies of Scope national agroforestry programs that target specified species must
also satisfy farmers' preferences in order to kindle their interest and
Our estimate for the economies of scope based on Eq. (13) was participation for achieving the desired ecological objectives. To
−0.16. This suggests the existence of significant economies of scope facilitate this process, further research is required to determine how
between cocoa and the other crops for the average multiple crop different amounts of crop species influence biodiversity and what the
cocoa farms. The estimate suggests that, on the average, a 10% thresholds of shade cover are. Further investigation is also required
increase in the production of other crops is associated with a 1.6% on the issue of scope economies to determine which crop
decrease in the marginal cost of producing cocoa. It needs to be combinations provide better cost complementarities and also meet
emphasized that this is an average figure and as such will not the biodiversity objectives. If, for example, timber species are found
necessarily hold true for every multiple crop cocoa farm. Nevertheless, to provide higher scope economies, then there would be need for a
it indicates that there are prospects for cost complementarities review of the current laws pertaining to ownership of and harvesting
between the production of cocoa and other crops on the same plot. rights to timber species on farmlands. This would therefore provide
The majority of the farmers in our survey showed a preference for economic incentives to farmers to grow and nurture native timber
fruit trees as shade due to lack of economic incentives to nurture species on their lands.
native timber species. Results of this study show that intercropping Although not a central aim of this research, our finding that the
cocoa with fruit trees offers opportunities for economies of scope. aging cocoa farming population could reduce farm technical efficiency
However, Sonwa et al. (2009) argue that although fruit trees enable draws attention to the need for a national dialog on approaches to
farmers to gain additional income, they tend to reduce the forest address this potential threat to the cocoa industry.
biodiversity of traditional agroforests. Therefore, to promote biodi-
versity conservation and restoration, there is a need for the 2
For example, some of the farmers interviewed indicated that avocado provides
government to review the current legislation regarding ownership excellent shade for cocoa but also harbors epiphytes such as mistletoes that spread to
and harvesting rights to native timber species. cocoa trees.
A. Ofori-Bah, J. Asafu-Adjaye / Ecological Economics 70 (2011) 1508–1518 1517

Appendix A Beer, J., Muschler, R., Kass, D., Somarriba, E., 1998. Shade management in coffee and
cacao plantations. Agroforestry Systems 38, 139–164.
Bentley, J., Boa, E., Stonehouse, J., 2004. Neighbor trees: shade, intercropping and cocoa
in Ecuador. Human Ecology 32 (2), 241–270.
Table A.1 Binam, J.N., Gockowski, J., Nkamleu, G.B., 2008. Technical efficiency and productivity
Maximum likelihood estimates of the single-output input distance function. potential of cocoa farmers in West African countries. The Developing Economies
XLVI-3, 242–263.
Variable Coefficient Estimated Standard T-ratio Bobo, S., Waltert, M., Sainge, M., Njokagbor, J., Fermon, H., Muhlenberg, M., 2006. From
coefficient error forest to farmland: species richness patterns of trees and understorey plants along
a gradient of forest conversion in Southwestern Cameroon. Biodiversity and
Production variables Conservation 15, 4097–4117.
Constant β0 2.400 0.341 7.043 Brummer, B., Glauben, T., Lu, W., 2003. The impact of policy reform on productivity
ln (WCT/LS) β1 0.865 0.147 5.889 and efficiency in Chinese agriculture: a distance function approach. Working
ln (TL/LS) β2 0.233 0.701 3.333 Paper FE 0402. Department of Food Economics and Consumption Studies,
ln (PI/LS) β3 0.166 0.061 2.691 University of Keil.
ln C α1 − 0.614 0.046 − 13.305 Burki, A., Terrell, D., 1998. Measuring production efficiency of small farms in Pakistan.
.5 ln (WCT/LS)2 β4 0.208 0.151 1.377 World Development 26, 155–169.
.5 ln (TL/LS)2 β5 0.365 0.126 2.878 Cadbury, 2010. Ghana's Cocoa Region. Accessed on 2 November 2010 at http://www.
.5 ln (PI/LS)2 β6 0.131 0.052 2.525 skillsspace.co.uk/geography/cocoa_farming/ghana_cocoa_region.asp.
Coelli, T., 1996. A guide to FRONTIER version 4.1: a computer program for stochastic
ln(WCT/LS) × ln(TL/ β7 0.456 0.146 3.126
frontier production and cost function estimation. CEPA Working Papers No.7/96.
LS)
Department of Econometrics, University of New England, Armidale.
ln(WCT/LS) × ln(PI/ β8 0.535 0.143 3.745
Coelli, T., Fleming, E., 2004. Diversification economies and specialisation efficiencies in
LS) mixed food and coffee smallholder farming system in Papua New Guinea.
ln(TL/LS) × ln(PI/LS) β9 − 0.033 0.041 − 0.805 Agricultural Economics 31, 229–239.
Sigma squared σ2 0.108 0.016 6.636 Coelli, T., Perelman, S., 1996. Efficiency measurement, multiple output technologies and
σu/σv + σu γ 0.999 0.047 21.269 distance functions: with applications to European railways. Working Paper 96/05.
Centre de Rechereche en Economie Publique et en Economie de la population
Efficiency variable (CREPP).
Constant δ0 0.462 0.329 1.401 Coelli, T., Rao, P.D., O'Donnell, C., Battese, G., 2005. An Introduction to Productivity and
Extension contact δ1 0.071 0.049 1.460 Efficiency Analysis. Springer Science and Business Media, New York.
Years of farming δ2 0.009 0.065 0.144 Duguma, B., Gockowski, J., Bakala, J., 2001. Smallholder cacao (Theobroma cacao Linn.)
experience cultivation in agroforestry systems of West and Central Africa: challenges and
opportunities. Agroforestry Systems 51, 177–188.
Education of δ3 − 0.175 0.106 − 1.643
ECAM Consultancy, 2006. Ghana/EU Cocoa Sector Programme. A Report for the
respondent
European Union, Accra.
Full time farming δ4 0.207 0.141 1.462
Färe, R.S., Grosskopf, S., Lovell, C., Yaisawarng, S., 1993. Deviation of shadow prices for
Male farmer δ5 0.315 0.200 1.573 undesirable outputs: a distance function approach. Review 75, 374–380.
Log likelihood − 27.9 Likelihood-ratio test 13.70 Farrell, M., 1957. The measurement of productive efficiency. Journal of the Royal
function of one sided error Statistical Society A CXX, 253–290 Part 3.
Feder, G., Murgai, R., Quizon, J.B., 2004. Sending farmers back to school: the impact of
farmer field schools in Indonesia. Review of Agricultural Economics 26 (1), 45–62.
Fleming, E., Lummani, J., 2001. Analysis of technical effeciency of cocoa smallholders in
the Gazelle peninsula. East New Britain Province. Occasional Paper, Centre for
Acknowledgments Economic Policy Analysis, Department of Econometrics. University of New England,
Armidale.
We would like to sincerely thank two anonymous reviewers for Forestry Commission, 2009. Timber Resource Management Act, 1997 — Act 547. The
Forestry Commission of Ghana, Accra. Accessed on 19 January 2011 at http://www.
their constructive comments. Contributions by Knox Lovell and Tim fcghana.com/publications/laws/act_547/index.html.
Coelli for the specification of the analytical model are deeply Frimpong, K.O., Asase, A., Yelibora, M., 2007. Cocoa farming and biodiversity in Ghana.
appreciated. Funding support from the African Economic Research An Annual Project Report for the Earthwatch Institute, Accra.
Ghana New Agency, GNA, 2010. Cocoa farmers laud government over cocoa price
Consortium (AERC) for this research is gratefully acknowledged.
increment. Ghana New Agency, Accra. Accessed on 10 October 2010 at http://www.
ghanaweb.com/GhanaHomerPage/NewsArchive/artikel.php?ID=194888.
Gimbol, C., Battese, G., Fleming, E., 1995. Technical efficiencies of smallholder cocoa
References producers in Papua New Guinea: a stochastic frontier analysis. Quarterly Journal of
International Agriculture 34 (4), 337–357.
Adu-Ampomah, Y., Frimpong, E.B., Abaitey, M.A., Ofori-Frimpong, K., Opoku, G., Yeboah, Gravelle, H., Rees, R., 2004. Microeconomics. Pearson Education Limited, Hants, UK.
S.B., 1998. Occurrence and distribution of light crop (category G) beans in the main Hajarasht, G., Coelli, T., Rao, P.D., 2008. A dual measure of economies of scope.
crop season. A report submitted to the Ghana COCOBOD. Economics Letters 100, 185–188.
Aigner, D., Lovell, C., Schmidt, P., 1977. Formulation and estimation of stochastic Hung, S.T., Coelli, T., Fleming, E., 1993. Analysis of the technical efficiency of state rubber
frontier production models. Journal of Econometrics 6, 21–37. farms in Vietnam. Agricultural Economics 9, 183–201.
Alene, D., Manyong, V., Gockowski, J., 2006. The production efficiency of intercropping Illukpitiya, P., 2005. Technical efficiency in agriculture and dependency on forest
annual and perennial crops in southern Ethiopia: a comparison of distance resources: an economic analysis of rural households and the conservation of
functions and production frontiers. Agricultural Systems 91, 51–70. natural forests in Sri Lanka. Economy and Environment Program for Southeast Asia
Alvim, P.T., 1977. Cacao. In: Alvim, P.D., Kozlowski, T. (Eds.), Ecophysiology of Tropical (EEPSEA) Technical Paper. Accessed on January 18 2010 at http://www.idrc.ca/
Crops. Academic Press, New York, pp. 279–313. uploads/user-S/11521683271Prabodh%28Tech%29.pdf.
Alvim, R., Nair, P., 1986. Combining cocoa with other plantation crops. Agroforestry Systems 4, Institute of Statistical Social and Economic Research (ISSER), 2009. The State of the
3–15. Ghanaian Economy in 2008. ISSER, University of Ghana, Legon.
Amanor, K., 1996. Managing trees in the farming system: the perspective of framers. Isaac, M., Ulzen-Appiah, F., Timmer, V., Quarshie-Sam, S., 2007. Early growth and
Forest Farming Series. Forestry Department, Ghana. nutritional response to resource competition in cocoa-shade intercropped systems.
Anglaaere, L., 2005. Improving the sustainability of cocoa farms in Ghana through Plant Soil 298, 243–254.
utilization of native forest trees in agroforestry systems. PhD Thesis, University of Jones, H., McNamara, N., Mason, W., 2005. Functioning of mixed species stands:
Wales, Bangor, UK. evidence from long-term forest experiment. In: Scherer-Lorenzen, M., Korner, C.,
Anim-Kwapong, G.J., 2003. Potential of some Neotropical Albizia species as shade trees Schulze, E. (Eds.), Forest Biodiversity and Function: Temperate and Boreal Systems.
when replanting cacao in Ghana. Agroforestry Systems 58 (3), 185–193. Springer-Verlag, Berlin, pp. 111–130.
Asare, R., 2005. Cocoa agroforests in West Africa: a look at activities on preferred trees Jose, S., Gillespie, A., Pallardy, S., 2004. Interspecific interactions in temperate
in the farming systems. Forestry and Landscape Working Paper, Arboretum agroforestry. Agroforestry Systems 61, 237–255.
Working Paper, No. 6. Forest and Landscape Denmark. Kay, N., 1997. Patterns of Corporate Evolution. Oxford University Press, New York.
Battese, G.E., Coelli, T.J., 1992. Frontier production functions, technical efficiency and panel Kelty, M., 1992. Comparative productivity in monocultures and mixed-species stands.
data: with applications to paddy farmers in India. Journal of Productivity Analysis 3, In: Kelty, M., Larson, B., Oliver, C. (Eds.), The Ecology and Silviculture of Mixed-
153–169. Species Forests. Kluwer Academic Publishers, Dordrecht, pp. 153–170.
Battese, G.E., Coelli, T.J., 1995. A model for technical inefficiency effects in a stochastic Kodde, D., Palm, F., 1986. Notes and comments: Wald criteria for jointly testing equality
frontier production function for panel data. Empirical Economics 20, 325–332. and inequality restrictions. Econometrica 54, 1243–1248.
Baumol, W., Panzar, J., Willig, R., 1988. Contestable Markets and the Theory of Industry Kumbhakar, S., Lovell, K., 2000. Stochastic Frontier Analysis. Cambridge University
Structure. Harcourt Brace Jovanivich, Orlando. Press, New York.
1518 A. Ofori-Bah, J. Asafu-Adjaye / Ecological Economics 70 (2011) 1508–1518

Leakey, R.R.B., 1998. Agroforestry in the humid lowlands of West Africa: some Scherer-Lorenzen, M., Korner, C., Schulze, E., 2005a. The functional significance of forest
reflections on future directions for research. Agroforestry Systems 40, 253–262. diversity: the synthesis. In: Scherer-Lorenzen, M., Korner, C., Schulze, E. (Eds.),
Lovell, C., Richardson, S., Travers, P., Wood, L.L., 1994. Resources and functionings: a Forest Diversity and Functions: Temperate and Boreal Systems. Springer, Berlin,
new view of inequality in Australia. In: Eichhorn, W. (Ed.), Models and Heidelberg, pp. 377–389.
Measurement of Welfare and Inequality. Spinger Verlag Press, Berlin. Scherer-Lorenzen, M., Korner, C., Schulze, E., 2005b. The functional significance of forest
Meeusen, W., van den Broeck, J., 1977. Efficiency estimation from Cobb–Douglas production diversity: the starting point. In: Scherer-Lorenzen, M., Korner, C., Schulze, E. (Eds.),
functions with composed error. International Economic Review 18, 435–444. Forest Diversity and Functions: Temperate and Boreal Systems. Springer, Berlin,
Morrison-Paul, C., Johnston, W., Frengley, G., 2000. Efficiency in New Zealand sheep and Heidelberg, pp. 3–12.
beef farming: the impact of regulatory reform. The Review of Economics and Schroth, G., 1999. A review of below-ground interactions in agroforestry, focusing on
Statistics 82 (2), 325–337. mechanisms and management options. Agroforestry Systems 43, 5–34.
Nair, P.K., 1993. An Introduction to Agroforestry. Kluwer Academic Publishers, Schroth, G., da Fonseca, G., Harvey, C., Gaston, C., Vasconcelos, H., Izac, A., 2004.
Dordrecht. Agroforestry and Biodiversity Conservation in Tropical Landscapes. Island Press,
Nyanteng, V., 1995. Prospects of Ghana's cocoa industry in the 21st century. In: Ruf, F., Washington D.C.
Siswoputroto, P. (Eds.), Cocoa Cycles: The Economics of Cocoa Supply. Woodhead Shephard, R., 1953. Cost and Production Functions. Princeton University Press, Princeton, NJ.
Publishing, Cambridge, UK, pp. 179–207. Shephard, R., 1970. The Theory of Cost and Production. Princeton University Press,
Panzar, J., Willig, R., 1977. Economies of scale and multi-output production. Quarterly Princeton, NJ.
Journal of Economics 91, 431–493. Sonwa, D.J., Wise, S.F., Schroth, G., Janssen, M.J.J., Shapiro, H., 2009. Market and
Parrish, J.D., Reitsma, R., Greenberg, R., 1998. Cocoa as a crop and conservation tool: livelihoods demand implications on plant diversity management inside cocoa
lessons from the Talamanca region of Costa Rica. 1st International Workshop on agroforest in forest landscape of West and Central Africa. XIII World Forestry
Sustainable Cocoa Growing. Panama City, March 30–April 2. Congress, Buenos Aires, Argentina, 18–23 October, pp. 1–11.
Pretzsch, H., 2005. Diversity and productivity in forests: evidence from long-term Takane, T., 2000. Incentives embedded in institutions: a case of shared contract in
experimental plots. In: Scherer-Lorenzen, M., Korner, C., Schulze, E. (Eds.), Forest Ghanaian cocoa production. The Developing Economies 37 (3), 374–397.
Diversity and Function: Temperate and Boreal Systems. Springer-Verlag, Berlin, Teal, F., Zeitlin, A., Maamah, H., 2006. Ghana Cocoa Farmers Survey 2004. Report to Ghana
pp. 41–64. Cocoa Board. Centre for the Study of African Economies, University of Oxford.
Rice, R.A., Greenberg, R., 2000. Cacao cultivation and the conservation of biological Vigneri, M., 2007. Drivers of cocoa production growth in Ghana. Project Briefing No 4.
diversity. Ambio 29 (3), 167–173. Overseas Development Institute, London.
Rolim, S., Chiarello, A., 2004. Slow death of Atlantic forest trees in cocoa agroforesty in Villano, R., Lucas, M., Pandey, S., 2005. An analysis of scope economies and technical
southeastern Brazil. Biodiversity and Conservation 13, 2679–2694. efficiency in intensive rainfed lowland rice-based cropping system in Northwest
Ruf, F., Schroth, G., 2004. Chocolate forests and monocultures: a historical review of Luzon, Philippines. Annual Conference of the Australian Agricultural and Resource
cocoa growing and its conflicting role in tropical deforestation and forest Society, Coffs Harbour, New South Wales.
conservation. In: Schroth, G., Fonseca, G., Harvey, C., Gascon, C., Vasconcelos, H., Wessel, M., Gerritsma, W., 1994. Re-thinking the shade policy for cocoa growing in
Izac, A.-M. (Eds.), Agroforestry and Biodiversity Conservation in Tropical West Africa. Proceedings of the 11th International Cocoa Research Conference,
Landscapes. Island Press, Washington, D.C., pp. 107–134. 18–24 July, 1993, Yamoussoukro, Côte d'Ivoire, pp. 681–686.
Ruf, F., Zadi, H., 1998. Cocoa: from deforestation to reforestation. Proceedings of the
First International Workshop on Sustainable Cocoa Growing. Panama City, March
30–April 2. Accessed on Dec 2009 at http://www.si.edu/smbc.

You might also like