1 s2.0 S2666154321001101 Main

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 47

Journal Pre-proof

The impact of adoption of artificial pollination technology in cocoa production:


Evidence from Ghana

Camillus Abawiera Wongnaa, Isaac Akurugu Apike, Suresh Babu, Dadson Awunyo-
Vitor, Afrane Baffour Kyei

PII: S2666-1543(21)00110-1
DOI: https://doi.org/10.1016/j.jafr.2021.100208
Reference: JAFR 100208

To appear in: Journal of Agriculture and Food Research

Received Date: 11 April 2021


Revised Date: 6 September 2021
Accepted Date: 7 September 2021

Please cite this article as: C.A. Wongnaa, I.A. Apike, S. Babu, D. Awunyo-Vitor, A.B. Kyei, The impact
of adoption of artificial pollination technology in cocoa production: Evidence from Ghana, Journal of
Agriculture and Food Research (2021), doi: https://doi.org/10.1016/j.jafr.2021.100208.

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. Please note that,
during the production process, errors may be discovered which could affect the content, and all legal
disclaimers that apply to the journal pertain.

© 2021 Published by Elsevier B.V.


The Impact of Adoption of Artificial Pollination Technology in Cocoa
Production: Evidence from Ghana

Camillus Abawiera Wongnaa1*, Isaac Akurugu Apike1, Suresh Babu2, Dadson Awunyo-Vitor1,
Afrane Baffour Kyei1
1
Department of Agricultural Economics, Agribusiness and Extension, Kwame Nkrumah University
of Science and Technology, Private Mail Bag, University Post Office, Kumasi, Ghana.
2
International Food Policy Research Institute, 1201, Eye Street, NW Washington, DC, USA

*Corresponding Author can be contacted through: wongnaaa@yahoo.com

of
ro
-p
re
lP
na
ur
Jo
Graphical Abstract
1a: NNM: Productivity 2a: NNM: Poverty

0 .2 .4 .6 .8 1 0 .2 .4 .6 .8 1
Propensity Score Propensity Score

Untreated Treated Untreated Treated

1b: NNM: Food Security 2b: NNM: Income

f
r oo
-p
re
lP

0 .2 .4 .6 .8 1
Propensity Score
0 .2 .4 .6 .8 1
Untreated Treated Propensity Score
na

Untreated Treated

Histogram of propensity score matching for Artificial Pollination Adoption


ur
Jo
The Impact of Adoption of Artificial Pollination Technology in Cocoa Production:
Evidence from Ghana

This study analyzed the impact of the adoption of artificial pollination on

productivity, income, poverty, and food security among cocoa farmers in Ghana.

Primary data was collected from 206 cocoa farmers, drawn through a multi-stage

sampling technique and analyzed using Propensity Score Matching. The study

of
revealed that households who adopted artificial pollination had improvement in their

ro
productivity, income, poverty, and food security. It was also revealed that adopters

-p
increased their productivity by close to 15.34% on average, earns between GHC

2756.84 to GHC 11074.38 more on average in terms of income, reduced their


re
poverty by an average of between 0.83% and 3.53%, and improved the food security
lP

by approximately 3% compared to non-adopters. Leveraging on the positive impact


na

of the adoption of artificial pollination, policymakers should take steps in

implementing artificial pollination to increase the yield of cocoa.


ur
Jo

Keywords: Adoption; Cocoa Production; Artificial Pollination; Propensity Score

Matching; welfare effects


1. Introduction

Cocoa continues to be a global commodity that has seen an annual production of

approximately 4 million tonnes worldwide since the year 2010 (Shahbandeh, 2019).

Demand for cocoa mainly emanates from Western Europe and developing economies in

Asia. For instance, in 2017, chocolate confectionery from sustainable cocoa sources

accounted for 8 percent of the total global retail market value of cocoa, this was mainly

of
fueled by increased demand from Western Europe and North America (Voora et al., 2019).

ro
In the past decade, West Africa has unequivocally been seen as the major producer

-p
of cocoa in the world, with Cote d'Ivoire, Ghana, Nigeria, and Cameroon contributing about
re
70% of global cocoa production (Shahbandeh, 2021). The average cocoa production in
lP

Africa has seen an increase of about 3% per year since the year 2000 (Wessel & Quist-
na

Wessel,2015). It is asserted by World Cocoa Foundation (2014) that Ghana is the second
ur

major producer of cocoa in the world, with Cote d'Ivoire being the first and that Ghana’s
Jo

cocoa production represents about 20% of world cocoa production. In Ghana, the cocoa

sector provides a source of livelihood to over four million households (Ghana Statistical

Service [GSS], 2018; Bangmarigu & Qineti 2018; Iddrisu et al., 2020; Danso-Abbeam et

al., 2020). The Institute of Statistical, Social and Economic Research (ISSER), (2017)

asserts that cocoa contributes about 80% to the country’s agricultural export. In this way,

the contribution of cocoa to the livelihoods of smallholder farmers cannot be

overemphasized.
The forest agro-ecological zones of Ghana are the main suitable areas for cocoa

production, and the western region of Ghana contributes about 56.5% of the total annual

cocoa production in Ghana (Codjoe et al., 2013). Cocoa offers basic earnings for the buying

of food and it plays a very important role in areas where there are food security issues

(Osei-Bagyina, 2012). Cocoa cultivation is encouraged due to its high prices, its all-year-

round yields, and its extraordinary commitment to smallholder livelihoods (Franzen &

of
Mulder, 2007).

ro
Averagely, the yield of dried cocoa beans in Ghana remains at 350 kg/ha. This is

-p
low as compared to Malaysia that records 1700kg/ha of cocoa and Ivory Coast that records
re
800kg/ha (Bosompem et al., 2011). Aneani & Ofori-Frimpong (2013) assert that poor
lP

pollination, low level of farmers’ adoption of agricultural technology as well as wrong use
na

of agricultural innovation are some of the causes of the low yield of cocoa recorded
ur

annually. Mainly, cocoa farms in Ghana continue to use natural and simple methods and
Jo

technologies even though there are a lot of technologies that could improve the yield of

cocoa (Bosompem et al., 2011).

Cocoa is pollinated by insects and the pollen is normally transferred by

Forcipomyia midges (Soria, 1980). The pollen transfer is such that fluctuations in rainfall

can influence midges which in turn also impact the efficacy of midges especially in moist

or humid conditions (Claus, 2018). To support the growth of cocoa production, the

ecosystem of cocoa must be considered to enhance system parameters that directly affect
cocoa yield (Gockowski & Sonwa, 2011); one very important determinant of the yield of

cocoa pollination. Cocoa pollination has been a global issue since 1925 (Glendinning 1972;

Adjaloo et al., 2013). This is because about 90% of flowers on cocoa trees fall off after

opening which leads to just about 10% well pollinated. Midges are the main determinants

of pollination, midges also depend heavily on how moist or humid the farm is (Bos et al.,

2007).

of
Artificial pollination is the best solution to natural pollination and its deficiencies

ro
(Vera-Chang et al., 2016; Forbes et al., 2019; Toledo-Hernández et al., 2020). Artificial

-p
pollination is said to have taken place where human intervention is involved in the
re
pollination process. It is therefore a mechanical process facilitated by human beings to
lP

pollinate plants. It is mostly used when natural pollination is limited. It is utilized to


na

increase the number as well as the quality of fruits on the tree (Forbes et al., 2017).
ur

Artificial pollination is the best solution to the problems associated with natural pollination,
Jo

it is important to analyze the impact of its adoption to further strengthen the implementation

or otherwise of the technology. A lot of researches on artificial pollination are centered on

the science and the benefits (Groeneveld et al., 2010; Chautá-Mellizo et al., 2012; Forbes

et al., 2019) whilst other studies focus on the awareness of the technology by farmers

(Frimpong-Anin et al., 2013; Elisante et al., 2019; Nyamekye & Danso, 2021; Mpondo et

al., 2021). There are also a lot of studies on the impact of agricultural technology adoption

(Adekambi et al., 2009; Wu et al., 2010 and Awotide et al. 2015). These studies provide
very great findings, however, the impact of the adoption of artificial pollination by cocoa

farmers in Ghana has been largely unexplored. This paper, therefore, seeks to answer the

question; what is the impact of artificial pollination on productivity, income, poverty, and

food security on the households of smallholder cocoa farmers?

The contribution of this study is two-fold, foremost, this paper contributes to the

literature on artificial pollination by broadening existing knowledge on the impact of

of
artificial pollination. Secondly, the findings of this paper will influence stakeholder actions

ro
and inactions within the cocoa sector towards the adoption of artificial pollination which

-p
will further improve farm productivity, income, and food security of farmers. A brief
re
literature review is presented in the next section, followed by the research methodology.
lP

The results and discussions also follow. Conclusion and recommendations are made in
na

section five.
ur
Jo

2. Literature review

The role of Agricultural innovations in increasing overall farm income is not

debatable. Artificial pollination is one such innovation. Kassie et al. (2009) assert that

artificial pollination reduces production costs, improves environmental benefits, and

increases crop yield, among others. According to Forbes et al. (2019), artificial pollination

regardless of the extent led to a significant increase in the yield of cocoa. During the season

where most trees display abundant flowers, Sánchez-Estrada & Cuevas (2020) on their part

found that artificial pollination resulted in a tremendous increase in the final fruit sets which
resulted in higher yields and profits. Three different methods of artificial pollination were

studied in clonal cacao relative to natural pollination by Vera-Chang et al. (2016) in

Mexico. It was found that amongst all the methods of pollination studied, natural

pollination had the lowest number of flowers that were pollinated and also the least fruit

weight, making artificial pollination, no matter the method used more superior and

rewarding. Toledo-Hernández et al. (2020) found that about 13% of easily accessible trees

of
and flowers that were partially hand-pollinated without application of fertilizer or

ro
insecticides led to an increase in the yield of cocoa of about 51% in Indonesia. Total

-p
pollination of the entire tree led to a 161% increase in yield likewise an increase in the net
re
income. A study by Gupta et al. (2017) also found that artificial pollination results in higher
lP

yields likewise food quality while hastening the physiological maturity period.
na

Some studies have looked at the impact of agricultural technology on the welfare
ur

of farmers; Adekambi et al. (2009) studied the impact of agricultural technology adoption
Jo

on poverty using NERICA rice varieties as a case in Benin. They found that the adoption

of NERICA varieties had a positive and significant impact on household expenditure.

Similarly, Wu et al. (2010) also assessed the impact of agricultural technology adoption on

farmers’ well-being using propensity score matching in rural China. They found that

improved upland rice technology had a robust and positive effect on the well-being of

farms. They measured well-being by income levels and the incidence of poverty.

Furthermore, Hailu et al. (2014) also found agriculture technology to have a positive impact
on farm income in northern Ethiopia. Awotide et al. (2015) in a similar study also found

adoption of improved varieties of cassava to have a positive and significant impact on asset

ownership.

Iddrisu et al. (2020) also studied the impact of participation in a UTZ-RA voluntary

cocoa certification scheme on smallholder welfare. They found that the UTZ-RA cocoa

certification program increased yield and income. It however had a negative impact on the

of
food security of smallholder cocoa farming households.

ro
Nakano et al. (2018) on their part also studied the impact of training on rice farming

-p
productivity in Tanzania. They found the training of rice farmers to have a positive impact
re
on the yield per hectare of rice farms.
lP

Habtewold (2021) also studied the impact of climate-smart agricultural technology


na

on multidimensional poverty in rural Ethiopia. He found that the adoption of climate-smart


ur

agricultural technology led to a high reduction in deprivation through an increase in income


Jo

emanating from improvement in production.

Finally, Lu et al. (2020) also studied the impact of rice varieties on household food

security in northern Ghana. They found consistent positive effects of the adoption of

improved rice varieties on household food security.


3. Methodology

3.1. Study Area

The study was conducted in the Amenfi West district of the Western North region

of Ghana. The Amenfi West District has a total land surface area of 1,287.265 sq. km.

The district is part of the wet semi-equatorial climatic zone. The area encounters two

main seasons that are dry and wet. The wet season comes about between April to October

of
whereas the dry season is between November to March. Figures for mean annual rainfall

ro
range between 1,250mm - 2,000mm. The mean annual temperature also ranges from

-p
25.5°C to 26.5°C. The district records high relative humidity between 75% - 90% in the
re
course of the wet season to 70% - 80% in the course of the dry season (GSS, 2014). The
lP

research area was chosen based on the fact that a comparatively enormous section of the
na

labor force in the district is involved in the production of cocoa and it is one of the
ur

foremost cocoa producing districts in Ghana. The district has a total population of about
Jo

92,152 and this represents approximately 3 percent of the region’s population (Ghana -

Population and Housing Census, 2010).

Economically, 76.9% of the whole population of the district are in active age

(15 years and above) of which 96.4% are working. The leading employer (95.1%) is the

private informal industry. The main economic activity of the people in the district is

cocoa production, with approximately 65–70% of the vegetation cover being cocoa

farms. This has affected land apportionment for food crops production (GSS, 2014).
3.2. Data

Primary and secondary data were employed. Data was collected using a detailed

structured questionnaire that contained both opened and closed-ended questions.

Secondary data was sourced from the Cocoa Health and Extension Division of COCOBOD

among others. The target of the study was all cocoa farmers within the Amenfi West

District. 10,021 farmers were identified in the 2010 population and housing census who

of
are cocoa farmers, constituting about 70% of the entire population of farmers in the district.

ro
206 cocoa farmers were selected among the 10,021 cocoa farmers as the sample for the

-p
study by using the sample selection formula used by Mendenhall et al. (1993):
re
𝑁𝑍 2 𝑃(1−𝑝)
lP

𝑛= (1)
𝑑2 (𝑁−1)+𝑍 2 𝑃(1−𝑃)
na

Where 𝑛 is the sample size, 𝑁 represents the population of cocoa farmers in the Amenfi
ur

West District, 𝑍 represents the 95% confidence level Z-statistic (1.96), 𝑑 represents the
Jo

margin of error at 95% confidence level, 𝑝 represents an estimated population proportion

(0.5) while 𝑞 represents the difference between 1 and p (𝑞 = 1 – 𝑝). The level of margin

of error was set to ± 0.0686, this is within the acceptable margin of error for sample size

determination (Suresh & Chandrashekara, 2012). Furthermore, since the exact certainty

of the proportion is not known, a value of 0.5 (50%) was used (Mendenhall et al., 1993).

This is because given the level of precision, ‘𝑝’ of 0.5 has the largest sample size.

Substituting the values in the formula, the sample size for the cocoa farmers was

computed as follows:
10021(1.96)2 0.5(1−0.5) 9624.17
𝑛 = (0.0686)2 (10021−1)+(1.96)2 0.5(1−0.5) = = 200 (2)
48.1141

However, on-field, there was an oversampling of 28 to make room for any incomplete

questionnaire, 22 of the questionnaires were incomplete, the total sample achieved was

206 instead of the proposed 200 sample size.

A multi-stage sampling technique was employed in selecting 206 cocoa farmers. The first

stage involved the purposive selection of the Amenfi West district. Amenfi West District

of
was selected due to its popularity in cocoa production in Ghana. The second and final stage

ro
-p
involved the random selection of 12 farming households from each of the 19 towns within
re
the district with the help of a list of cocoa farmers that was obtained from the district office
lP

of the Ministry of Food and Agriculture. In all, about 6 adopting farming households as
na

well as 6 non-adopters from each community were selected for the study. In all, 206
ur

farming households formed the sample size for the study. The additional six that were
Jo

added to the original sample size of 200 had even sides (3 adopters and 3 non-adopters)

3.3. Outcome variables and Analytical Frameworks

3.3.1 Outcome variables

The study employed propensity score matching (PSM) in analyzing the impact of

the adoption of artificial pollination in cocoa production for the major cocoa season in 2018.

The analysis focused on the impact of the technology on smallholder welfare. The welfare

indicators were productivity, income, poverty, and food security of the farming households.
Productivity was measured as a ratio of output per unit land area. Income was measured by

multiplying the quantity of cocoa produced by the price per bag of cocoa. For this study,

three variables were used as a proxy to assess the poverty levels of the smallholders. These

variables were their ability to pay for their kid’s school fees, their ability to open a bank

account to save excess money, and having a building with an iron roofing sheet. These

three variables were weighted to get a single indicator for measuring the farmers’ poverty

of
level (Alkire & Jahan 2018). For this study, farmers ability to pay fees and open a bank

ro
account were weighted ‘5’ if farmer answers yes to either of them and ‘0’ otherwise

-p
whereas their ability to have a building with an iron roofing sheet was weighted ‘10’ if the
re
farmer answers yes and ‘0’ if otherwise. The scores were then summed to give a unit
lP

measure for the poverty level of the farming household. A good score per this study should
na

be greater or equal to 50% of the overall score. According to FAO (1996), food security is
ur

a state where “all people at all times have both physical and economical access to sufficient
Jo

food to meet their dietary needs for a productive and healthy life”. For this study, the

Household Food Insecurity Access Scale (HFIAS) was used to assess the food security of

the farming households in the past four weeks before the study (Pandey & Bardsley, 2019).

In line with Coates et al. (2007) and Iddrissu et al. (2020), USAID’s standard method of

measuring HFIAS was adopted in this study as a proxy measure of the food security status

of the households. The HFIAS was measured using 9 attributes or questions, the questions

asked under the HFIAS framework have been presented in Appendix 2.0. All the nine
questions were rated on a 4-point scale, viz. 0=No; 1=Rarely (once or twice in the past 4

weeks); 2=Sometimes (three to ten in the past 4 weeks) and 3=Often (more than ten times

in the past 4 weeks). The HFIAS score variables were calculated for each farming

household by summing the codes for each frequency of occurrence question. All those who

code ‘no’ are given a score of 0, ‘Rarely’ given code 1, and on till the last scale ‘Often’

which is scored 3. The acceptable region of the score is between 0 – 27 (Pandey & Bardsley,

of
2019). The closer the score to 0 the more food secured the household and the closer the

ro
score to 27 the more insecure the farming household.

3.3.2 Analytical Framework


-p
re
To determine the impact of adopting artificial pollination on the outcome variables
lP

discussed above, Propensity Score Matching (PSM) was used to assess the difference in
na

outcome between adopting and non-adopting households.


ur

PSM was applied in two stages. A probit model was used to estimate the probability of
Jo

adopting artificial pollination. The adopting decision of respondents ranges between 0 and

1, with 1 indicating the probability of adoption and 0 indicating otherwise (Iddrisu et al.,

2020). The first stage of the PSM, which is the probit model is specified as follows;

𝐴𝑃𝐴 = 𝛽0 + 𝛽1 𝐺𝑒𝑛𝑖 + 𝛽2 𝐴𝑔𝑒𝑖 + 𝛽3 𝑀𝑆𝑡𝑎𝑡𝑖 + 𝛽4 𝐸𝑑𝑢𝑖 + 𝛽5 𝐻𝐻𝑆𝑖 + 𝛽6 𝐹𝑆𝑖𝑧𝑒𝑖 +

𝛽7 𝐸𝑥𝑡𝑖 + 𝛽8 𝐹𝐵𝑂𝑖 + 𝛽9 𝐶𝑟𝑒𝑑𝑖 + 𝛽10 𝑂𝑓𝑓𝑖𝑛𝑐𝑖 + 𝜀𝑖 (3)

Where APA is Artificial Pollination Adoption as well as the dependent variable, defined

as 1 for adopters and 0 for nonadopters.


𝛽0 is the intercept,

𝛽1 + 𝛽2 … … 𝛽𝑘 are the coefficients of the X variables.

𝜀𝑖 is the error term.

Following Zeweld et al. (2015), the second part of PSM involved the use of

matching econometrics to ascertain whether or not artificial pollination adoption has

unambiguously led to a significant increase in each of the aforementioned outcome

of
variables. In assessing the impact of any intervention, finding a comparable group of

ro
treated and a control group is one of the problems. In any case, coordinating econometrics

-p
gives a promising instrument to cure that while assessing the normal treatment impacts
re
(Zeweld et al., 2015). Nian et al. (2019) also asserted that matching is a method widely
lP

used in estimating the average treatment effects of a binary treatment on a continuous scalar
na

outcome. It utilizes a non-parametric regression procedure to construct the counterfactual


ur

based on the assumption of selection on observables. Artificial pollination adoption is a


Jo

binary treatment with productivity, income, food security, and poverty status as outcome

variables. Smallholder cocoa farmers who adopted this technology are the treatment group

and non-adopters are the control group. Propensity score matching matches the treated and

control group households’ farmers with similar observable characteristics to estimate the

effect of adoption as the difference in the mean value of the outcome variables. Following

the empirical literature of program evaluation, let 𝑌1 be the outcome variables (productivity,

income, food security, and poverty status) when a farmer 𝑖 is subject to treatment (𝐶 = 1)
and 𝑌0 be the productivity, income, food security, and poverty status when a farmer 𝑖 is in

the control group (𝐶 = 0). The observed outcome therefore becomes:

𝑌 = 𝐶𝑌1 + (1 − 𝐶)𝑌0 (4)

𝑌1 is observed when 𝐶 = 1; and 𝑌0 is observed when 𝐶 = 0. The average effect of treatment

(artificial pollination) on the treated is to be identified and it is defined as:

𝐴𝑇𝑇 = 𝐸(𝑌1 − 𝑌0 |𝐶 = 1) = 𝐸(𝑌1 |𝐶 = 1| − 𝐸(𝑌0 |𝐶 = 1)) (5)

of
Unlike the observed 𝐸(𝑌1 |𝐶 = 1) 𝐸(𝑌0 |𝐶 = 1)d, cannot be observed since it is not found

ro
in the data. This problem can be solved by creating the counterfactual 𝐸(𝑌0 |𝐶 = 1) that is

-p
what would have been the productivity, income, food security and poverty status of
re
farmers' households adopting artificial pollination had they not adopted, by matching the
lP

treatment and the controlled farming households. Certain assumptions must be met for
na

matching to be valid and the primary assumption underlying matching estimators is the
ur

Conditional Independence Assumption (CIA). Conditional Independence Assumption


Jo

states that the decision to adopt is randomly conditional on observed covariates X. That is,

(𝑌1, 𝑌0 ) ⊥ 𝐶|𝑋 (6)

The assumption implies that the counterfactual outcome in the treated group is the same as

the observed outcomes for the control group;

𝐸(𝑌0 |𝑋, 𝐶 = 1) = 𝐸(𝑌0 |𝑋, 𝐶 = 1) = 𝐸(𝑌0 |𝑋) (7)

The assumption rules out of the selection into the program based on unobservable gains

from adopting. The Conditional Independence Assumption requires that the set of X’s
should contain all the variables that jointly influence both the outcome with non-treatment

and treatment. The average treatment effect, ATT can be computed under the CIA as

follows:

𝐴𝑇𝑇 = 𝐸(𝑌1 − 𝑌0 |𝑋, 𝐶 = 1) = 𝐸(𝑌1 |𝑋, 𝐶 = 1| − 𝐸(𝑌0 |𝐶 = 1)) (8)

Rosenbaum & Rubin (1983) demonstrate that as opposed to coordinating along 𝑋, one can

coordinate along 𝑃(𝑋), a solitary record variable that abridges covariates. This index is

of
known as the propensity score and it is used to overcome the curse of dimensionality arising

ro
from too many dimensions of the covariates making matching of farmers based on

-p
observed covariates undesirable or not feasible. The propensity score (response probability)
re
is the conditional probability that farming household 𝑖 adopts artificial pollination given
lP

covariates:
na

𝑝(𝑋) = 𝑝𝑟(𝐶 = 1)|𝑋 (9)


ur

Equation 9 can then be written as:


Jo

𝐴𝑇𝑇 = 𝐸(𝑌1|𝑃(𝑋), 𝐶 = 1) − 𝐸(𝑌0 |𝑃(𝑋), 𝐶 = 1) (10)

The first step in the PSM approach is the estimation of the propensity score. The propensity

score is estimated by a simple binary regression model. According to Caliendo & Kopeinig

(2005), for a binary treatment case, the probability of participation versus non-participation

needs to be estimated and this can be done using either the logit or probit model as they

usually yield similar results. The binary probit regression model is employed in this study.

Nian et al. (2019) suggested that since the p-score is a continuous variable, exact matches
may rarely be attained and certain distances between the treated and untreated farmers have

to be accepted. They proposed to solve this problem, the treated and non-treated farmers

are matched based on their score using nearest neighbors, kernel, and radius matching

estimators. These three methods identify for each farmer the closest propensity score in the

opposite intervention status and then estimates the investment effect as the mean difference

of farmers’ productivity, income, food security, and poverty status between each pair of

of
matched farmers.

ro
Three main PSM approaches, also mentioned above, are used in this analysis, viz.

-p
Nearest Neighbor Matching (NNM), Radius Matching, and Kernel Based Matching (KBM).
re
Using several matching procedures widens the possibility of getting the best match since
lP

this is based on the fact that the process of coming up with the best match is more or less a
na

trial-and-error exercise.
ur

3.3.2.1 Sensitivity Analysis for Matching Estimators


Jo

One of the assumptions of the sensitivity examination is that the treatment task is not

unconfounded given a lot of covariates. This assumption reveals the weakness of the PSM

approach which faces the challenge of endogeneity due to omitted confounding variables.

This has made the checking of estimated treatment effects a very critical issue in assessing

the impact of interventions. Jawara (2020) asserts that the suitable control approach of

hidden bias is to examine the sensitivity of significance levels. To better explain what

happens where there is a hidden bias, 𝑒 𝑦 presents a measure of the degree of departure from
a study that is free from hidden bias, and this assesses the sensitivity of the estimated

participating effect. Several values of 𝑒 𝑦 are calculated to test the sensitivity of the

significance levels. This test helps to identify the critical impact level at which significance

would be lost. This represents the level of unobservable covariate at which the inference

about the treatment effect will be undermined. The sensitivity of the results obtained from

the PSM results was done using rbounds syntax (Diprete & Gangl, 2004; Iddrisu et al.,

of
2020).

ro
4. Results and Discussion

-p
4.1. Socio-Economic Characteristics of Respondents
re
Table 1 details the socioeconomic characteristics of the respondents. The mean age of the
lP

cocoa farmers was 39.40 years and the mean age of adopters was 36.19 years, while the
na

mean age of non-adopters was 42.60 years. These outcomes demonstrate that most of the
ur

respondents are in their prime age. Although there were wide contrasts in their ages,
Jo

adopters were more youthful than non-adopters. Additionally, as indicated by Ajewole

(2010), heads of the family unit who are youthful are bound to adopt artificial pollination

than the older family heads. This may presumably be because the more youthful farmers

may have been exposed to new agricultural and farming innovations and wouldn't fret

giving it a shot (Mwangi & Kariuki, 2015). Older farmers mostly will adhere to their

traditional methods of cultivating and may be somewhat reluctant to changes. Along

these lines, they are less prone to change and inventive regarding the adoption of

technology (Enete & Igbokwe, 2009).


It is evident from Table 1 that the average years of education of the respondent were 9.71

years. Specifically, the average years of education for adopters and non-adopters were

9.64 and 9.80 respectively. This shows that the greater part of the respondents had

education up to Junior High School. As argued by Orinda (2013), training and education

could almost certainly influence farmers' readiness to be the early adopters to take

advantage of innovations and technology. Moreover, as evident from appendix 1, 16% of

of
respondents had no formal education. Among the adopters, 12% had no formal education

ro
whiles 19% out of non-adopter also had no formal education. This is indicative of the fact

-p
that education was higher among the adopters than the non-adopters. This could be
re
because the literate relatives have a better ability to decipher information and have a
lP

propensity of influencing the household’s choice to adopt artificial pollination (Kassie et


na

al., 2009).
ur

The average family size of farmers was 11.15 out of which an average of 6.47 were
Jo

adults aged 18 years or more and 5.00 were children under 18 years. The mean household

size was 11.93 among adopters, with 7.86 being 18 years and above and 4.16 under 18

years, on the other hand, non-adopters had a mean family size of 10.36 with 5.57 being

18 years and above and the remaining being below 18 years. The family unit size was

high among the adopters. This could presumably be because artificial pollination requires

intensive and bigger labor, subsequently, bigger family size can compensate for the

additional work required (Ajewole, 2010).


Moreover, on the experience of cocoa farmers, it is evident from Table 1 that the mean

years of experience of a respondent is 15.50 years. The adopters’ mean year of

experience in cocoa farming is 12.53 years while non-adopters had average years of

experience of 8.46 years. More years of experience in cocoa farming helps farmers in

assessing the benefits of technologies and be the early adopters of innovations (Obisesan,

2014). Experienced farmers appear to have better knowledge and information aggregated

of
over a long time. The mean farm size of the respondents was 4.30 hectares of which an

ro
average of 3.22 hectares is allocated to cocoa production. Adopters had an average

-p
agricultural farmland of 3.63 hectares of which 2.77 hectares are used for cocoa
re
production. Non-Adopters on the other hand had an average of 5.00 hectares of farmland
lP

with an average of 3.70 hectares allocated for cocoa production. The average yield was
na

373.7 kg/ha of cocoa in the major season and 138 kg/ha in the minor season. Adopters
ur

had an average yield of 533 kg/ha during major seasons and 202.7 kg/ha during the minor
Jo

seasons.

Nonadopters make a mean yield of 255 kg/ha of cocoa during the main season and 90

kg/ha during the minor season. An average amount of GH¢406.08 is made every month

from other economic activities aside from farming. The mean income acquired by

adopters
from other economic activities aside farming was GH¢395.83 every month and GH¢416.33

every month for non-adopters. This finding shows that even though adopters had relatively less

farmland size and its relating allocation for cocoa production juxtaposed with non-adopters, they

make relatively more harvest than non-adopters both in major and minor seasons. This could

presumably be a result of the direct results of the adoption of artificial pollination (Gelgo et al.,

2016). Likewise, non-adopters get a greater amount of their income from other economic

activities aside from farming than adopters. This presumably is as a result of their non-adoption

of
of agricultural technology, this leads to lower yield, hence lesser income, and for that matter,

ro
they go into other income-generating ventures aside from farming.

-p
Finally, the average distance to the closest market was 4.45km. The mean distance to the nearest
re
market was 5.85km among adopters while the mean distance for the non-adopters was 2.90. This
lP

means that markets of non-adopters are relatively far. This contradicts the findings of IFPRI
na

(2012) and Martey et al. (2014), who assert that there is a positive relationship between distance

from markets and agricultural technology adoption.


ur

4.2. Factors influencing Artificial Pollination Adoption in Cocoa Production


Jo

Table 2 shows the probit results of the factors that influence the adoption of artificial

pollination in cocoa production. It is evident from Table 2 that age, extension services,

experience, off-farm income, and household size were the factors that influence the adoption of

artificial pollination. The result shows that as the age of a household head increases by one, the

likelihood of adopting artificial pollination reduces by 7.4%. this means that as the age of

household heads increases, they are discouraged to adopt artificial pollination. This may be a

result of the fact that older household heads do not have the strength to employ and explore new

technologies. This result is consistent with the findings of Akinwale & Folarin (2018) who
studied factors that influence farmers’ adoption of cocoa hybrid technology in Oyo State,

Nigeria, they found the age of farmers to be a significant factor in the adoption of cocoa hybrid

technology.

It is also evident from Table 2 that an additional visit by extension officers to farmers

increases the probability of farmers adopting artificial pollination. This depicts a positive

relationship between the extent of extension visits and the likelihood of adopting the technology.

Gelgo (2016) also found that extension officers normally prepare demonstration plots to give the

of
farmers hands-on training and can help farmers experiment on new agricultural technologies. This

ro
leads to an increase in the likelihood of adopting agricultural technologies.

-p
The experience of the heads of households has a positive relationship with the likelihood
re
of adopting artificial pollination. This means that an increase in the number of years of experience
lP

will lead to an increase in the likelihood of adopting artificial pollination. This is probably a result
na

of the fact that farmers with more years of experience can assess the merits and demerits of the

agricultural technology due to their accumulated knowledge over years (Obisesan, 2014). This
ur

finding is consistent with the findings of Avane et al. (2021) who studied the perception and
Jo

determinants of the adoption of organic fertilizer for cocoa production in Ghana. They found that

the experience of the farmer had had a positive influence on the adoption of organic fertilizer.

Off-farm income was also significant in determining the likelihood of adopting artificial

pollination. Off-farm income has a positive relationship with the likelihood of adopting artificial

pollination. Off-farm income earners can be seen to be entrepreneurial and risk-takers, hence their

likelihood of adopting artificial pollination. Farmers who engage in off-farm economic activities

also have the financial ability to employ new technologies. This finding is in line with the findings
of Iddrisu et al. (2020) who found non-farm income to positively influence the participation of

cocoa farming households in the UTZ-RA voluntary cocoa certification scheme.

Household size also has a positive relationship with the likelihood of adopting artificial

pollination. This means that an addition to a particular household increases the likelihood of

adopting artificial pollination. It could be as a result of the fact that a bigger household provides a

bigger workforce. Farmers are likely to adopt artificial pollination when they have a bigger family

size that can provide a workforce on their farms. This study supports the findings of Zheng et al.

of
(2021), who found the household size to have a positive influence on the adoption of organic soil

ro
amendments in China.

-p
4.3. Productivity, Income, Poverty and Food Security of Adopters and Non-Adopters
re
Table 3 compares the performance of adopters and non-adopters of artificial pollination in
lP

terms of productivity, income, poverty, and food security. The results showed that cocoa farmers
na

saw improvement in their productivity, income, poverty level, and food security after the adoption

of artificial pollination and these were significant at 1%.


ur

4.4. Testing the Balancing Properties of Propensity Scores and Covariates


Jo

To estimate the impact of artificial pollination adoption on income, the balancing property

scores should be checked to test if the observations have had the same distribution of propensity

scores or not. Tolemariam (2010) argues that the balancing test seeks to examine if, at each value

of the propensity score, a given characteristic has the same distribution for the treated and

comparison groups. The results are presented in Table 4 below. The results showed that 10

variables exhibited a significant mean difference before matching while no variable showed a

significant mean difference after the match. This indicates covariate balance between the sample

participants and non-participants of artificial pollination. Thus, it can be concluded that the
specification was successful in terms of balancing the distribution of covariates between the

matched adopters and non-adopters of artificial pollination. Further to the results from Table 4, the

overall balance test statistics showed that the likelihood ratio test was insignificant at p>chi 2 =

0.806 confirming both adopter and non-adopters had the same distribution after matching. The

result is present in Table 5.

4.5. Impact of Artificial Pollination Adoption in Cocoa Production

Taking participation (adoption decision) as 1 if the household has been participating in the

of
adoption of artificial pollination, and 0 otherwise, all the variables that were assumed to influence

ro
adoption decision were included to predict the probability of each household. The overall

-p
estimated propensity score was between 0.00 and 1.00 as depicted in Table 6. Amongst the
re
adopters, the score lies between 0.375 and 1.000 but for non-adopters, it lies between 0.00 and
lP

1.00. This indicates that the region of common support would lie between 0.375 and 1.000.
na

Furthermore, Table 6 reveals that the overall average propensity score among the respondents was

0.50 implying that the average probability of participants adopting the artificial pollination
ur

technology in their cocoa farms was 50%.


Jo

Figure 1 is also the visual presentation of the density distribution of the estimated

propensity scores for the adopters and non-adopters. It shows the number of cocoa farmers on

support and those not on support. The figure reveals substantial overlap in the distribution of the

propensity scores of both adopters and non-adopters. The impact of the adoption of artificial

pollination on smallholder productivity, income, poverty level, and food security were estimated

using the PSM and the results are presented in Table 7. The results revealed that adoption of

artificial pollination led to a 1.93% to 15.34% increase in yield, GHC 2792.96 to GHC 7133.90

increase in income, 0.83% to 3.53% reduction in poverty, and 1 to 3% increase in food security.
This implies that the adoption of artificial pollination is vital to increasing smallholder welfare.

This finding is consistent with those of similar adoption and welfare studies such as Iddrisu et al.

(2020), who found the participation in the UTZ-RA voluntary certification program to have a

positive impact on yield and income, Lu et al. (2020) also found the adoption of rice varieties to

have a positive impact on food security.

4.6 Sensitivity Analysis

This section analyses the robustness of the results from the treatment effects using rbounds

of
syntax. The main result was to check the degree to which the estimated treatment effects were free

ro
of unobserved covariates. This was done by comparing the baseline propensity score and treatment

-p
effect (Lu, 2020). This approach was to check how strongly an unmeasured variable can influence
re
the selection process to undermine its implications of the matching analysis. The results are
lP

summarized in Table 8. From Table 8, even though all the matching techniques showed robust
na

treatment effect estimates in terms of hidden bias on the unobserved covariates on productivity,

the radius matching technique was more robust among them. Even though there were differences
ur

in the matched pair (𝑒 𝑦 = 2) in the unobserved characteristics, the impact of artificial pollination
Jo

adoption would remain significant at a 1% significance level. That notwithstanding, it should be

noted that the results from the sensitivity analysis are extreme instances even though they indicate

information about uncertainty within the matching estimators of treatment effects (Rosenbaum,

2002).

5.0 Conclusion and Recommendations

This study analysed the impact of the adoption of artificial pollination using productivity,

income, poverty, and food security as indicators of welfare. The number of extension visits, the
years of experience of heads of farming households, and the size of farming households had a

positive significant relationship with the adoption of artificial pollination. The age of the household

heads also had a negative significant relationship with the adoption of artificial pollination.

The study also revealed that households who adopted artificial pollination had

improvements in their productivity, income, poverty, and food security. The results revealed that

adoption of artificial pollination led to a 1.93% to 15.34% increase in yield, GHC 2792.96 to GHC

7133.90 increase in income, 0.83% to 3.53% reduction in poverty, and 1% – 3% increase in food

of
security. This implies that artificial pollination has a positive impact on the welfare of cocoa

ro
farmers and for that matter a great tool for dealing with the welfare concerns of cocoa farmers.

-p
This study makes the following recommendations; first of all, extension visits to cocoa
re
farms are highly recommended since it allows demonstrating the importance of adopting artificial
lP

pollination, hence government and other stakeholders should invest in extension services to help
na

explain and demonstrate the technology to farmers. Secondly, governments, NGO’s and other

stakeholders that are interested in changing the livelihoods of cocoa farmers are encouraged to
ur

support cocoa farmers financially and technically to implement artificial pollination since this
Jo

study has shown the overriding effect of the technology on farmer welfare.

References
Adekambi, S. A., Diagne, A., Simtowe, F., & Biaou, G. (2009). The impact of Agricultural
Technology adoption on Poverty: The case of NERICA rice varieties in Benin (No. 1005-
2016-78928, 1-16).
Adjaloo, M., Banful, B. K. B., & Oduro, W. (2013). Evaluation of breeding substrates for cocoa

of
pollinator, Forcipomyia spp. and subsequent implications for yield in a tropical cocoa
production system. American Journal of Plant Science, 4,203-210.

ro
Ajewole, O. C. (2010). Farmers response to adoption of commercially available organic

-p
fertilizers in Oyo state, Nigeria. African Journal of Agricultural Research, 5(18), 2497-
re
2503.
Akinwale, J. A., & Folarin, O. E. (2018). Factors Influencing Farmers’ Adoption of CRIN TC
lP

Cocoa Hybrid Technology in Oyo State, Nigeria. Pelita Perkebunan (a Coffee and Cocoa
na

Research Journal), 34(3), 191-198.


Ali, E. B., Awuni, J. A., & Danso-Abbeam, G. (2018). Determinants of fertilizer adoption among
ur

smallholder cocoa farmers in the Western Region of Ghana. Cogent Food & Agriculture,
Jo

4(1), 1538589.
Alkire, S., & Jahan, S. (2018). The new global MPI 2018: aligning with the sustainable
development goals. OPHI Working Paper 121, University of Oxford.
Aneani, F., & Ofori-Frimpong, K. (2013). An analysis of yield gap and some factors of cocoa
(Theobroma cacao) yields in Ghana. Sustainable Agriculture Research, 2(526-2016-
37857).
Asamoah, M., & Baah, F. (2003). Improving research-farmer linkages: The role of CRIG. In 4th
International Seminar on Cocoa-Pests and Diseases (INCOPED), Accra, 19-21.
Awotide, B. A., Alene, A. D., Abdoulaye, T., & Manyong, V. M. (2015). Impact of agricultural
technology adoption on asset ownership: the case of improved cassava varieties in
Nigeria. Food Security, 7(6), 1239-1258.
Bangmarigu, E., & Qineti, A. (2018). Cocoa production and export in Ghana (No. 2038-2018-
3066).
Bos, J. L., Rehmann, H., & Wittinghofer, A. (2007). GEFs and GAPs: critical elements in the
control of small G proteins. Cell, 129(5), 865-877
Bosompem, M. A. R. T. I. N., Kwarteng, J. A., & Ntifo-siaw, E. D. W. A. R. D. (2011). Towards
the implementation of precision agriculture in cocoa production in Ghana: Evidence from
the cocoa high technology programme in the Eastern region of Ghana. Journal for
Agrcultural Research and Development, 10(1), 11-17.
Breisinger, C., Xinshen, D., Shashidhara, K., & James, T. (2008). The Role of Cocoa in Ghana’s

of
Future Development; Ghana Strategy Support Program (GSSP), Background Paper No.
GSSP 11, International Food Policy Research Institute (IFPRI), Washington, D.C. USA.

ro
Chautá-Mellizo, A., Campbell, S. A., Bonilla, M. A., Thaler, J. S., & Poveda, K. (2012). Effects

-p
of natural and artificial pollination on fruit and offspring quality. Basic and Applied
re
Ecology, 13(6), 524-532.
Claus, G., Vanhove, W., Van Damme, P., & Smagghe, G. (2018). Challenges in cocoa
lP

pollination: the case of Côte d’Ivoire. Pollination in plants, 39-58.


na

Coates, J., Bilinsky, P., & Coates, J. (2007). Household Food Insecurity Access Scale (HFIAS)
for Measurement of Food Access: Indicator Guide VERSION 3 Household Food
ur

Insecurity Access Scale (HFIAS) for Measurement of Food Access : Indicator Guide
Jo

VERSION 3. August.
Codjoe, F. N. Y., Ocansey, C. K., Boateng, D. O., & Ofori, J. (2013). Climate change awareness
and coping strategies of cocoa farmers in rural Ghana. Journal of Biology, Agriculture
and Healthcare, 3(11), 19-29.
Danso-Abbeam, G., & Baiyegunhi, L. J. (2017). Adoption of agrochemical management
practices among smallholder cocoa farmers in Ghana. African Journal of Science,
Technology, Innovation and Development, 9(6), 717-728.
Danso-Abbeam, G., Baiyegunhi, L. J., & Ojo, T. O. (2020). Gender differentials in technical
efficiency of Ghanaian cocoa farms. Heliyon, 6(5), e04012.
Djokoto, J. G., Owusu, V., & Awunyo-Vitor, D. (2016). Adoption of organic agriculture:
Evidence from cocoa farming in Ghana. Cogent Food & Agriculture, 2(1), 1242181.
Elisante, F., Ndakidemi, P. A., Arnold, S. E., Belmain, S. R., Gurr, G. M., Darbyshire, I., ... &
Stevenson, P. C. (2019). Enhancing knowledge among smallholders on pollinators and
supporting field margins for sustainable food security. Journal of Rural Studies, 70, 75-
86.
Enete, A. A., & Igbokwe, E. M. (2009). Cassava market participation decisions of producing
households in Africa. Tropicultura, 27(3), 129-136.
Food and Agriculture Organization of the United Nations, 1996. Rome declaration on world food
security and World Food Summit plan of action. In: World Food Summit; Rome, Italy;
November 1-17.

of
Forbes, S. J., & Northfield, T. D. (2017). Increased pollinator habitat enhances cacao fruit set
and predator conservation. Ecological applications, 27(3), 887-899.

ro
Forbes, S. J., Mustiga, G., Romero, A., Northfield, T. D., Lambert, S., & Motamayor, J. C.

-p
(2019). Supplemental and Synchronized Pollination May Increase Yield in Cacao.
re
HortScience, 54(10), 1718-1727.
Franzen, M., & Mulder, M. B. (2007). Ecological, economic and social perspectives on cocoa
lP

production worldwide. Biodiversity and Conservation, 16(13), 3835-3849.


na

Frimpong-Anin, K., Adjaloo, M. K., Kwapong, P. K., & Oduro, W. (2014). Structure and
Stability of Cocoa Flowers and Their Response to Pollination. Journal of Botany.2014:1-
ur

6.
Jo

Frimpong-Anin, K., Kwapong, P. K., & Gordon, I. (2013). Cocoa farmers awareness of
pollination and its implication for pollinator-friendly practices. Res. and Rev. Biosci,
7(12), 504-512.
Gelgo, B., Mshenga, P., & Zemedu, L. (2016). Analysing the determinants of adoption of
organic fertilizer by smallholder farmers in Shashemene District, Ethiopia. International
Journal of Agricultural Economics, 4(1), 117-124.
Glendinning, D. R. (1972). Natural pollination of cocoa. The New Phytologist, 71(4), 719-729.
Gockowski, J., & Sonwa, D. (2011). Cocoa intensification scenarios and their predicted impact
on CO 2 emissions, biodiversity conservation, and rural livelihoods in the Guinea rain
forest of West Africa. Environmental management, 48(2), 307-321.
Groeneveld, J. H., Tscharntke, T., Moser, G., & Clough, Y. (2010). Experimental evidence for
stronger cacao yield limitation by pollination than by plant resources. Perspectives in
Plant Ecology, Evolution and Systematics, 12(3), 183-191.
Gupta, A., Godara, R., Sharma, V., & Panda, A. (2017). Artificial pollination: a tool for
improving fruiting traits in date palm (Phoenix dactylifera L.). Chemical Science Review
and Letters, 6, 1312-1320.
Habtewold, T. M. (2021). Impact of climate-smart agricultural technology on multidimensional
poverty in rural Ethiopia. Journal of Integrative Agriculture, 20(4), 1021-1041.
Hailu, B. K., Abrha, B. K., & Weldegiorgis, K. A. (2014). Adoption and impact of agricultural

of
technologies on farm income: Evidence from Southern Tigray, Northern Ethiopia.
International Journal of Food and Agricultural Economics (IJFAEC), 2(1128-2016-

ro
92058), 91-106.

-p
Iddrisu, M., Aidoo, R., & Wongnaa, C. A. (2020). Participation in UTZ-RA voluntary cocoa
re
certification scheme and its impact on smallholder welfare: Evidence from Ghana. World
Development Perspectives, 20, 100244.
lP

Ilesanmi, J. O., & Afolabi, J. A. (2020). Determinants of Adoption of Improved Cocoa


na

Technologies in Ekiti State, Nigeria. International Journal of Agricultural Economics,


5(2), 36.
ur

International Food Policy Research Institute (IFPRI). 2012. IFPRI-Africa Lead report on
Jo

Assessment of Nigeria’s Agriculture Transformation Agenda(ATA) and Capacity


Building Needs. Available at: http://pdf.usaid.gov/pdf_docs/PA00K1SH.pdf [Accessed:
2021, August 06]
Jawara, H. (2020). Access to savings and household welfare evidence from a household survey
in The Gambia. African Development Review, 32(2), 138-149.
Kassie, B. T., Hengsdijk, H., Rötter, R., Kahiluoto, H., Asseng, S., & Van Ittersum, M. (2013).
Adapting to climate variability and change: experiences from cereal-based farming in the
Central Rift and Kobo Valleys, Ethiopia. Environmental Management, 52(5), 1115-1131.
Kassie, M., Zikhali, P., Manjur, K., & Edwards, S. (2009, August). Adoption of sustainable
agriculture practices: Evidence from a semi‐ arid region of Ethiopia. In Natural
Resources Forum (Vol. 33, No. 3, pp. 189-198). Oxford, UK: Blackwell Publishing Ltd.
Lavison, R. K. (2013). Factors influencing the adoption of organic fertilizers in vegetable
production in Accra (Doctoral dissertation, University of Ghana).
Lu, D., Guo, F., & Li, F. (2020). Evaluating the causal effects of cellphone distraction on crash
risk using propensity score methods. Accident Analysis & Prevention, 143, 105579.
Martey, E., Etwire, P. M., Wiredu, A. N., & Dogbe, W. (2014). Factors influencing willingness
to participate in multi-stakeholder platform by smallholder farmers in Northern Ghana:
implication for research and development. Agricultural and Food Economics, 2(1), 1-15.
Martey, E., Wiredu, A. N., Asante, B. O., Annin, K., Dogbe, W., Attoh, C., & Al-Hassan, R. M.
(2014). Factors influencing participation in rice development projects: the case of

of
smallholder rice farmers in Northern Ghana. International Journal of Development and
Economic Sustainability, 1(2), 13-27.

ro
Mendenhall, C. L., Moritz, T. E., Roselle, G. A., Morgan, T. R., Nemchausky, B. A., Tamburro,

-p
C. H., ... & Veterans Affairs Cooperative Study Group 275. (1993). A study of oral
re
nutritional support with oxandrolone in malnourished patients with alcoholic hepatitis:
results of a Department of Veterans Affairs cooperative study. Hepatology, 17(4), 564-
lP

576.
na

Mpondo, F. T., Ndakidemi, P. A., & Treydte, A. C. (2021). Balancing Bees and Livestock:
Pastoralist Knowledge, Perceptions and Implications for Pollinator Conservation in
ur

Rangelands, Northern Tanzania. Tropical Conservation Science, 14,


Jo

19400829211028127.
Mwangi, M., & Kariuki, S. (2015). Factors determining adoption of new agricultural technology
by smallholder farmers in developing countries. Journal of Economics and sustainable
development, 6(5), 208-216.
Nakano, Y., Tsusaka, T. W., Aida, T., & Pede, V. O. (2018). Is farmer-to-farmer extension
effective? The impact of training on technology adoption and rice farming productivity in
Tanzania. World Development, 105, 336-351.
Nian, H., Yu, C., Ding, J., Wu, H., Dupont, W. D., Brunwasser, S., ... & Wu, P. (2019).
Performance evaluation of propensity score methods for estimating average treatment
effects with multi-level treatments. Journal of applied statistics, 46(5), 853-873.
Nyamekye, E., & Dansoh, A. (2021). Cocoa farmers' knowledge and perception of hand
pollination and its effect on their practices and yield. A case study of Tafo cocoa district,
Eastern region-Ghana (Doctoral dissertation).
Obisesan, A. (2014). Gender differences in technology adoption and welfare impact among
Nigerian farming households. MPRA Paper No. 58920.
Orinda, M. A. (2013). Analysis of factors influencing sweet potato value. Analysis of factors
influencing sweet potato value addition amongst smallholders farmers in rachuonyo
South District, Kenya. Agriculture Journal, 3(2), 133-145.
Osei-Bagyina, A. (2012). Assessment of the impact of mining on the land use systems and

of
livelihoods in the Obuasi municipality (Doctoral dissertation).
Pandey, R., & Bardsley, D. K. (2019). An application of the Household Food Insecurity Access

ro
Scale to assess food security in rural communities of Nepal. Asia & the Pacific Policy
Studies, 6(2), 130-150.
-p
re
Rosenbaum, P. R. (2002). Covariance adjustment in randomized experiments and observational
studies. Statistical Science, 17(3), 286-327.
lP

Sánchez-Estrada, A., & Cuevas, J. (2020). Profitability of Artificial Pollination in


na

‘Manzanillo’Olive Orchards. Agronomy, 10(5), 652.


Shahbandeh, M. (2019). Cocoa production worldwide from 1980/81 to 2018/19 (in 1,000 tons)
ur

Retrived from https://www.statista.com/statistics/262620/global-cocoa-production/


Jo

Shahbandeh, M. (2021). Cocoa bean production worldwide 2018/19 & 2020/21, by country
Retrieved from https://www.statista.com/statistics/263855/cocoa-bean-production-
worldwide-by-region/
Sianesi, B. (2004). An evaluation of the Swedish system of active labor market programs in the
1990s. Review of Economics and Statistics, 86.133–55.
Soria, S. D. J., Wirth, W. W., & Chapman, R. K. (1980). Insect pollination of cacao in Costa
Rica, 1: Preliminary list of the ceratopogonid midges collected from flowers. Revista
Theobroma (Brasil) v. 10 (2) p. 61-68.
Suresh, K. P., & Chandrashekara, S. (2012). Sample size estimation and power analysis for
clinical research studies. Journal of human reproductive sciences, 5(1), 7.
Tedla, T. H. (2011). Factors determining fertilizer adoption of the peasant farm sector in
northern Ethiopia, Tigray region (Master's thesis).
Toledo-Hernández, M., Tscharntke, T., Tjoa, A., Anshary, A., Cyio, B., & Wanger, T. C. (2020).
Hand pollination, not pesticides or fertilizers, increases cocoa yields and farmer income.
Agriculture, Ecosystems & Environment, 304, 107160.
Tolemariam, A. (2010). Impact assessment of input and output market development interventions
by IPMS Project: The case of Gomma Woreda (Doctoral dissertation, Haramaya
University).
Vera-Chang, J., Cabrera-Verdezoto, R., Morán-Morán, J., Neira-Rengifo, K., Haz-Burgos, R.,
Vera-Barahona, J., ... & Cabrera-Verdesoto, C. (2016). Evaluation of three methods of
artificial pollination in clones of cocoa (Theobroma cacao L.) CCN-51. IDESIA, 34(6),

of
35-40.
Voora, V., Bermúdez, S., & Larrea, C. (2019). Global Market Report: Cocoa. International

ro
Institute for Sustainable Development.

-p
Walls, H., Baker, P., Chirwa, E., & Hawkins, B. (2019). Food security, food safety & healthy
re
nutrition: are they compatible?. Global Food Security, 21, 69-71.
Wessel, M., & Quist-Wessel, P. F. (2015). Cocoa production in West Africa, a review and
lP

analysis of recent developments. NJAS-Wageningen Journal of Life Sciences, 74, 1-7.


na

World Cocoa Foundation (2014). World cocoa update. Retrieved from WCF
http://worldcocoafoundation.org/wp-content/uploads/Cocoa-Market-Update-as-of-4-1-
ur

2014.pdf[Accessed: 2021, August 06]


Jo

Wu, H., Ding, S., Pandey, S., & Tao, D. (2010). Assessing the impact of agricultural technology
adoption on farmers' well‐ being using propensity‐ score matching analysis in rural
China. Asian Economic Journal, 24(2), 141-160.
Zeweld, W., Huylenbroeck, G. V., Hidgot, A., Chandrakanth, M. G., & Speelman, S. (2015).
Adoption of small‐ scale irrigation and its livelihood impacts in Northern Ethiopia.
Irrigation and Drainage, 64(5), 655-668.
Zheng, H., Ma, W., & Li, G. (2021). Adoption of organic soil amendments and its impact on
farm performance: evidence from wheat farmers in China. Australian Journal of
Agricultural and Resource Economics, 65(2), 367-390.
Jo
ur
na
lP
re
-p
ro
of
Tables

Table 1: Descriptive Statistics (N=206)


Total Adopters Non-Adopters
t-
Mean SD Min Max Mean SD Min Max Mean SD Min Max statistics
Household Characteristics
Age 39.4 11.42 16 75 36.19 7.03 24 57 42.6 13.85 16 75 4.007***
Actual number of years of formal
education 9.71 3.86 1 20 9.64 3.60 1 17 9.80 4.15 1 20 0.062

f
oo
Household size 11.15 4.66 1 25 11.93 3.76 4 21 10.36 5.32 1 25 2.209*
Number of Adult members in the

r
household (>18 years) 6.47 3.23 0 16 7.36 3.04 3 16 5.57 3.18 0 15 3.759***

-p
Number of Children in the household

re
(<18 years) 5.00 2.69 0 15 4.16 2.000 0 12 4.84 3.24 0 15 0.77
How many years have you been

lP
cultivating cocoa 15.5 10.7 3 50 12.53 7.95 3 40 18.46 12.22 4 50 4.038***
Farm Characteristics

na
Agricultural Land Size 10.62 6.57 3 40 8.96 3.2 3 27 12.28 8.43 3 40 3.508***
Cocoa Land Size 7.97 6.23
ur 2 40 6.85 2.98 3 24 9.09 8.17 2 40 2.384**
Number of Parcels of Cocoa Farm 1.91 0.74 1 3 1.98 0.79 1 3 1.84 0.69 1 3 1.061
Number of Bags Harvest (Major Season) 25.11 13.09 2 84 30.27 13.89 12 84 19.94 9.87 2 50 6.028***
Jo
Number of Bags Harvest (Minor
Season) 9.28 4.96 2 28 11.5 5.13 4 28 7.06 3.62 2 21 7.127***
Non-Farming Activity Income 406.08 257.12 10 1200 395.83 246.4 10 1200 416.33 267.84 10 1100 2.802***
Institutional Factors
Distance to nearest market (km) 4.45 5.05 1 18 5.85 4.71 1 14 2.9 4.98 1 18 9.871***

Source; Survey Estimation, 2019 (+ P<0.1; *P<0.05; **P<0.01; ***P<0.001)


Table 2: Probit Regression Results

Adoption Coef. Std. Err. dy/dx


Gen -0.008 0.279 -0.0012
Age -0.074 0.019 -0.011***
MStat 0.73 0.671 0.112
FBO 0.215 0.455 0.033
Ext 1.119 0.514 0.172*
Experience 0.657 0.173 0.101***
Distance -0.065 0.117 -0.01
Offinc 1.8 0.455 0.277***
Yield -0.661 0.351 -0.102+
FSize -0.172 0.392 -0.027
HHS 0.897 0.382 0.138*
Edu -0.029 0.196 -0.005

of
Constant -5.622 1.792
Sigma

ro
Number of obs = 206
LR chi2(12) = 171.4

-p
Prob > chi2 = 0.000
Pseudo R2 = 0.6
re
Log likelihood = -57.047648
Source; Survey Estimation, 2019(+ P<0.1; *P<0.05; **P<0.01; ***P<0.001) while dy/dx denote
lP

marginal effect
na

Table 3: Comparison of Productivity, Income, Poverty and Food Security


ur
Jo

Non-Adopters Adopters
t-
Mean SD Min Max Mean SD Min Max statistic

Productivity 30.19 14.14 7 84 41.90 38.26 2 200 2.961***

Income 13760.89 8196.66 0 38000 17216.19 16680.53 0 76000 2.531***

Poverty 9.21 4.23 5 20 14.01 5.70 5 20 7.005***


Food
Security 0.65 0.48 0 1 0.92 0.27 0 1 4.491***
Source; Survey Estimation, 2019(+ P<0.1; *P<0.05; **P<0.01; ***P<0.001)
Table 4: Balancing Test Results on Nearest Neighbor Matching Method

Before Matching (N=206) After Matching (N=206)


t-test t-test
Covariate Treated Control Statistics Treated Control Statistics
Gen 0.56 0.58 -7.44* 0.61 0.83 3.27
Age 35.83 42.64 47.70* 38.11 41.83 48.16
MStat 0.98 0.77 1.89 1.00 1.00 17.84
FBO 0.87 0.40 -5.90* 0.72 0.56 5.08
Ext 0.98 0.58 -1.83 0.89 0.89 10.89
Dist 1.46 0.56 2.25* 0.77 0.83 7.22
Offinc 5.69 5.62 63.71* 5.04 5.50 68.60
Yield 3.65 3.17 69.52* 3.20 3.34 79.50

of
FSize 1.84 1.94 26.57* 1.66 1.84 35.82
Exp 2.34 2.70 37.34* 2.58 2.92 45.48

ro
HHS 2.43 2.16 37.40* 2.51 2.39 46.86

-p
Edu 1.99 1.74 16.81* 1.63 1.68 22.82
Source; Survey Estimation, 2019(+ P<0.1; *P<0.05; **P<0.01; ***P<0.001)
re
lP

Table 5: Overall Balance Test Results


na

Overall Balance test (Hansen & Bowers, 2010)


ur

chi-square df p-value
Jo

Overall 6.911 11 0.806


Source; Survey Estimation, 2019(+ P<0.1; *P<0.05; **P<0.01; ***P<0.001)

Table 6: Distribution of Estimated Propensity Score

Variable Obs Mean Std. Dev. Min Max


Non-Adopters 105 0.13 0.2051 0.0000 0.7555
Adopters 101 0.87 0.2383 0.0375 1.0000
Total 206 0.4961 0.4331 0.0000 1.0000
Source; Survey Estimation, 2019
Table 7: Propensity Score Matching Results

Treated Control ATT Std. Err. t


Income
NNM 101 105 7133.90 1728.45 4.13*
RM 101 105 8224.30 3713.87 2.21*
Kernel 101 43 2792.96 1472.8 1.90*
Productivity
NNM 101 105 1.93 7.701633 0.25
RM 101 105 15.34 7.723635 1.99*
Kernel 118 88 5.00 5.471 0.914*

of
Poverty

ro
NNM 101 105 3.53 1.054285 3.35*
RM 101 105 0.83 0.558637 1.48
Kernel 84 122
-p 1.18 0.965 -1.218*
re
Food Security
NNM 101 105 0.02 0.028608 -0.85
lP

RM 101 105 0.03 0.03702 -0.79


Kernel 84 122 0.01 0.039 0.32
na

Source; Survey Estimation, 2019(+ P<0.1; *P<0.05; **P<0.01; ***P<0.001)


ur
Jo
Table 8: Sensitivity Analysis Results

Upper bounds on the significance level for different values of


𝑒𝑦 = 1 𝑒 𝑦 = 1.25 𝑒 𝑦 = 1.50 𝑒 𝑦 = 1.75 𝑒 𝑦 = 2.00
Using the single closest neighbor
Impact of artificial
pollination on
productivity 1.05878 0.776974 0.438807 0.776974 0.837229
0.000004 0.000523 0.033501 0.000523 0.000218
Using all neighbors within a caliper of 0.25
Impact of artificial
pollination on
productivity 1.10306 1.38083 1.72593 1.38083 1.32448

of
0.0000017 2.4E-09 7.90E-14 2.40E-09 1.00E-08
Using a kernel function and a smoothing parameter of 0.06

ro
Impact of artificial
pollination on
productivity 4.66461
0.0000015
-p
4.14533
0.000017
3.73694
0.000093
4.33947
0.0000071
3.81127
0.000069
re
Source; Survey Estimation, 2019
lP
na
ur
Jo
Figure

1a: NNM: Productivity 2a: NNM: Poverty

of
0 .2 .4 .6 .8 1 0 .2 .4 .6 .8 1
Propensity Score Propensity Score

ro
Untreated Treated Untreated Treated

1b: NNM: Food Security 2b: NNM: Income

-p
re
lP
na
ur

0 .2 .4 .6 .8 1
Propensity Score
0 .2 .4 .6 .8 1
Untreated Treated Propensity Score
Jo

Untreated Treated

Figure 1: Histogram of propensity score matching for Artificial Pollination


Adoption

Appendix

Appendix 1.0: Educational Level


Total Adopters Non-Adopters
Education Frequency Percentage Frequency Percentage Frequency Percentage
No Formal 32 16% 12 12% 20 19%
Primary 110 53% 63 61% 47 46%
JHS 48 23% 20 19% 28 27%
Secondary 16 8% 8 8% 8 8%
Tertiary 0 0% 0 0% 0 0%
Total 206 103 103

Source; Survey Estimation, 2019

of
Appendix 2.0: Occurrence questions used in the computation of HFIAS score

ro
Occurrence questions Response
S/N

1. In the past four weeks, did you worry -p 0 = No


re
that your household would not 1 = Rarely (once or twice in the past four weeks.
lP

have enough food? 2 = Sometimes (three to ten times in the past four
weeks
na

3 = Often (more than ten times in the past four


ur

weeks).
Jo

2. In the past four weeks, were you or 0 = No


any household member not 1 = Rarely (once or twice in the past four weeks.
able to eat the kinds of foods 2 = Sometimes (three to ten times in the past four
you preferred because of a lack weeks
of resources? 3 = Often (more than ten times in the past four
weeks).

3. In the past four weeks, did you or any 0 = No


household member have to eat 1 = Rarely (once or twice in the past four weeks.
a limited variety of foods due 2 = Sometimes (three to ten times in the past four
to a lack of resources? weeks
3 = Often (more than ten times in the past four
weeks).

4. In the past four weeks, did you or any 0 = No


household member have to eat 1 = Rarely (once or twice in the past four weeks.
some foods that you really did 2 = Sometimes (three to ten times in the past four
not want to eat because of a weeks

of
lack of resources to obtain 3 = Often (more than ten times in the past four

ro
other types of food? weeks).

5. In the past four weeks, did you or any-p 0 = No


re
household member have to eat 1 = Rarely (once or twice in the past four weeks.
lP

a smaller meal than you felt 2 = Sometimes (three to ten times in the past four
na

you needed because there was weeks


not enough food? 3 = Often (more than ten times in the past four
ur

weeks).
Jo

6. In the past four weeks, did you or any 0 = No


household member have to eat 1 = Rarely (once or twice in the past four weeks.
fewer meals in a day because 2 = Sometimes (three to ten times in the past four
there was not enough food? weeks
3 = Often (more than ten times in the past four
weeks).

7. In the past four weeks, was there ever 0 = No


no food to eat of any kind in 1 = Rarely (once or twice in the past four weeks.
your household because of lack 2 = Sometimes (three to ten times in the past four
of resources to get food? weeks
3 = Often (more than ten times in the past four
weeks).

8. In the past four weeks, did you or any 0 = No


household member go to sleep 1 = Rarely (once or twice in the past four weeks.
at night hungry because there 2 = Sometimes (three to ten times in the past four
was not enough food? weeks

of
3 = Often (more than ten times in the past four

ro
weeks).

9. In the past four weeks, did you or any -p0 = No


re
household member go a whole 1 = Rarely (once or twice in the past four weeks.
lP

day and night without eating 2 = Sometimes (three to ten times in the past four
na

anything because there was not weeks


enough food? 3 = Often (more than ten times in the past four
ur

weeks).
Jo

Source; (Coates et al., 2007)


 Adoption of artificial pollination is negatively influenced by age of farmer.

 Extension, Experience, Off farm income and Household size increase adoption.

 Adoption of artificial pollination increased smallholder productivity.

 Adoption of artificial pollination improves food security of farmers.

 Adoption of artificial pollination reduces poverty.

of
ro
-p
re
lP
na
ur
Jo
The authors declare that they have no conflict of interest

of
ro
-p
re
lP
na
ur
Jo

You might also like