Journal of Natural Sciences Research www.iiste.

org
ISSN 2224-3186 (Paper) ISSN 2225-0921 (Online)
Vol.6, No.16, 2016

Production Frontier of Small Scale Pearl Millet Farmers: A

Comparison of Conservation and Traditional Agricultural
Practices in the Northern Namibia

Mr. Bonolo, P. Montle

Post graduate (Honours) student, Department of Agriculture & Natural Resources Sciences, Namibian
University of Science and Technology
Tel. +264816082873, E-mail: [email protected]

Dr. David, I. Uchezuba (Corresponding author)

Namibian University of Science and Technology, Department of Agriculture & Natural Resource Sciences, 13,
Storch Street, Windhoek, Namibia. Tel +264810425716,
E-mail: [email protected]

Mr. Salomo Mbai.

Namibian University of science and Technology, Department of Agriculture & Natural Resource Sciences, 13,
Storch Street, Windhoek, Namibia.Tel. +264813902376, E-mail: [email protected]

Abstract

This study fits a stochastic Cobb-Douglas production frontier of the pearl millet smallholder farmers and
examined their technical efficiency comparing Conservation and Traditional Agriculture practices. The data was
collected using a structured questionnaire administrated to 100 randomly selected smallholder farmers in
Omusati, Ohangwena, Oshikoto, Oshana and Kavango regions during the 2014-2015 planting season. The
estimated parameter of the model shows that land availability, the level of fertilizer use and tractor power
explains variations in the production of pearl millet. The efficiency analysis result shows there is no statistically
significant difference in the technical efficiency of farmers who were exposed to conservation agriculture
compared to their traditional method of agriculture. The inefficiency model indicates that farm experience, farm
size, and farm training have significant positive effect on efficiency. In addition, the study examined farmers
willingness to pay for extension services, the predicted probability of getting farmers who are willing to pay is
60%. Some socio-economic factors such as farm size, herd size, and membership of a cooperative were found to
influence farmers’ willingness to pay for extension service. The study recommends that Conservation
Agriculture should be continued over a long period of time so that the impact can be felt. Capacity building,
training, extension services, information on agronomic practices and farmer’s education are factors that policy
should address.
Keywords: technical efficiency, Conservation Agriculture, pearl millet, stochastic production frontier.

1 Introduction and background

The Namibian population at large has a large number of people that are rural dwellers. About 70 percent of these
rural dwellers depend on agriculture for sustenance (UNEP, 2012). The agriculture sector is a major contributor
to employment. It employs about twenty-seven percent of the country’s workforce and fifty-eight per cent of the
workforce live in the rural areas (UNEP, 2012). Nevertheless, there has been a declining trend in agricultural
productivity and often the practice is seldom sustainable. In 2013, a report by the World Food Programme on
food security indicated that crop production in Namibia was threatened by continued drought. Consequently, a
significant drop in cereal output was recorded in 2015. Maize production declined by 73% from the above-
average yield in 2014. Production of sorghum and millet also decreased by 60 and 65 percent respectively (Food
and Agriculture Organisation (FAO), 2015). This resulted in food shortages, leading to about 30% of households
adopting survival strategy of reducing the number of their food ration to one meal per day (FAO, 2015). This led
to reduced availability and reduction in dietary diversity to about 46% among the households (Emergency Food
Assessment in Communal and Resettlement Areas of Namibia (EFA), 2013). As a result, an estimated number of
330 925 people were found to be food insecure (EFA, 2013).

76
Journal of Natural Sciences Research www.iiste.org
ISSN 2224-3186 (Paper) ISSN 2225-0921 (Online)
Vol.6, No.16, 2016

Due to the declining trend in food production, attention is now focussed on methods of improving agricultural
productivity. This is because the improvement of agricultural productivity is an important tool towards
increasing household food security and alleviating rural poverty (Owour, 2000). Despite the various past efforts,
food security continues to be a challenge in Namibia as is the situation in a number of Sub-Saharan African
countries. This is so because of low and stagnant agricultural productivity growth associated with major crops
like pearl millet which are predominantly produced by smallholder farmers under rain-fed conditions (United
Nation Partner Framework (UNPAF), 2014). More effort is required for total eradication of poverty and food
insecurity. In this regard, the key area of interventions is to enhance efficiency and productivity growth, make
land available to the poor through land reform and maximize the potential of available land through the use of
improved soil fertility, and the adoption of technical innovations that enhance technical change (UNPAF, 2014).
These landmarks are currently not achieved under the Traditional Agricultural (TA) system where few available
lands are marginally utilized, resulting in soil degradation and a decline in productivity. In the wake of this
situation, there is a need for a re-thinking about the best way of utilizing land potentially. As a result, an
agricultural practice (conservation agriculture) that inculcates the principle of conservation has recently been
introduced to the farmers in some parts of the Northern Namibia. This type of agricultural practice aims at
improving efficiency and productivity by preventing loss or damage to the soil and soil components, thereby
enhancing the preservation and careful management of the environment and of natural resources, such as land.
As a pilot project, the conservation agricultural practice was introduced to the farmers (by a non-governmental
organization NGO) who applied it alongside their Traditional Agricultural (TA) practice, that is, farmers were
selected to utilize the two methods simultaneously.

Therefore, the aim of this study is to compare the production frontier of the two practices to determine the
response of agricultural output to the inputs under both practices and in addition, measure the farmers’ efficiency
of production. The measurement of farm efficiency is an important factor for productivity growth and it will be a
viable option in the developing countries such as Namibia where resources are scarce (Kibaara, 2005). As
mentioned previously, the administration of this pilot project is facilitated by an NGO whose services are
currently offered without cost to the farmers. However, it is envisaged that in the future, the farmers might be
required to pay for the services. Therefore, the study also assesses the farmers’ willingness to Pay (WTP) for
such services in the future. In agriculture, WTP studies have been used to evaluate demand and cost curves for
extension services delivery through commercial agents (Oladele, 2008). A similar approach will be adopted in
this study because the extent to which the farmers are willing to pay for extension services has not been
conducted in this field of study in Namibia. The study is important because the outcome will aid policy makers
and stakeholders towards the generation of a pool of knowledge about farm practice that is optimal and can
maximize land use. This will form important anecdote to the mandate of the national development plans
(Mushunje, 2005).

2 Materials and Methods

2.1 Data and Sampling

The study was conducted in five administrative regions of Northen Namibia namely; Oshana, Omusati,
Oshikoto, Ohangwena, and Kavango region. The selected regions are well suited for this study because their
inhabitants derive their sustenance mainly from crop production especially pearl millet which is their major
staple. Also, information regarding conservation agricultural (CA) practices in these regions is not known and
has not been investigated before. Primary data was collected in a farm-level survey using structured
questionnaire administered to 100 randomly selected household heads in the study area during the 2014-2015
cropping season. The data consist of production information, farmer’s support, and farmers’ socio – economic
characteristics.

2.2 Empirical Model

2.2.1 Technical efficiency
According to Alene and Hassan (2003), production frontier is specified to represent the maximum output from a
given set of inputs and existing production technology. Failure to attain the frontier output implies the existence
of technical inefficiency. Both production frontier and technical inefficiency models were fit. The frontier model
specification was developed by Battese and Coelli (1995) who proposed a stochastic frontier model where
technical inefficiencies can be expressed as a function of explanatory variables and a random error (Battese and
Coelli, 1995). The Cobb-Douglas stochastic frontier production function was specified as follows:

Ln Yi = β0i + β1 LnX1i + β2 Ln + β LnX3i + β4 LnX4 +Vi +Ui ………………………………… (1)

77
Journal of Natural Sciences Research www.iiste.org
ISSN 2224-3186 (Paper) ISSN 2225-0921 (Online)
Vol.6, No.16, 2016

Where:
Output (Y) is yield of pearl millet in kg;
X1 is planted land size (ha)
X2 is the fertilizer application (kg)
X3 is the seed quantity (kg)
X4 is the total labour (man days)
X5 is tractor power

The technical inefficiency model was used to identify factors that influence the efficiency among pearl millet
farmers in the study area, and the model was estimated as follows;

= + + + + + + + ……………………… (2)

Where:
Z1= farm experience; Z2 = education level; Z3= extension services; Z4 = off farm income; Z5 = farm size; Z6 =
Cooperative membership, Z7= household income less than 2000, Z8= household size, Z9 = household status, Z10
= farm training, Z11= farm credit (loan).

2.2.2 Willingness to pay for extension services

Probit regression model was fit to assess the factors that may be associated with willingness to pay (WTP) for
the extension services. The independent variables are, farm experience, age, education level, extension services,
off-farm income, farm size, cooperative membership, household income, and household size. The dependent
variable WTP is a dichotomous variable which takes the value of one if a farmer is willing to pay otherwise,
zero. The main issue is to identify the individual household characteristics that influenced the farmers’ WTP.
Following Nagler (2002), the probit model takes the form:

Pr =1 = + ………………………………………………………………… (3)
Where, Y is a dichotomous dependent variable which can assume the value of 0 or 1. It measured the farmer’s
willingness to pay for extension services. Xi = n x k matrix of explanatory variables (farm experience, age,
education level, extension services, off farm income, farm size, cooperative membership, household income,
household status). =k x 1 vector of parameters to be estimated and is the error term.

3 Results and Discussion

3.1 Descriptive statistics
The descriptive statistics of the variables included in the models are presented in Table 1. The existence of
multicollinearity was tested which was detected as some variables correlated with others. The autocorrelation
was corrected by allowing for correlated errors in the variance-covariance matrix (VCE) estimation.

78
Journal of Natural Sciences Research www.iiste.org
ISSN 2224-3186 (Paper) ISSN 2225-0921 (Online)
Vol.6, No.16, 2016

Table 1: Summary descriptive statistics

Variable Mean Standard. Deviation Min Max
Wtp520 0.6 0.4923 0 1
Farmexp 31.07 12.4634 2 61
Age 61.16 13.3088 22 100
Edu 0.76 0.4292 0 1
Extserv 0.44 0.4988 0 1
NFI 0.66 0.4760 0 1
Farmsize 6.464 4.4041 1 25
HHsize 9.46 5.2308 1 28
Memcop 0.43 0.4975 0 1
Income < 2000 0.66 0.4760 0 1
HHstatus 0.76 0.4292 0 1
From the results, it can be deduced that the average farmer is more than sixty years old. The average farming
experience is about 31 years, which means that most of the farmers have farmed for the most part of their life.
This is an indication that pearl millet production has been in existence for a number of years as the majority of
the small-scale farmers have been in pearl millet production for more than 31 years. The age of the farmer is an
important factor of production as older people tend to be too stereotyped always sticking to what they are used to
(that is, they prefer to use old methods of planting than adopt new technology). It is assumed that older farmers
are more experienced in farming activities and are in a better position to assess the risks involved in farming than
younger farmers. The household size plays an important role in pearl millet production because most farmers
depend on family labour. The data shows that the average household size is 9. Invariably, the large family size
means that more labour will be available for farming. The average farm size for the farmers is 6.4 hectares. This
indicates that the land capacity of the sampled farmers is small, however; the aim of the study is to determine
how efficient they are given their farm size and the introduction of conservation agriculture.

3.2 Maximum likelihood estimates of production frontier parameters

Table 2 presents the results of the Cobb-Douglas production function for both methods (Conservation and
traditional agriculture). In the case of conservation agriculture, the estimated production function parameters
indicated that the area planted, fertiliser and the use of animal plough significantly affect the pearl millet yield.
Labour and seed were found to be insignificant. In traditional agriculture, fertiliser and the use of animal plough
were shown to significantly affect yield while seed, area planted and labour was insignificant. These results are
consistent with findings by Baloyi (2011), Musaba and Bwacha (2014) in a study carried out in South Africa and
Zambia respectively. The insignificant effect of seed quantity could imply that farmers in the study area do not
use improved seeds hence; they resort to the indigenous varieties and the use of inappropriate seed rate. The
insignificant effect of labour with a negative sign is consistent with Musaba et al (2014), this could imply that
there is abundant household labour in the study area and any increase in this labour will lead to diminishing
returns. The positive and significant coefficient for fertiliser indicates that pearl millet output increases with a
unit increase in fertiliser input. The higher significance level under CA implies that farmers were advised well on
the application of both basal and top dressing and application was also done in rows in close accessibility to a
plant. While on TA, farmers mostly applied basal dressing fertilisers through the broadcasting method which
makes this result acceptable. This is in agreement with a study by Tchale and Sauer (2007) which emphasised
the essence of obtaining efficiency in more fertile areas. For CA, the area planted was significant at 5% while on
TA it was insignificant. The result shows that access to land is important in explaining the differences in yield of
the farmers. This finding is supported by Kimhi (2003) who found a positive relationship between maize yield
and plot size, indicating that economies of scale are dominant throughout the plot size distribution.
The estimated coefficients for production frontier parameters have the expected positive signs except labour,
suggesting that labour had less influence in the production of millet. This is consistent with the study by
Ogundele & Okoruwa (2006). This scenario is expected as the level of pearl millet production depends largely
on a number of factors including climate and not only limited to production inputs. However, all other variables
for both technologies were positive, implying that the pearl millet in the study area is positively influenced by
these factors. The seed quantity being insignificant is quite unexpected but given that farmers in the study area
do not know the correct amount of seeds to be used in a unit area, the result is acceptable. With regards to land
preparation, both in TA and CA the variable was significant but the sign was not as expected.

79
Journal of Natural Sciences Research www.iiste.org
ISSN 2224-3186 (Paper) ISSN 2225-0921 (Online)
Vol.6, No.16, 2016

Table 2: Maximum Likelihood Estimation of the Cobb-Douglas stochastic production frontier for the
comparison of Conservation Agriculture vs Traditional Agriculture
Conservation Agriculture Traditional Agriculture
(CA) (TA)
Variable Coefficients Coefficients
-0.0458 -0.1279
Total labour
(0.5180) (0.5290)
0.1846** 0.1171
Land
(0.0570) (0.2620)
0.0592*** -0.0551**
Fertilizer
(0.0080) (0.0120)
0.0410 0.0253
Seed
(0.4020) (0.6250)
-0.2229** -0.2309**
Tractor power
(0.0140) (-0.0260)
6.5761*** 0.1228
Constant
(0.000) (0.6660)
-3.8945 -3.6696
Lnσ v2 (0.0000) (0.0000)
2 -0.7741 -0.7922
Lnσ u (0.0000) (-0.00060)
0.1427 0.1596
σv (-0.0370) (-0.0662)
0.6791 0.6729
σu (-0.0810) (-0.0973)
Wald Test: Joint Significance:
Wald Chi-square(11) 20.030 15.650
Prob > Chi-square 0.001 0.008
LR Test: Sigma_u = 0 29.46 7.02
(0.0000) (0.004)
Note: Figures in parenthesis are the p-values. The notation ***, **, and * denote statistical significance at the
1%, 5% and 10% level of significance.

3.3 Constant returns to scale (CRS)

For constant return to scale, the sum of the technical coefficients β must be equal to one, for an increasing return
to scale, it should be greater than one, and for decreasing return to scale it should be less than one. For the case
of this study, the constant return to scale hypothesis was rejected indicating that the sum of the technical
coefficients is not equal to one. This is an indication that the hypothesis is not supported by the data. Reasons
attributed to this are that of a shorter period of trials to examine the impact of CA, and that the labour input is not
efficiently utilised by the farmers due to the communal nature of their farming. Therefore, the output cannot
double if inputs levels are doubled.

3.4 Hypothesis testing

The standard deviation of the two error components σv and σ u , and their log likelihood estimates Lnσ v2 and
2
Ln for the two models are respectively given in the post estimation Table 2. The result shows that they are all
u
statistically significant. To test the presence of technically inefficiency effects, the log-likelihood-ratio (LR) test
was adopted. The null hypothesis of no technical inefficiency effects in pearl millet production was strongly
rejected indicating that the production frontiers of the farmers are characterized by technical inefficiency effects
(Table 2). The Wald test statistics for joint significance of the variables in the two models are 20.03 and 15.65.
The null hypothesis of joint zero coefficients was rejected. Based on the result of the two tests, it can be
concluded that the included variable contribute to explain production frontier and that technical inefficiency
effects are present in the model thus confirming the need to fit inefficiency model.

80
Journal of Natural Sciences Research www.iiste.org
ISSN 2224-3186 (Paper) ISSN 2225-0921 (Online)
Vol.6, No.16, 2016

3.5 Technical inefficiency scores

This section discusses the technical efficiency estimates obtained from the stochastic frontier model. Table 3
presents summary statistics of the technical inefficiency scores based on two farming methods as well as by
region.

The mean technical efficiency of 32% under conservation agriculture indicates that on average the respondents
are able to obtain over 30% of potential output from a given mix of production inputs. The result is consistent
with Diiro (2013) and Kibaara (2005). This implies that, in the longer term, there is a potential for pearl millet
producers to increase their efficiency by about 68% by utilising existing farm resources better and following the
appropriate principles of conservation agriculture so as to be on the optimal production frontier. While under
TA, the mean technical efficiency of 33% indicate that on average there is a potential for pearl millet producers
to increase their efficiency by about 67% utilising existing farm resources better and adopting improved
technology and techniques. The regional analysis (Table 4) shows that Kavango region was the most efficient
region in both technologies as evidenced by the farmers’ inefficiency scores. The most inefficient were Oshikoto
and Ohangwena regions for both methods of farming. The differences in efficiency levels between regions could
be attributed to factors such as climate, soil fertility, availability of planting materials such as seeds, poverty
prevalence, management and socio-economic factors. The intra-region differences between CA and TA are also
very small an indication that the effects of the CA cannot be realized in the short-run. Generally, the result
indicates that there is a need to practice CA over a longer period of time in order to observe the desired impact.

Table 3: The summary statistics of Technical Inefficiency scores of sampled farmers

Farming method Obs Mean Std. Dev Min Max

CA 100 0.6791 0.6285 0.0619 2.8432

TA 100 0.6729 0.5547 0.0641 2.7142

Table 4: Mean technical inefficiency scores by regions

Oshikoto Omusati Oshana Kavango Ohangwena
CA 1.2171 0.5741 0.6439 0.2563 0.7038
TA 0.8824 0.5943 0.7511 0.3234 0.8136

3.6 Determinants of technical efficiency

The analysis of the estimated coefficients of the inefficiency model explains the contribution of the variables to
technical efficiency in the study area. Using the SFA model, the sources of inefficiency were examined using the
identified determinants of inefficiency effects. Table 5 shows the results for both conservation and traditional
agriculture technical inefficiency model. The negative sign on the estimated parameters in the technical
inefficiency model implies that the associated variable reduces inefficiency, meaning it has a positive effect on
technical efficiency; as a result, it increases productivity level. A positive sign indicates that the associated
variables increase inefficiency or have a negative effect on technical efficiency.

In the case of conservation agriculture, household size, cooperative membership, income>2000, farming
experience are significant at 5% and 10%. Although these variables are significant, they have positive signs
except farming experience which implies that they have a negative effect on efficiency. This finding is consistent
with studies carried out by Kibaara (2005), Mango, Makate, Lundy (2015) and Diiro (2013). The coefficient sign
for farming experience is negative and significant; suggesting that this variable reduces technical inefficiency.
This further entail that experienced farmers tend to be more efficient because of good managerial skills which
they have learnt over time, and more efficient than younger ones. This result is supported by Khairo and Battese
(2005) who found that the farming experience coefficient was negative and significant which means that farmers
tend to decrease their technical inefficiencies as they become more experienced. The estimated positive
coefficient of the household size which is significance at 5% implies that smaller families are efficient compared
to larger ones because large family size exerts pressure on the limited resources a farmer has (Mango et al,
2015).

The insignificant level for extension service and training is not as expected, however; this could be attributed to
slow rate of adoption and understanding of the intervention by first time participants’ farmers (Mkhabela, 2005).
The positive coefficient sign for Income<2000 indicate that farmers who have a household income less than
2000 are inefficient compared to the ones earning more than this amount. This is because farmers with

81
Journal of Natural Sciences Research www.iiste.org
ISSN 2224-3186 (Paper) ISSN 2225-0921 (Online)
Vol.6, No.16, 2016

income<2000 will not have the financial resources to purchase necessary inputs for farming that may increase
their technical efficiency; this result is consistent with Oladimeji and Abdulsalam (2013).

Table 5: Determinants of technical efficiency for Conservation Agriculure vs Traditional Agriculture

Conservation Agriculture Traditional Agriculture
(CA) (TA)
Variable Coefficients Coefficients
Farmexp -0.0105* -0.0122***
(0.0952) (0.0180)
Edu -0.0845 0.0437
(0.623) (0.7730)
Extserv -0.0029 0.1071
(0.623) (0.3390)
Nfi -0.1512 -0.1537
(0.381) (0.2440)
Farmsize -0.0128 -0.0210*
(0.29) (0.0700)
Hhsize 0.0319** 0.0263**
(0.026) (0.0230)
Memcop 0.2454* 0.1784
(0.072) (0.1470)
Income<2000 0.3508*** 0.2207
(0.016) (0.1430)
Hhstatus 0.1802 0.1573
(0.246) (0.2140)
Training -0.1789 -0.3505***
(0.225) (0.0140)
Loan 0.1783 0.3828**
(0.357) (0.0230)
Constant 0.4859 0.7711
(0.266) (0.0510)
F (12, 87) 5.88 6.84
Prob > F 0.000 0.000
R - Squared 0.139 0.1942
Root MSE 0.6219 0.5312
Note: Figures in parenthesis are the p-values. The notation ***, **, and * denote statistical significance at the
1%, 5% and 10% level of significance.

The result for the cooperative membership shows that farmers who are non-cooperative members were more
efficient than the ones who are members. This result is not expected because; being a member of the cooperative
gives the farmers opportunities for training, information, collective bargaining power and credit sales. This could
be due to the fact that most of the farm cooperative does not offer these services hence; members do not have
much leverage over non-members.

In traditional agriculture, the estimated coefficient sign of the variables from the inefficiency model shows that
only farmers’ training, farm experience, and farm size are statistically significant with the correct signs. This
indicates that these variables have a positive influence on technical efficiency. With regards to training, the
finding implies that farmers who have access to training classes are more efficient than the ones who do not. The
training sessions farmers have received over time by the Ministry officials on production related information

82
Journal of Natural Sciences Research www.iiste.org
ISSN 2224-3186 (Paper) ISSN 2225-0921 (Online)
Vol.6, No.16, 2016

tend to increase their efficiency levels. This result is consistent with Mango et al (2015). Moreover, a negative
sign for farm size implies that farmers with large arable land tend to be more efficient than smallholder farms as
they are able to diversify their activities i.e. practice crop rotation and other integrated farming systems that
eventually increase their income and improve their efficiency. This same result was found by Mango et al
(2015). Although the household size and financial credit are significant, their coefficients are positive which
implies they have a negative influence on technical efficiency. Larger and poor households are more likely to
default in honouring their financial obligation, more credit, in this case, mean more liability; therefore, credit-
strapped farmers are more likely to be inefficient because increase credit gets them more entangled in debts.
Debt financing strips them of the available cash flow.

3.7 Factors influencing farmer’s willingness to pay (WTP) for extension services
To identify determinants of willingness to pay for agricultural extension services, probit model was estimated.
The result shows that farm size, household income (Income < 2000), cooperative membership and household
size have a significant relationship with farmer’s willingness to pay. The predicted probability of getting farmers
who are willing to pay for extension services is 60 %, (Table 6). The probability is high, an indication of the
likelihood that future CA practices will be adopted by the majority of the farmers.

Table 6: Marginal effects test

Delta method
Margin Std. Err z P>|z| [95% Conf. Interval]
_cons 0.6008 0.0312 19.21 0.000 0.5395 0.6621

Table 7: Maximum likelihood estimates of the probit model

Robust
wtp520 Coef. Std. Err z P>|z| [95% Confidence Interval]
Farmexp 0.0185 0.0255 0.72 0.469 -0.0315 0.0685
Age 0.0050 0.0283 0.18 0.858 -0.0504 0.0605
Edu 0.4847 0.4044 1.2 0.231 -0.3080 1.2774
Extserv -0.1653 0.3410 -0.48 0.628 -0.8337 0.5030
Nfi 0.5880 0.4284 1.37 0.17 -0.2516 1.4277
Farmsize -0.0929 0.0430 -2.16 0.031 -0.1773 -0.0084
Hhsize -0.1106 0.0378 -2.92 0.003 -0.1849 -0.0364
memcop -2.5604 0.3419 -7.49 0.000 -3.2307 -1.8901
Inc<2000 -2.0064 1.0036 -2.00 0.046 -3.9736 -0.0393
hhstatus 0.1146 0.4137 0.28 0.782 -0.6963 0.9255
_cons 3.0895 1.6845 1.83 0.067 -0.2121 6.3911

Farm size was found to be statistically significant at 5% level with the willingness to pay for an extension.
Negative sign implies that farmers with small farm size are more likely to pay than those with larger farms. This
could be attributed to the fact that the farmers are in a communal leasehold farming system with small land
capacity. If farm size increases they will pay more premium as the payment for extension is made per hectare.
Household size was significant at 1% level and has negative signs. The result is in agreement with the a priori
expected sign because if household size increases, the cost of living increases and the purchasing power of the
farmer decline. This finding is supported by Tolera et al (2014) who revealed that negative sign implies that
small size households were likely to pay more than larger households.

Cooperative membership was significant at 1% level. A negative sign implies that farmers who are non-
cooperative members are more likely to pay than unionized farmers. The finding is consistent with results by
Oladele (2008). Cooperative member who gets services and other benefits from cooperatives are less likely to
pay because they would rely on the cooperative assistance than pay additional money to other service providers.
Household income (< NAD 2000) is statistically significant but has negative signs. The result shows that
resource-poor farmers with income less than N$ 2000 would still not pay even if income increases beyond N$

83
Journal of Natural Sciences Research www.iiste.org
ISSN 2224-3186 (Paper) ISSN 2225-0921 (Online)
Vol.6, No.16, 2016

2000. This implies that payment for extension services does not depend on income but on the need to understand
the importance of the services rendered which most farmers find difficult to comprehend.

Table 8: Marginal effects of the covariates

Variable dy/dx Std.Err. z P>|z| [95% Confidence Interval]
Farmexp 0.0066 0.0091 0.72 0.469 -0.0113 0.0245
Age 0.0018 0.0101 0.18 0.858 -0.0180 0.0217
Edu 0.1737 0.1450 1.20 0.231 -0.1106 0.4580
Extserv -0.0592 0.1216 -0.49 0.626 -0.2976 0.1791
Nfi 0.2107 0.1525 1.38 0.167 -0.0882 0.5097
Farmsize -0.0333 0.0152 -2.18 0.029 -0.0632 -0.0033
Hhsize -0.0396 0.0131 -3.02 0.002 -0.0653 -0.0139
Memcop -0.9175 0.1323 -6.93 0.000 -1.1770 -0.6580
Inc>2000 -0.7190 0.3616 -1.99 0.047 -1.4278 -0.0102
Hhstatus 0.0410 0.1481 0.28 0.782 -0.2492 0.3314

Since the extension service will be based on cost per hectare serviced, farmers with larger farms will pay more
than smaller farms. The result shows that an increase in one unit of farm size will lead to a decrease in the
willingness to pay for extension services by 3%. Household size was also found to have similar effects. An
increase in household size by one person decreases the farmer’s willingness to pay by 4%. As the household
enlarges, the farmer will have a huge responsibility of catering for a large family and may not have extra
resources to pay for extension services. If a farmer belongs to a cooperative, the likelihood that they will pay is
lower by 91% compared to when they are not a member. This is because the farmer will expect the cooperative
to subsidise and cater for a larger percentage of their extension service needs. Farmers with an income less than
2000 are less likely to pay for extension services than those that have higher.

4. Conclusion and recommendations

The result of the production frontier under conservation agriculture shows that the area planted, tractor power
and fertiliser application significantly affect the pearl millet yield. On the other hand, fertilizer and tractor power
influenced production under traditional agriculture method. No statistically significant effect was observed for
seed and labour used for both methods. The insignificant labour coefficient can be attributed to low labour
productivity among low-skill workers in sub-Saharan Africa. Optimal labour use cannot be identified due to the
use of family labour thus; output may be affected if additional labour is employed. Similar effects can be
observed with regards to seed application. Non-optimal use of seed may result in negative effects on output. In
addition, the use of the wrong seed variety that has low germination rate may affect output. The results show that
there is a technical efficiency of 32% under conservation agriculture indicating that on average the respondents
are able to obtain over 30% of potential output from a given mix of production inputs. While under traditional
agriculture, the mean technical efficiency of 33% was calculated, indicating that on average there is a 67%
allowance for efficiency improvement by addressing important constraints that affect farmers’ levels of technical
efficiency and productivity in the study area. The results show that variables such as farm size, cooperative
membership, household size and household income were statistically significant but have unexpected signs.
This study recommends that conservation agriculture should be practised over a long period of time so that its
impact can be felt. Skill development should be considered by policy-makers or institutions with the
responsibility of designing programs towards pearl millet farming improvement in the study area. It is
recommended that farmers’ financial and social capacity be improved through the provision of hi-tech planting
materials such as fertilizers, seeds, and increased extension training.

References

Alene, A.D., and Hassan, R.M. 2003, “The determinants of farm-level technical Efficiency among adopters of
improved maize production technology in Western Ethiopia.” Agrekon, Vol 42, No 1 (March 2003).

Baloyi, R. 2011, “Technical Efficiency in Maize Production By Small Scale Farmers in South Africa: Case study
from Gamothiba, Limpopo Provience”.

84
Journal of Natural Sciences Research www.iiste.org
ISSN 2224-3186 (Paper) ISSN 2225-0921 (Online)
Vol.6, No.16, 2016

Battese, G. E., and Coelli T. J. 1995, “A Model for technical inefficiency effect in a stochastic frontier
production function for panel data”. Empirical Economics, 20: 325-32.

Diiro G.M. 2013, “Impact of Off-farm Income on Agricultural Technology Adoption Intensity and Productivity-
Evidence from Rural Maize Farmers in Uganda.” Uganda Strategic program-IFPRI: Working Paper 11 -January
2013.

Emergency Food Assessment in Communal and Resettlement Areas of Namibia (EFA). 2013, World Food
Programme on Food Security in Namibia.

FAO. 2015, “Global information and early warning of food agriculture. Country brief for Namibia Kimhi A.
(2003). Plot Size and maize productivity in Zambia; the inverse relationship re-examined. Discussion paper
No.10.13. Hebrew University of Jerusalem.

Kibaara, B.W. 2005, “Technical Efficiency in Kenyan’s Maize Production: An Application of the Stochastic
Frontier Approach”. Unpublished M.Sc. Department of Agricultural and Resource Economics, Colorado State
University, Fort Collins. Colorado.

Mango, N., Makate, C., Mlambo, B., Siziba, S., & Lundy, M. 2015, “A Stochastic Frontier Analysis of
Technical Efficiency in Smallholder Maize production in Zimbabwe: The post-fast-track land reform outlook.

Mkhabela, T. 2005, “Technical efficiency in a vegetable based mixed-cropping sector in Tugela Ferry, Msinga
District, KwaZulu-Natal”. Agrekon, 44, (2):187-200.

Musaba E and Bwacha, I. 2014, “Technical Efficiency of Small Scale Maize Production in Masaiti District,
Zambia”: A Stochastic Frontier Approach: Journal of Economics and Sustainable Development: Vol.5, No.4,
2014. ISSN 2222-1700 (Paper) ISSN 2222-2855 (Online).

Mushunje A. 2005, “Farm efficiency and land reform in Zimbabwe.” PhD thesis: Department of agricultural
economics and extension, University of fort hare, Alice, South Africa.

Nagler, J. 2002, “Interpreting probit analysis”. New York University.

Webpage: www.nyu.edu/classes/nagler/quant1/probit1_post.pdf. Accessed 18 March, 2016.

Oladele, O.I. 2008, “Factors determining farmers willingness to pay for extension services in
Oyo State, Nigeria”. Agricultural Tropica Et Subtropica 41: (4)165- 169

Oladimeji, Y.U. and Abdulsalam, Z. 2013, “Analysis of Technical Efficiency and its Determinants among Small
Scale Rice Farmers in Patigi LGA of Kwara State, Nigeria”. IQSR Journal of Agriculture and Veterinary
Science. 3(3) Pp 34-39.

Ogundele O. O and Okoruwa, O.V. 2006, “Technical efficiency differential in rice production technologies in
Nigeria”, Africa Research Consortium, Research paper 154.

Owuor, S.O. and Foeken, D. 2000, “Urban farmers in Nakuru, Kenya. International Symposium. Berlin,
German, July 7-9.

Tchale, H. and Sauer, J. 2007, “The efficiency of maize farming in Malawi: A bootstrapped translog frontier.
Faculty of Social Sciences, Kent Business School”. Agri-Environment Economics. Cahiers d’economie et
Sociologie Rurales.82-83:34- 56.

Tolera, T., Temesgen, D. and Rajan, S. 2014, “Factors Affecting Farmers Willingness to Pay for Extension
services in Haramaya District of Ethiopia”. International Journal of Agricultural Science Research. Vol 3(12),
pp. 268-277.

United Nations Environment Programme (UNEP) 2012, “Green Economy Sectoral Study: BioTrade – A catalyst
for transitioning to a green economy in Namibia”

United Nation Partner Framework (UNPAF). 2014, “United Nations Development Programme in Namibia”.

85