1. Introduction
Pesticide, a damage control input to safeguard from insects and other pests, is considered to improve nutrition in food, and its use is assumed an economic, labour-saving, and efficient tool for pest management [
1]. Furthermore, pesticide is believed to improve competitive advantage in agriculture [
2]. This is because pesticide use is deemed essential for retaining current production and yield levels, as well as maintaining a high-quality standard of life [
2]. There is a widespread acceptance that the use of modern agricultural technologies has led to a sharp increase in pesticide use, along with other modern inputs, in the developing economies [
3,
4,
5]. However, there is a widespread claim that pesticides are harmful to human health and the environment [
6,
7]. The environmental and social impact of pesticide use in the USA alone is estimated at USD 10 billion per year [
7]. An estimated 1–5 million farm workers suffer from pesticide poisoning every year, and at least 20,000 die annually from exposure, mostly in developing countries [
8].
Both Asia and Latin America experienced dramatic increases in agricultural productivity due to rapid and widespread adoption of Green Revolution (GR) technologies, which incorporate widespread use of modern agricultural inputs and agro-chemicals [
9]. However, Sub-Saharan Africa (SSA) did not or could not participate in this drive for GR technologies of the 1970s–1980s, and therefore could not gain from the application of modern agricultural inputs and agro-chemicals [
10]. In fact, the low use of modern inputs, including pesticides, is assumed to be the norm in SSA agriculture, which led to the setting up of policy directives and programs such as the Comprehensive Africa Agriculture Development Program (CAADP), Abuja Declaration and Malabo Declaration [
10].
Nigeria, the largest economy in Africa, is largely dependent on its agricultural sector for the supply of raw materials, food, and foreign exchange, and employs over 70% of the labor force [
11]. Small-scale semi-subsistence farmers comprising more than 70 million farmers/rural citizens [
11] also dominate the sector. The agricultural sector is characterized by low level of productivity and modern technology adoption [
12,
13].
Cassava and yam are the main staple food crops in Nigeria with a wide range of industrial and commercial uses as well [
14]. The country is one of the leading producers of cassava and yam in the world, supplying more than 68% of global yam production [
15,
16]. However, over the past two decades, rice has also been introduced as a major staple food crop in Nigeria, growing at an annual rate of 14% from 1990 onward [
17]. Manyong et al. [
18] noted that the major constraints on improving agriculture in Nigeria is the subsistence production system, the low level of modern technology adoption, land fragmentation, and crop failure, which increases production risk. Lack of the use of damage control inputs, e.g., pesticides, further increases the risk of crop losses. This is because about 20–40% of potential food produced is lost to insects and other pests in Africa [
19].
Explaining variation in pesticide use intensity at the farm level is quite complex and not well explored in the literature [
20]. A limited number of studies are available that examine various aspects and/or determinants of pesticide use at the farm level in Africa [
10,
19,
21,
22,
23,
24]. Sheahan and Barrett [
10] utilized a large-scale multi-country nationally representative dataset generated through the Living Standard Measurement Study-Integrated Surveys on Agriculture (LSMS-ISA) with 22,565 households and 62,387 plots from six countries (Ethiopia, Malawi, Niger, Nigeria, Tanzania and Uganda) collected during 2010–2012. The project aimed to investigate the broader question of the current level of modern input use in agriculture, which also includes the use of agro-chemicals. One of their main conclusions is that although modern input use may be relatively low in aggregate, it is not equally low across six countries, particularly regarding the use of inorganic fertilizers and agro-chemicals [
10]. Anang and Amikuzuno [
19], using a sample of 300 rice farmers from Northern Ghana, reported that a number of socio-economic factors influence farmers’ decision to choose pesticides. Adeniyi et al. [
21], using a sample of 100 cocoa farmers from Osun State, Nigeria, noted that pesticide price is an important determinant of pesticide use. Oesterlund et al. [
23], using a sample of 317 small-scale farmers in Uganda, concluded that the farmers do not use the most dangerous WHO Class 1a and 1b pesticides but mostly use WHO Class II pesticides, and have poor knowledge of the level of toxicity and poor protection practices. Mwatawala and Yeyeye [
22], using a sample of 91 tomato farmers from Morogoro region in Tanzania, noted that the farmers are generally aware of the laws, environment, and consumer health, but could not name a single act, and only 21% of them used the correct dose of pesticides. Idris et al. [
24], using a sample of 50 cocoa farmers from Ogun State, Nigeria, noted that most of the farmers applied fungicides because of the black pod disease. It is clear from the aforementioned brief review that the studies were mainly conducted on a single crop (e.g., cocoa, tomato, and rice) with limited sample sizes, and Sheahan and Barrett [
10] did not provide any detailed information on the factors influencing agro-chemical use. In addition, none has examined the use of pesticides on major food staples, i.e., yam and cassava, in Africa in general and Nigeria in particular.
Given this backdrop, the present study examines the influence of a range of price and socio-economic factors on pesticide use in producing multiple food crops (rice, yam, and cassava) by using a survey data of 400 farm households from two states (Ebonyi and Anambra) of South-eastern Nigeria. Our specific contributions to the existing literature are as follows: (a) We have examined the extent of pesticide use by type and/or combinations of food crops produced by the farmer, which may provide further insight into whether pesticide use varies across cropping portfolio; (b) We have tested the farm-size and pesticide use relationship; and (c) We have incorporated a wide range of variables including prices, socio-economic factors, and variables representing commercial motive of the farmers to explain pesticide use at the farm level, which are not seen commonly in the literature. However, we could not use agroecological, climatic, land elevation, and/or political-economic variables applied by Rahman [
25] and Galt [
20] because of a lack of information in these areas, which are also important in explaining the complexities of pesticide use in crops.
The paper is organised as follows.
Section 2 presents the analytical framework, the description of the study area, data, and the empirical model.
Section 3 presents the results.
Section 4 provides conclusions and draws policy implications.
2. Methodology
2.1. Theoretical Framework
The study utilizes a farm production model based on the profit maximizing behaviour of the farmers adopted by Rahman [
4,
25,
26]. Consider a model with two variable input vectors: pesticides,
H, and ‘other inputs’,
X, and one fixed input of land,
L, to produce
n number of crops (
i = 1 …
n), in which
Li is land area allocated to the
ith crop.
Farmer
j maximizes total profits:
The first order conditions lead to the corresponding demand functions for pesticides (
Hj) and for ‘other inputs’ (
Xj) for individual crops:
in which p’s and w’s are output and input prices, respectively.
We can aggregate the pesticide demand functions of individual crops
(Qj) as follows:
in which
H’j = aggregate pesticide demand.
The assumption of the separability of inputs (pesticide on one hand, and all ‘other inputs’ on the other) enables the pesticide demand equation to be estimated separately [
4,
25,
26].
2.2. Study Area and the Data
A multi-stage sampling procedure was utilized for this study. First, two states, Ebonyi and Anambra, from Southeastern Nigeria were purposively selected. Then, three local government areas (LGAs) from each state were selected randomly based on the cell structure developed by the Agricultural Development Program. Then, 10 communities/villages from each LGA were chosen randomly. Finally, a simple random sampling procedure was applied to choose farmers from these communities. Using the total number of farm households in each village as the sample frame, the sample size (
n) of households was determined [
27]:
in which
n = sample size,
N = total number of farm households,
z = confidence level (at 95% level
z = 1.96);
p = estimated population proportion (0.5, this maximizes the sample size), and
d = error limit of 5% (0.05).
The required total sample size by applying the sampling formula in Equation (6) is 450. However, due to difficulty of data collection in developing countries and usability of the returned questionnaires, a reserve of 33% sample was added. As such, 600 questionnaires were distributed (300 in each state with 30 in each community), of which 290 from Ebonyi and 190 from Anambra states were returned. However, complete information was available in only 249 and 141 questionnaires from Ebonyi and Anambra, respectively. Therefore, the final sample size stands at 400 households. Details on input and output data for each of the three major food crops (i.e., cassava, yam, and rice) by the farmers were recorded separately. In addition, key demographic and socio-economic information from each of the farm households were also collected. The field survey was very intensive and carried out during the months of October and November 2011. The questionnaire was pre-tested and modified as required prior to final administration. Farmers were asked to provide details of their production activities, level of inputs used, and outputs produced individually for each of the major food crops covering the crop year 2010–2011 based on their recall. Therefore, all quantity and price data used in the analysis are actual data provided by the farmers specific to each crop produced. The co-author and two research assistants who are the final year agricultural undergraduate students were used for collecting primary data. The co-author trained the research assistants on the questionnaire and survey methodology prior to data collection.
2.3. The Empirical Model
Since not all farmers use pesticides in their production process, meaning that the dependent variable is censored at zero, the Tobit model provides a suitable method for estimating the pesticide demand equation in this case, as it allows for zero use of inputs [
4,
25,
26].
The stochastic model underlying Tobit may be expressed as follows:
H’j* is a latent variable such that:
in which the disturbances
uj are an error term and are independent and identically distributed as
N(0,
σ2). The econometric software STATA V. 10 (StataCorp., College Station, TX, USA) was used for the analysis.
2.4. Variables
The dependent variable in the econometric model is the total amount of pesticides (measured in litres of concentrated pre-prepared form as purchased from the market) applied to each of the three major food crops (i.e., rice, yam, and cassava). Brief details of the type of pesticides used in these three major food crops are presented in
Section 3. We did not include the price of pesticides because of the unavailability of correct information, although Rahman [
4] reported the correct negative impact of own price of pesticides on its demand.
The variables included in the pesticide demand function were: (a) input prices—weighted average prices of inorganic fertilizers applied to each of the three food crops, weighted average of the wages of labour used in each of the three food crops, and weighted average per unit cost of ploughing services used (this is mainly the cost of labour used exclusively for land preparation) in each of the three food crops; (b) output prices—price of rice, price of yam, and price of cassava; (c) total amount of manure applied to all three crops; (d) a set of socio-economic characteristics that includes total farm operation size, experience of the farmer, average education of the farmer, average family size of the household, amount of agricultural credit, share of land rented in for cultivation, number of extension contacts, and distance to nearest agricultural extension office; (e) dummy variable to represent gender of the farmer; (f) share of rice area; (g) share of cassava area; and (h) two variables representing motives (i.e., high profit and high yield) behind adopting modern inputs and technologies (i.e., use of high yielding varieties of seed, inorganic fertilizers and pesticides) were included in the model, since adoption of modern technology influence pesticide use [
4,
25].
Table 1 presents the definition, measurement, and summary statistics of all the variables used in the econometric model. The choice of these variables was based on the literature and justification thereof [
4,
10,
19,
20,
21,
25,
26].
All price variables used in this study were reported by the farmers for purchase of inputs and sale of outputs. For the family-supplied inputs, such as family labour, the market wage paid by the farmers was imputed. Ploughing price, which is mainly the cost of labour used exclusively for land preparation, was treated as a separate variable, because wages varied for this operation as compared to the wages paid for other farming operations. Again, for family-supplied labour for this operation, corresponding market wage was imputed.
The share of rice area and the share of cassava area were included in the pesticide demand function. This is because although both are staple food crops, rice is mainly destined for the market and provide an indication of the level of commercialization in food production, which is not commonly seen in the literature. Finally, two variables representing motivation behind using modern agricultural inputs were included to check their independent influence on pesticide use, reflecting commercialization motive of the farmers. Farmers were asked about the motivation for making their crop choice decision and to rank each of the motives (e.g., high yield and high profit) on a five-point Likert scale (i.e., 1 for least important motive and 5 for most important motive). The variables are the weighted average rank values of the motives.
2.5. Variance Analyses
The study also used other forms of analysis. First, a One Way Analysis of Variance (ANOVA), was planned to be utilised to examine existence of systematic variation in pesticide use rates and cost of pesticide use per ha across various categories of classification, e.g., by crop combinations and by farm size categories. The underlying assumption in conducting ANOVA is that the population of farmers from which the sample was drawn who were producing major food crops and applying pesticides in their farming operations is normally distributed. Although, the graphical plots of the pesticide use rate and cost of pesticide used per ha for each crop in a histrogram with normal curve imposed showed an approximate normal distribution with a few outliers, the Levene’s test for the Homogeneity of Variance showed that the variances are significantly different. An additional Brown-Forsythe’s robust test for the equality of means showed that the means between categories are significantly different. Nevertheless, due to some concerns of the robustness of Brown-Forsythe test in the literature, we have also conducted the non-parametric Kruskal-Wallis test to identify systematic differences across categories (details of the test results were presented in the bottom panel of Tables 3 and 4 below). The SPSS V. 24 (IBM Corporation, Armonk, NY, USA) was used to conduct cross tabulation and variance analyses.
2.6. Multicollinearity
In a regression analysis with multiple variables, multicollinearity can be a problem [
20]. Therefore, it is important to check existence of collinearity amongst variables used in the econometric model. Pairwise correlation tests were conducted for all the variables used in the model. Results showed that although less than 50% of the variables are significantly correlated, all the correlation coefficients were under 0.4, except correlation between total farm size and rice price (
r = 0.53,
p < 0.05) and education and farming experience (
r = −0.55,
p < 0.05). A general rule of thumb of the presence of serious multicollinearity is to have correlation coefficient in excess of 0.6 for more than two variables [
28]. Therefore, multicollinearity is not an issue in this study.
4. Conclusions and Policy Implications
The principal aim of this study was to examine the level and extent of pesticide use in multiple food crops and identify the influence of prices and socio-economic factors on the pesticide demand of a sample of 400 farms from two states (Ebonyi and Anambra) from Southeastern Nigeria. Farmers produce multiple food crops, as proven by the fact that 68% of the farmers produced at least two food crops. Results show that pesticide use is strongly influenced by a host of price and socio-economic factors of the farmer and the farming household, with varied effects. Although the overall proportion of farmers applying pesticides is relatively low, estimated at 41%, there is a wide variation in the proportion of farmers using pesticides in various food crops and crop combinations. Significant variation also exists with respect to pesticide use rates and cost of pesticides per ha amongst farmers producing various crops and crop combinations.
Pesticide use rate and cost per ha is highest for the farmers producing yam followed by cassava, and for those who produced both yam and cassava followed by rice and cassava. Farmers treat pesticides as substitutes for labor and ploughing services. The implication is that a rise in labor wage and ploughing price will induce a significant rise in pesticide use mainly to reduce the amount of labor for various farm operations and ploughing activities. On the other hand, an increase in the price of yam, which is desirable for increasing income of the farmers, would lead to a significant increase in pesticide use. Nevertheless, since actual pesticide use rate is relatively low in Nigeria, the level of increase in pesticide use relative to a rise in yam price will not be very large.
Inverse farm size–pesticide use rate exists in the study areas, i.e., the pesticide use rate is highest for the small farmers (p < 0.01), estimated at 1.16 L/ha costing Naira 1788.70 per ha. The pesticide use rate is relatively higher for a crop that is mainly produced for the market but which is also a staple crop of the economy. This was confirmed econometrically by significant influence of the share of rice area on pesticide use. In contrast, production of cassava uses significantly fewer pesticides. Significant gender differences exist with respect to pesticide use as male farmers use significantly higher pesticides. Farming experience significantly increases pesticide use.
The following policy implications can be drawn from the results of this study. First, land reform policies aimed at increasing the farm operation size of individual farmers could lead to a reduction in pesticide use. Galt [
20] also noted that land reform with local backing could reduce pesticide use in vegetable farming in Costa Rica. Second, investment in programmes to promote expansion of cassava, e.g., Cassava Plus project, will significantly reduce pesticide use in major staple food crop production, as was also noted by Galt [
20] for vegetable farming in Costa Rica. Third, policies to encourage female farmers to engage in farming are likely to reduce pesticide use in food production.
Although the effective implementation of these policies is challenging, a significant reduction in pesticide use is important for sustaining the agricultural sector, as well as for safeguarding the farming population, which is a worthwhile goal to pursue.