Assessing the impact of the adoption of agroforestry technology on food production and poverty reduction among farming households in Oyo State, Nigeria

Poverty in Nigeria has been on the increase, notwithstanding several governmental and nongovernmental interventions to reduce it through poverty alleviation/reduction programmes and projects (Adepoju and Okunmadewa, 2010). Estimate of poverty incidence from 2009–10 was 53.5% according to the international poverty line of $1.90 per person per day (2011 Purchasing Power Parity (PPP)). In 2016, it was projected to have fallen to 48.4%. But owing to slow growth, poverty has been on the rise (World Bank, 2018). The United Nations Human Development Report (2019) declares that Nigeria’s Human Development Index (HDI) value for 2018 is 0.534 which puts the country in the low human development category, positioning it at 158 out of 189 countries and territories.


Introduction
Poverty in Nigeria has been on the increase, notwithstanding several governmental and nongovernmental interventions to reduce it through poverty alleviation/reduction programmes and projects (Adepoju and Okunmadewa, 2010). Estimate of poverty incidence from 2009-10 was 53.5% according to the international poverty line of $1.90 per person per day (2011 Purchasing Power Parity (PPP)). In 2016, it was projected to have fallen to 48.4%. But owing to slow growth, poverty has been on the rise (World Bank, 2018). The United Nations Human Development Report (2019) declares that Nigeria's Human Development Index (HDI) value for 2018 is 0.534 which puts the country in the low human development category, positioning it at 158 out of 189 countries and territories.
In Nigeria, poverty has persisted and many interventions have not yielded noticeable improvement in the country's Human Development Index. Plagued with the challenges of the effects of Covid-19, unemployment crises, climate change, conflict, fragility and violence, Nigeria (the most populous country in Africa) stands at a grave risk if poverty is not tackled (Danaan, 2018). Therefore, reducing poverty is an important development policy issue because economic growth is obviously associated with poverty reduction (Osowole, 2011).

Assessing the impact of the adoption of agroforestry technology on food production and poverty reduction among farming households in Oyo State, Nigeria
Poverty, which has been defined as the deprivation of well-being related to lack of material income or consumption, low levels of education and health, vulnerability and exposure to risk, lack of agency, and powerlessness, has remained a threat and challenge to humanity in all ramification (World Bank, 2001;Maduka, 2007). In Africa, poverty remains an important topic and has been connected to climate change and its associated hazards and food insecurity (Beegle et al., 2016;Thornton et al., 2006;Charles, 2019). Therefore, reducing poverty by means of improvement in the environment, sustainable use of natural resources and sustainableland management practices, such as agroforestry, should be at the forefront of major global initiatives (Adams et al., 2004;Griggs et al., 2013).
Agroforestry has earned a distinct identity as an approach to sustainable landuse (Rahman et al., 2010). Garrity and Stapleton (2011) noted that agroforestry is one of mankind best hopes to create a climate-smart agriculture, increase food security, alleviate rural poverty and achieve a truly sustainable development. Meanwhile, Agroforestry has been defined as a land use system in which woody perennials are grown with food crops and/or livestock leading to many beneficial, ecological and economic interactions between trees and non-trees components (Idumah et al., 2018). The International Council for Research in Agroforestry (ICRAF) now World Agroforestry Centre also defined agroforestry as a "dynamic ecologically based natural resources management system that through interactions of trees on farm and in the agricultural landscape diversifies and sustains production, enhancing social, economic and environmental benefits for land users at all levels" (Idumah et al., 2018;Rahman et al., 2010;Mutua et al., 2014;Kennedy et al., 2016).
Agroforestry not only helps to increase food and fodder but also protects the existing forest where the unemployed and the poor earn their livelihoods. Agroforestry also helps to lift rural poor from poverty through marketdriven, locally led tree cultivation systems that generate income and build assets; conserve biodiversity through integrated conservation-development solutions based on agroforestry technologies (Rahman et al., 2010;Kandji et al., 2006). Furthermore, it can protect forest through agroforestry-based solutions; assist the rural poor to better adapt to change and derive benefits from emerging carbon markets, through tree cultivation (Rahman et al., 2010;Ajayi and Catacutan, 2012;Maren and Carolyn, 2011).
Studies have shown that the adoption of agroforestry technology in some parts of Nigeria has proven to prevent some environmental problems like soil erosion, flooding and desert encroachment in addition to ensuring that several thousand hectares of marginal lands are salvaged and planted with economic trees as well as securing land for the farming population for food production (Adeola, 2015). Further studies have also revealed that agroforestry practices enhance farmers' income, food security and improve on poverty status of the farming households (Tiwari et al., 2017;Olajuyigbe, 2016;Kareem et al., 2017).
Despite these evidences that revealed the benefits of agroforestry technology's adoption, not many farmers have adopted these technologies. This may probably be due to several factors like inadequate knowledge of these technologies, non-availability of seed/seedlings, lack of appropriate skills, insecure land tenure/land ownership pattern, inability to understand agroforestry practices, decrease in crop yield, lack of access to credit facilities, inadequate source of information on the practice etc. (Akinwalere, 2016;Owombo and Idumah, 2017).
This study therefore, attempts to address the gaps in the previous studies by assessing the significant impact of the adoption of agroforestry technology on income generation, farmers' output and poverty reduction strategies as well as the factors militating against the adoption of agroforestry among farming households in Oyo State, Nigeria, under the following specific objectives: 1. To determine the impact of agroforestry practices on food production and income generation in the study area. 2. To determine the contribution of agroforestry practices to poverty reduction in the study area. 3. To identify factors militating against the adoption of agroforestry in the study area.

Study area
The study was carried out in Oyo State, Nigeria. Oyo State is located between latitude 20,381 and 40,351 east of the Greenwich meridian. Oyo State is located in the south-west zone of Nigeria. The state consists of thirtythree Local Government Areas (LGAs) and covers an area of 28,454 square kilometers. Agriculture is the main occupation of the people and small-scale traditional farming system predominates in the area. The bulk of the produce come from annually cultivated rain-fed farms. Oyo state has four Agricultural Development Programme (ADP) zones, namely; Ibadan/Ibarapa zone, Oyo zone, Saki zone and Ogbomoso zone (see Figure 1). Ibadan/ Ibarapa zone has fourteen LGAs/Blocks, Oyo with only five, Saki zone with nine zones and Ogbomoso has only five LGAs/Blocks in their zones.

Sampling procedure
A multi-stage sampling technique was used to select the respondents from the study area. The first stage was the random selection of two Agricultural Development Project (ADP) zones out of the four agro-ecological or ADP zones in Oyo state since they all have a dominance of rural farmers. The selected zones were Ibadan/Ibarapa and Saki zones. The second stage was the random selection of Local Government Areas also known as ADP Blocks from the selected zones. Six Local Government Areas (LGAs) were selected from Ibadan/Ibarapa zone while four LGAs were chosen from Saki Zone. The selection was done based on the proportion of number of local government areas in each of the selected zones. The third stage was the random selection of two communities in each Local Government Area, making a total of twenty communities in all. The fourth stage was the random selection of twenty farmers in each Local Government Area making a total of two hundred copies of questionnaires. However, only 199 copies of the questionnaire were eventually utilized for the analysis due to non-recovery of a copy of the administered questionnaire.

Propensity score analysis
This is an important tool to identify imbalance in covariates with regard to treated and comparison groups. The objective of establishing a propensity score is for covariate balancing between individuals who did and did not receive a treatment, thereby ensuring easy isolation of the effect of a treatment.
A propensity score (p) for an individual (i) is defined in Rosenbaum and Rubin (1983) as the conditional probability (P) of assigning a participant to a particular treatment or comparison group (T) given a set of covariates (X), expressed as: In theory, pertinent pre-treatment variables are used to obtain probabilities of group membership which are used to match participants in treatment and comparison groups in a way that both groups have equal means or likelihoods of receiving treatment. After matching, differences between groups should reflect real treatment effects in the population and similar to the interpretation of randomized designs.
As covariates are identified, probabilities of group membership or propensity scores are estimated for all participants. Logistic regression is the most commonly used estimation technique (Guo and Fraser, 2010;Thoemmes and Kim, 2011) and is relatively easy to interpret given that the predicted probabilities (P) of group membership (T) are the propensity scores (p) for a given set of covariates (X). (2)

Figure 1
Map showing the Four ADP zones in Oyo State, Nigeria Source: Cartographic Laboratory, IFSERAR, FUNAAB, 2016 This study therefore adopted the use of Propensity Score Matching (PSM) model to analyze the impact of the adoption of agroforestry technology on income and food production of farmers in the study area. Propensity Score Matching simply means matching treated and untreated observations on the estimated probability of being treated (propensity score). The ideal comparison group is selected such that it matches the treatment group using a comprehensive baseline either survey or time invariant characteristics. The matches are selected based on of similarities in observed characteristics. This assumes no selection bias based on unobserved characteristics.
Therefore, let P(X) = Pr (z = 1|x) represent the probability of adopting agroforestry technology, that is the propensity score. Propensity Score Matching will then construct a statistical comparison group by matching observations on the agroforestry adopters to non-adopters for similar values of propensity score. Rather than create a match for each adopter with exactly the same value of X, we can therefore match the probability of adoption.
Therefore, to evaluate the impact of Agroforestry technology adoption on income and food production (in terms of crop yield/year), a measure of the impact was used to compare the outcome of those who adopted agroforestry practices and those who did not adopt: Let Y 1 = adopters of agroforestry technology, Y 0 = non-adopters of agroforestry technology.
The impact of the adoption will therefore be the change in the mean outcome caused by adopting the technology.
Since it may not be possible to estimate individual treatment effects in equation 1 directly, the evaluation parameter, which is the Average impact of the treatment on the treated (ATT), was introduced where: Z -an indicator variable, showing whether a respondent actually adopted agroforestry technology or not. It is equal to 1 if respondent adopted and 0 if otherwise. X denotes a vector of control variables. STATA 12 Version was used for this analysis

Foster Greer Thorbecke (FGT) analysis
To assess the impact of agroforestry technology adoption on poverty reduction in the study area, poverty line was first constructed. Poverty line is described as a borderline that distinguishes poor from non-poor households in terms of their level of welfare. Two approaches are commonly adopted to determine the poverty line namely: expenditure and income approach.
To compare the level of poverty among the adopters and non-adopters of agroforestry technology, Foster Greer Thorbecke measure of poverty was adopted.
Following Foster, Greer and Thorbecke (1984), the model is expressed as: ( 5) where: Z -the poverty line defined as 2/3 of Mean annual per capita expenditure; y -the annual per capita expenditure -poverty indicator/welfare index per capita; q -the number of poor households in the population of size n; a -the degree of poverty aversion; a = 0 -is the Headcount index (P 0 ) measuring the incidence of poverty (proportion of the total population of a given group that is poor, based on poverty line); a = 1 -is the poverty gap index measuring the depth of poverty that is on average how far the poor is from the poverty line; a = 2 -is the squared poverty gap measuring the severity of poverty among households, that is the depth of poverty and inequality among the poor

Results and discussion
From Table 1, it was discovered that majority (about 65%) of the adopters of agroforestry technology were between 40 and 59 years of age while 68.61% of non-adopters were within the same age range. Majority (38.29%) of the adopters had secondary education while majority (35.23%) of non-adopters had tertiary education. Large proportion (95.74%) of the adopters was male while about 85% of the non-adopters were male, with female accounting for 4.26% and 15.24% of the adopters and non-adopters respectively. In addition, majority (72.68%) of the adopters had farm size of 8 ha and below while 83.81% of the non-adopters had equivalent land holdings. This is an indication that majority of the farm households in the study area are small scale farmers. This is because, according to Ozowa (2005), farm households with less than 10 ha of farmland are regarded as smallscale farmers. This is according to international standards measurement for farm sizes. It was also observed that the adopters of agroforestry technology had larger household size than the non-adopters with 73.4% of the adopters having household size from 6 to 15 while 64.76% of the non-adopters had same household range. This may not be unconnected to the fact that adopters with larger household sizes realized that they had more people to cater for, hence the need to add tree planting (especially multipurpose and fruit trees) to their farming  Figure 2 shows propensity score distribution and common support for propensity score estimation. The "treated" in the figure shows the observations in the adopters' group that have a suitable comparison. The balancing test was thereafter applied to find out if the differences in the covariates of the two categories in the matched sample have been eliminated, in which case, the matched comparison group can be considered a plausible counterfactual (Ali and Abdulai, 2010). Table 2 shows the results from the covariate balancing tests both before and after matching. The standardized mean difference of 28% before matching has now decreased to about 8.4% after matching. Therefore, it can be concluded that the matching process decreased the total bias. In addition, the ratio of variances of the propensity score and covariates from the adopters and non-adopters are all near 1 except for household size where the unmatched was significant.
But after matching, the bias was significantly reduced by about 87% and the ratio of the variances after matching is now not statistically different from 1, as shown in Table 3. Therefore, the treatment (adopters) and comparison (non-adopters) groups are said to be balanced. This is because, according to Rubin (2001), ratio of variances of propensity score and covariates from the treatment and comparison groups should be near one if the treatment and comparison groups are balanced.

Figure 2
Graphical representation of propensity score match among adopters and non-adopters of agroforestry technology   Table 5 shows the result on the impact of the adoption of agroforestry practices on farmers' income in the study area. The result from the PSM analysis shows that the adoption produces a positive and significant impact on the farmers' income. The Average Treatment Effect on the treated (ATT) of the entire population of the adopters in the study area was $950.33 per year. This implies that the average annual income of the adopters increased by $950.33 per year. In addition, the average effect of the treatment (ATE) on the entire population in the study area is $664.63. That is if any adopter of agroforestry practices is picked at random in the study area, the average annual income will be $664.63. The implication of this is that if both adopters and non-adopters of agroforestry technology were considered, income due to adopters will increase by about $664.63 per year. Furthermore, for the impact on the non-adopters, the average treatment on the untreated (ATU) had a positive impact but not significant effect on income which will increase by $408.86 provided they were treated. Therefore, adoption of agroforestry technology will lead to an increase in farmers' income in the study area. This finding agrees with Idumah and Akintan (2014) and Idumah and Owombo (2019) who reported that agroforestry enhanced the income of agroforestry farmers in Edo state, Nigeria. Likewise, Sarvade and Singh (2014) posited that agroforestry helps to enhance nutrition, health and income of rural poor and described the system as the best one option for tackling food security problem. This could be attributed to multipurpose benefits derivable from the practice of agroforestry which in addition to enriching soil nutrients; it serves as source of income to farmers, as farmers can generate income from sales of timber for construction or sales of wood as poles. Agroforestry also enables farmers to generate income fruit trees planted on their farms. All these therefore accounted for the positive and significant impact of agroforestry on the income of the adopters in the study area. Table 6 shows the value of income generated by the farming households in the study area as well as the value of the poverty line. The values of poverty line computed for the adopters and non-adopters of agroforestry technology were $183.25 per year and $102.21 per year respectively. The households that earn less than the value of poverty line was considered to be poor and those above the poverty line are regarded as non-poor. It was then discovered that about 27% of the adopters fell below the poverty line and were therefore regarded as poor while about 67% of the non-adopters fell below the poverty line and can therefore be described as poor.
Foster Greer Thorbecke (FGT) poverty index was used to show the extent of poverty among the farming households in the study area. The poverty aversion parameters used were P 0 , P 1 , and P 2 which translate to poverty incidence (headcount), poverty depth (gap) and poverty severity respectively. Table 7 revealed that the incidence of poverty (P 0 ) among the adopters was 0.2659 and 0.6667 for the non-adopters indicating that 26.6% of the adopters were poor and 66.67% of the non-adopters were poor, with reference to the poverty line. The value of the poverty depth (P 1 ) among the adopters was 0.4481 and 0.6982 for the non-adopters. This implies that an average poor adopter of agroforestry technology would require 44.8% of the poverty line to get out of poverty  while an average poor non-adopter would need 69.8% of the poverty line to get out of poverty. In addition, the value of poverty severity (P 2 ) among the adopters was 0.2204 and 0.4994 for the non-adopters. This therefore implies that the poverty severity among the adopters was 22.04% and 49.9% for the non-adopters. It was therefore discovered from the study that the adopters of agroforestry technology were faring better than the non-adopters of agroforestry technology, as they recorded lower poverty incidence and depth than the non-adopters of the technology. It could therefore be inferred that adoption of agroforestry technology plays significant role in poverty reduction in the study area.
There are many constraints militating against the adoption of agroforestry technologies by farmers in the study area. This study therefore highlighted some of the problems faced by farmers in the adoption of agroforestry in the study area. Some of the problems examined includes; insufficient land for tree planting, illegal felling of trees, long gestation period of trees, lack of technical assistance, lack of planting materials, lack of knowledge and skills as well as competition among   trees and arable crops on farmland. It was discovered that 59.30% of the respondents (both adopters and nonadopters) stated that lack of knowledge and required skills on agroforestry was a constraint to their adoption of agroforestry and this ranked highest among the constraints, as shown in Table 8. This possibly explains why the adopters of agroforestry technology in the study area could not efficiently explore the potentials of agroforestry technology. This therefore accounted for the reduction in farm outputs of the adopters relative to the non-adopters. Furthermore, ranking second highest among the constraints militating against the adoption of agroforestry technology among the farmers in the study area is the gestation period of agroforestry trees as 51.26% of them described the long gestation period of agroforestry trees as a constraint. In addition, 49.75% of the respondents stated insufficient land as a constraint to the adoption of agroforestry. This is where expertise of the extension agents and some subject matter specialists are needed to train and enlighten the farmers on how to make effective use of their land to accommodate both their tree crops and arable crops. This work is therefore in line with the study by Amonum and Bada (2019) where lack of land, lack of tree seedlings as well as inadequate extension personnel were stated as some of the constraints affecting the adoption of agroforestry in Katsina State of Nigeria.

Conclusions
It can be inferred that the impact of the adoption of agroforestry practices on farmers' income in the study area produces a positive and significant impact on the farmers' income while the result of the impact of adoption of agroforestry practices on farmers' output in the study area produces a negative but not significant impact on the farmers' yields. This may not be unconnected to improper practices of the technologies among the farmers. It was discovered that about 27% of the adopters fell below the poverty line and were therefore regarded as poor while about 67% of the non-adopters fell below the poverty line and can therefore be described as poor. It was therefore concluded that the adopters of agroforestry technology were faring better than the non-adopters of agroforestry technology. Agroforestry practices can therefore be said to play prominent role in the enhancement of farmers' income, reduction of poverty and the overall improvement in livelihood within the study area.
It is therefore, recommended that efforts should be geared towards increasing adoption of agroforestry technology through awareness and sensitization of farmers so as to enjoy the benefits of agroforestry practices. There is the need for training and enlightenment of the farmers, by extension agents and specialists on agroforestry technology, on how to make effective use of their land to accommodate both their tree crops and arable crops for improved yield and productivity. Farmers should also be provided with necessary and early-maturing tree seedling by government through the appropriate research institute (such as Forestry Research Institute of Nigeria, FRIN) to encourage and motivate the farmers to adopt agroforestry technology.