The Impact of Agricultural Technology Adoption on Poverty: The Case of Yam Minisetts Technology in Ekiti State, Nigeria

Data Sources

Researchers in Nigeria of the National Root Crops Research Institute (NRCRI), Umudike, and IITA, Ibadan, developed the minisett technique to overcome the critical problem of the unavailability of good quality seed yam by improving the rate of multiplication of white yam [7]. With the technique, the multiplication ratio can increase from the traditional 1:5 to 1:30 [8]. The study used primary data collected from a cross-sectional survey of yam farmers from Ekiti State with the aid of a structured questionnaire. The survey collected valuable information on several factors including household composition and characteristic, revenue and cost of operation of yam production. The revenue(sales from yam output)in naira, while cost of operation include: quantity and cost of inputs (yam minisetts(kg), labour(man-days), fertilizer(kg), land and non-land farm assets, yam production cost and socio-economic characteristics of the farmers in the study area (education, experience and family size). Also data were collected on general socioeconomic characteristics of respondents, which include age, gender, marital status, household size, educational level, major occupation, secondary occupations, farming experience, number of land assets owned and cultivated, residency type, and membership of association. In addition, information on the sources of income and food consumption expenditure were collected.

Sampling Procedure

A multi-stage sampling technique was used for this study. In the first stage, three Local Government Areas (LGAs) where yams are predominantly grown were purposively chosen (Gbonyin, Ijero and Efon local government). In the second stage, four largest yam producing communities were purposively selected from each LGA, resulting to the selection of 12 communities for the study. Then, farmers in the selected communities were stratified into adopters and non adopters of yam minisetts technology. The last stage was the random selection of 50 yam farmers from each of the 12 communities, giving a sample size of 600 yam farmers (300 from adopters and 300 from non adopters). For the analysis in the paper, both adopters and non adopters of the agricultural technologies were selected from the same geographical location, same geographical proximity, same production system, same weather conditions, and same altitude.

The Conceptual/Analytical Framework

We assumed that for a farmer to make decision on whether or not to adopt the yam minisetts technology, he/she must have first examined the benefit obtainable from the adoption and benefit derived from non-adoption. A farmer is likely to adopt the yam minisetts technology, if the expected utility derived from adoption is greater than the expected utility from non- adoption. Conditional on cross-sectional data availability, we estimated poverty level of yam farmers households based on their adoption of yam minisetts technology. The data obtained were subjected to descriptive and inferential statistical analysis. Descriptive statistics for this study include frequency tables, percentages and means. The inferential analyses adopted for the study are FGT poverty index measurement, Propensity Score Matching (PSM) and Probit model [9].

Or,

Propensity Score Matching (PSM)

To isolate the independent impact of technology adoption on poverty reduction, propensity score matching (PSM) was used. A logit model was used to estimate propensity scores using a composite of pre- intervention characteristics of the sampled households and matching was performed using propensity scores of each observation [10]. In estimating the logit model, the dependent variable was yam minisett technology adoption, which takes the value of 1 if a household is an adopter and 0 otherwise. The cumulative logistic probability function is specified as

1 𝑝𝐼= 𝐹(𝑍𝑖)= F 𝛼+ Σ𝛽𝑖 𝑋𝑡1)= (

1+𝑒− (𝑎+Σ𝛽𝑖 𝑋𝑡1) ) (1)

Where: e represents the base of natural logarithms (2.718) Xti represents the tth explanatory variable (t=1,2,…,m) for the ith individual Pi is the probability that ith individual will make a certain choice (in this case adoption of yam minisett technology) given m explanatory variables α & βt are parameters to be estimated (t = 1, 2,,,,,,,,m; m is number of explanatory variables). Interpretation of the coefficients will be understandable if the logistic model is written in terms of the odds and log of odds. The odds ratio implies the ratio of the probability that an individual would choose an alternative (Pi) to the probability that he/she would not choose it (1-Pi). But,

1 (1 - 𝑝𝐼 ) =

1+𝑒𝑧𝑖 (2) Therefore,

1+ 𝑒𝑧𝑖 𝑃𝑖 1− 𝑃𝑖 ) = (

1+𝑒−𝑧𝑖 )=𝑒𝑧𝑖 (3) (

1+ 𝑒𝑧𝑖 𝑃𝑖 1− 𝑃𝑖 ) =(

1+𝑒−𝑧𝑖 )= 𝑒(𝑎+ Σ𝛽𝑖 𝑋𝑡𝑖) (4)

( Taking the natural logarithm of equation above will result in what is known as the logit model as indicated below:

𝑃𝑖 1−𝑃𝑖) = 𝛼+ 𝛽1 𝑋𝑡1 + 𝛽2 𝑋𝑡2 + ----- + 𝛽𝑚𝑖 𝑋𝑚𝑖 (5)

𝑧𝐼 = In (

If the disturbance term _Ui_is taken in to account, the logit model becomes 𝑧𝐼= 𝛼+ Σ𝛽 𝑡 𝑋1𝑡 + 𝑈𝑖 (6) After running the logit model, then the common support region where the values of propensity scores of both adopters and comparison groups can be found was identified. The region of common support will be defined by dropping observations below the maximum of the minimums and above the minimum of the maximums of the balancing scores between the two groups [11]. Then the Average Treatment Effect on treated (ATT) are only determined in the region of common support [12].

The next step in propensity score matching is to get the matching algorithm which best matches the treated observations with untreated based on the propensity scores from the preceding step. Treatment, in this case, is adoption of technology. There are different matching estimators in theory. According to Caliendo and Kopeinig, the most commonly applied matching estimators are nearest neighbor matching Caliper and Radius matching, Stratification and Interval matching, Kernel and local linear matching [12]. All matching estimators contrast the outcome of a treated individual with outcomes of comparison group members [12]. To estimate the effect of yam minisett technology to a given outcome (consumption expenditure per adult equivalent), is specified as:

𝑝 NP

$$ \sum_ {j = 1} ^ {r} C _ {i j 1} - \sum_ {i = 1} ^ {M} C _ {i j 1 0} $$

$$ \frac {i = 1}{P} - \tag {7} $$

where, Cij1is the post intervention per capita

consumption expenditure of household j, Cij0is the per

capita consumption expenditure of the ith non-adopters

matched to the jth adopters, P is the total number of

adopters (adopters of yam minisett technology), NP is the

total number of non-adopters (non-adopters of minisett

technology) and C is difference in per capita consumption

expenditure in ekiti state. Then the average effect of

adoption of yam minisett technology on outcome

variables (consumption expenditure per adult equivalent)

will be computed and it is specified as:

C =

Ti= Yi(Di=1)- Yi(Di=0) (8)

Where Τi is treatment effect (effect due to adoption of minisett technology), Yi is the outcome on household, Di is whether household i has got the treatment or not (i.e. whether a household adopts minisett technology or not). However, one should notice that Ti=Yi(Di=1) and Ti= Yi(Di=0)_cannot be observed for the same household at the same time. Depending on the position of households in the treatment (intervention adopters), either _Ti=Yi(Di=1)_or _Ti=Yi(Di_=0)_is unobserved outcome (called counterfactual outcome). Due to this fact, estimating household's treatment effect is not possible. One has to shift to estimate the average treatment effect of the population than the individual one. The most commonly used average treatment effect estimation is the average treatment effect on the treated (τATT) and is specified as ƮATT = E (Ʈ|D=1) = E[Y(1)|D = 1-E[Y(0)|D] = 1 (9)

Estimation of Poverty Profile (Foster- Greer- Thorbecke)

Changes in poverty status of yam minisett technology adopters and non-adopters households was achieved by using the Foster- Greer- Thorbecke class of poverty measures (FGT) which include the Headcount Index (P0), the Poverty Gap Index (P1), and the severity of Poverty Index (P2) [9]. The three indices can be expressed into one general form and distinguish themselves for the different weights attributed to the distance between expenditure of the poor and the poverty line. P0 attributes equal weight to all expenditure of the poor while P1 and P2 attribute increasingly more weight to distance of expenditure of the poor from the poverty line. The FGT is presented below:

𝑞 𝑦𝑝 − 𝑦 𝑖 𝑃(𝑎) = 1 𝑛 𝑦 𝑝 }𝑎 (10)

𝑖=1 { where n is the number of sample households, yi is per capita consumption expenditure (consumption expenditure per adult equivalent) of the ith household, yp represents the poverty line defined as 2 3 of mean per capital expenditure, q is the number of households below the poverty line in the population size and α is the poverty aversion parameter. The poverty aversion parameter takes a value of 0, 1, or 2. If α = 0, then the result (p) is poverty head count ratio, which measures the incidence of poverty within the sample. When α = 1, the result (p) is a poverty gap index, which measures depth of poverty or the aggregate consumption shortfall of the poor from the poverty line. Finally, if α = 2, the result (p) is a squared poverty gap, which measures the severity or intensity of poverty.

Poverty Analysis: Estimation of Poverty Status among Yam Farming Households in Ekiti State

The estimated total per capita expenditure of the yam farmers’ household was ₦16,110,900.00 while the mean per capita expenditure was ₦ 134,257.50. The poverty line computed was ₦44,950.0 as the two thirds (2/3) of the per capita expenditure mean. Thus, the yam farming households whose per capital expenditure falls below poverty line were considered being poor, while those that falls above or equal to the value of poverty line were considered to be non-poor.

As shown in Table 2, the poverty incidence (P1) in the study area was 0.5642 indicating that 56.42% of the sampled yam farmers’ households were actually poor based on the poverty line. The poverty gap (P1) was 0.3173. This implies that the cost of eliminating poverty relative to the poverty line was 31.7%, because it shows how much would have to be transferred to the poor to bring their incomes or expenditures up to the poverty line (as a proportion of the poverty line).The poverty severity (P2) among the sampled yam farmers households was 0.2131, indicating that the poverty severity of poor yam farmers households was 21.3%.This implies the poorest among the poor farm households who require the attention of policy maker in the redistribution of the standard of living indicators, such as income generating activities.

Decomposition of Poverty based on Yam Minisett Technology Adoption

As depicted in Figure 1, the results of poverty profile based on yam minisett technology adoption shows that households that are non-adopters of yam minisetts technology were poorer than those of adopters in the study area. The incidence, depth and severity of poverty for non-adopters were 0.5742, 0.3333 and 0.2183 respectively while the corresponding figures for the adopters were 0.5600, 0.2989 and 0.2071 respectively.

Poverty incidence is slightly higher for non-adopters of yam minisetts technology than the adopters of yam minisetts technology (by about 1.42%). Hence, adopters of yam minisetts technology were less poor than the non- adopters of yam minisetts technology. The indices of depth (3.44%) and severity (1.12%) of poverty, were also higher among non-adopters of yam minisetts technology than those of the adopters of yam minisetts technology, revealing a high degree of income shortfall below the poverty line and a high degree of inequality among the poor. These findings conform to the work of Ayinde in his study on analysis of poverty level among farmers in Ogun State, Nigeria [19].

Click to enlarge

Impact of Yam Minisetts Technology on Poverty Alleviation among Yam-Based Farming Households

The unconditional summary statistics above suggest that agricultural technology may have a role in improving household well-being, but because adoption is endogenous, a simple comparison of the welfare indicators of adopters and non- adopters has no causal interpretation. That is, the above differences may not be the result of yam minisett technology adoption, but instead may be due to other factors, such as differences in household characteristics and the endowments mentioned earlier. To measure the impact of adoption, it is necessary to take into account the fact that individuals who adopt improved varieties might have achieved a higher level of welfare even had they not adopted. Therefore, multivariate analysis was applied to test the impact of yam minisett technology adoption on household welfare.

Determinants of Farmers’ Adoption of Yam Minisetts Technology

The results from the probit analysis of yam minisetts technology adoption and the variables used in the matching procedures are reported in Table 3. The standard errors in the result, indicates that adoption of yam minisett technology is clearly (nonlinearly) related to age at 1% significant level, and that the educational level appears not to be significant. Extension agents may be proxies for access to information. It is significant in explaining the variation in the adoption decision. Agricultural extension is the system of learning and building human capital of farmers through the provision of information and demonstrations, exposing farmers to technologies which can increase agricultural productivity and, in turn, income and welfare. Farmers who are frequently visited by extension agents tend to be more progressive and more likely to experiment with yam minisett technology. This positive effect of farmer technology awareness variable is consistent with Shiferaw, et al. for improved pigeonpea varieties in Tanzania and also with Kristjanson, et al. for cowpea varieties, while Kaliba, et al. also reported the same result for maize varieties and Gebreselassie and Sanders for sorghum in Ethiopia [20, 21, 22, 23]. Furthermore, household head income and access to credit also affects the propensity score, while household size, sex, primary occupation and farming experience does not seem to explain adoption. The adoption regression suggests the importance of age, household head income and access to income in influencing yam minisetts technology adoption.

Common Support Region: Estimation of the Propensity Score

After estimating values of yam minisetts technology adoption for adopters and non-adopters, the second step was estimating the propensity score by imposing a common support condition. Among adopters (treatment) and non adopters (control), the predicted propensity score ranged from 0.0638 to 0.9973, with a mean of 0.5428. The common support condition is satisfied. Thus, the common support assumption is satisfied in the region of [0.0638, 0.9973]. In other words, households whose estimated propensity scores are less than 0.0638 and larger than 0.9973 are not considered for the matching exercise. As a result of this restriction, 40 adopters and 85 non-adopters were discarded from the analysis. The density distributions of the propensity scores for adopters and non-adopters also support this result where there is a good overlap.

Estimation of Average Adoption Effect (ATT): Matching Algorithms

Table 5 reports the estimates of the average adoption effects estimated by Nearest Neighbor Matching (NNM), Kernel Based Matching (KBM), Radius Matching and Stratification methods. All the results were based on implementation of common support and caliper, so that the distributions of adopters and non-adopters were located in the same domain. As suggested by Rosenbaum and Rubin, we used a caliper size of one-quarter of the standard deviation of the propensity scores. Bootstrap standard errors based on 50 replications are reported [10]. Outcome variable used in the analysis is the natural logarithm of per capita consumption expenditure (hereafter consumption expenditure). The consumption expenditure is transformed into logarithmic because it is very right-skewed. The logarithmic transformation eliminates this skewness. The results indicate that adoption of improved yam minisett technology has a positive and significant effect on consumption expenditure and thus, has the power to lift the poor farmers above the poverty line.

consumption expenditure of similar pairs of households that belong to different technological status (i.e., adopters and non-adopters). The increase in consumption expenditure can help adopters reduce their poverty level. These findings are consistent with recent studies on the impact of modern crop varieties on household welfare. Hossain et al and Mendola in Bangladesh, Janaian et al in India, and Wu et al in China showed that the adoption of improved rice varieties has a significant positive impact on household income and a negative impact on poverty status [6, 24, 25]. Becerril and Abdulai using propensity score matching methods found that improved maize adoption significantly increases per capita expenditure and reduces poverty in Mexico [26]. Kijima et al also showed that NERICA rice adoption reduces poverty without deterioration in income distribution in Uganda [27]. Kassie, et al. using PSM methods found that the adoption of improved groundnut varieties in rural Uganda increase crop income and reduce poverty [27].

The ATT in nearest neighbor matching (NNM), Radius matching and stratification methods were found to be significant at 10%, 5% and 1% respectively. This indicates a positive relationship which increases the consumption expenditure per capita of the yam farming households using the algorithms. This is the average difference in

Estimation of the True Impact of Yam Minisett Technology Adoption on Poverty Status of Yam Farmers Household

The matching procedure applied to the probability of the yam farmers’ household to be poor indicated that adopters were less likely to be poor by about 6.8% - 91.9% across the ATT estimator (Table 6). This is consistent with a priori expectations. As consumption expenditure of yam farmers increases due to the effect of adoption, poverty level is expected to reduce because an increase in consumption expenditure shows that there is an increase in income. In this case, a unit increase in income due to adoption of yam minisett technology would decrease adopters’ propensity to fall below the poverty line. This finding qualitatively agrees with Mendola, and is statistically significant at 5% level [6].