1. To identify the certified chickpea seeds that are mostly preferred by smallholder farmers;
2. To assess the determinants of farm-level farmers’ participation decisions in the purchase of certified chickpea seeds in the central highlands of Oromia, Ethiopia;
3. To examine the determinants that influence the quantity of certified chickpea seed purchased by farmers in the central highlands of Oromia, Ethiopia.

The article’s conceptual framework aligns with established theories in agricultural development and market equilibrium (Ayoola 2023). The high input pay-off model assumes that farmers possess efficient resource allocation skills and respond to economic incentives, but face significant technical and economic constraints. Consequently, it proposes the need for support in the form of improved seeds, technical inputs, and fair output prices, which serve as the basis for price intervention policies. Furthermore, the market failure theory emphasizes the inadequacies of the agricultural-input market, necessitating intervention policies to address these deficiencies (Lipsey and Lancaster 1956).

The above literature indicates that the agricultural seed policy process encompasses two crucial aspects: supply and demand. On the supply side, policy authorities such as policymakers, policy service providers, policy analysts, and consultants play a vital role in shaping government actions and inactions within the seed market. On the demand side, stakeholders including farmers, seed producers, and seed marketers directly or indirectly experience the consequences of government interventions in the seed system. The interaction between these two sets of stakeholders in the agricultural seed marketplace, the outcomes of public policy interventions, are determined by the relative influence exerted by each side on industry-related issues. However, without a shared reference point for both sets of actors, a gap emerges in the policy marketplace for seeds, significantly impacting the overall performance of the seed industry. Therefore, establishing a demand-driven seed industry in Ethiopia requires an understanding of the demand level and the factors influencing the demand for agricultural seeds. This knowledge is crucial for enhancing the efficiency of the seed industry in the country as a whole and benefiting the stakeholder community.

The study was conducted in the Eastern and Southwestern Shewa Zones of the Oromia Regional State of Ethiopia during the 2022 crop season. These study areas are among the major chickpea-producing areas in the country where certified chickpea has been practiced by farmers. The Zones represent the high potential areas based on the area of production and yield of the chickpea crop (CSA 2022). Thus, the emphasis was given to these rural highlands of Ethiopia especially the Gimbichu, Becho, Lume, and Adea districts of Oromia, Ethiopia due to their agroecological suitability for chickpea production.

The sample size and the sample selection process should ensure the representativeness of the population. Sample size determination has a scientific approach. In this study, to determine the sample size, different factors such as research cost, time, human resources, accessibility, and availability of transport facilities were taken into consideration. The study used multi-stage procedures to collect the primary data. In the first stage, two zones were selected purposively based on the production potential of the crop. In the second stage, from the selected zones, 4 districts were randomly selected. In the third stage, 9 kebeles were randomly selected from the chosen districts. Finally, 186 farm households were interviewed using random sampling techniques. The study used primary cross-sectional farm household data. Data on farmer characteristics, chickpea farming practices, and environmental and institutional characteristics pertaining to the households were collected from the farm households. In addition to primary data, secondary data were collected from the Ministry of Agriculture, CSA, published documents, and other relevant institutions to corroborate the primary data. A structured questionnaire was designed to collect primary data, and experienced enumerators were trained to assist with the data collection process.

Following the completion of data collection, the responses were coded and entered into the SPSS version 26 software program and converted into STATA 14 software for statistical analysis. Descriptive statistical analyses such as mean, percentages, and standard deviations were used to understand farm households’ demographic characteristics, resource ownership, and institutional service, production characteristics and farm input use. The determinants of farm-level participation decision and quantity of certified chickpea seeds purchase were analyzed using a double hurdle model.

This section introduces the empirical model for estimating the factors that influence farm-level decisions regarding the purchase of certified chickpea seeds and the quantity of certified chickpea seeds bought in the central highlands of Ethiopia. The involvement of chickpea farmers in certified seed purchase activities is expected to be influenced by socioeconomic factors, institutional factors, as well as the demographic characteristics of the farmers. These factors were used collectively to determine whether a specific chickpea farmer finds it desirable to engage in certified seed purchase and, if so, the extent of their participation (i.e., the quantity of certified seed purchased). One of the objectives of this study is to analyze the motivating factors behind the participation of chickpea farmers in certified chickpea seed purchases. It is important to note that some chickpea farmers may choose not to participate in certified chickpea purchase activities, resulting in a portion of the dependent variable being zero. (Elhorst 1993) highlighted the difficulty in estimating models for farm household investments due to the prevalence of zero values in the dataset. The presence of only positive values in dependent variables can introduce sample selection bias, and using simple linear regression with ordinary least squares (OLS) estimation may result in biased and inconsistent estimates (Elhorst 1993; Worku and Mekonnen 2012).

Greene et al. (2008) suggested that it is necessary to use an approach that can incorporate both discrete and continuous components. To address the statistical issue associated with the dependent variable having a significant number of zero values, the conventional regression models used a binary dependent variable to assess the statistical issues related to the dependent variable. Numerous empirical researchers have examined factors influencing market participation decisions using discrete choice models: probit or logit estimators and evaluated the probability of a household’s decision by assuming a logistic or normal distribution (Aman et al. 2014; Bekele and Alemu 2015). Dedah et al. (2010) pointed out that the probit/logit approaches are useful tools to provide information on how different characteristics of farmers influence the probability of investment in farm activities. However, these models fail to provide information about the level of participation in the market.

The Tobin (1958) model, also called the censored regression model can handle this estimation problem and allows for the analysis of the factors affecting the joint decision. According to this author, a censored regression model is required for a dependent variable that has a considerable number of observations with zero value. Many researchers have applied this model because it has an advantage over models such as Linear Probability Models, Logit, and Probit in that it indicates both the probability of willingness to participate in the market and the intensity of participation (Bekele and Alemu 2015; Obayelu et al. 2016). However, the Tobit model is very restrictive in its parameterization and there are limitations concerning the use of this model when the proportion of zero values for the dependent variable is significant. In the Tobit model, the censored variable (participation) and expected value conditional on the quantity of certified seed purchases are estimated by the same factors. This model considers only the dependent variable to be censored at zero and ignores the source of zero observations (Greene 2002; Newman 2004). To overcome the weakness of the Tobit model,the Heckman (2013) model was designed to deal with estimation bias associated with censoring. The author made note of the selection bias that occurs from estimation on a chosen subsample. To address the issue caused by the zero observations produced by the nonparticipation decision, he suggests using the two-stage estimating technique (sometimes referred to as the Heckit model). By performing a complete sample Probit estimation in the first stage and a corrected self-selection estimation in the second stage, the Heckit model eliminates the selection bias. The model assumes that there are different sets of independent variables at play in these two stages and that there are no zero observations in the second stage.

Cragg (1971) proposed the double hurdle model, which generalizes the Tobit model by providing a second hurdle that must be passed before observing any positive values. Both models allow the possibility of estimating the first and second-stage equations using different sets of explanatory variables. The difference is that the double-hurdle model permits potential zero values in the second stage. The first hurdle refers to the participation decision/certified seed purchase decision, and the second hurdle refers to the level of participation decision of certified seed purchased by the farm households. The second stage in the double hurdle model is open to potential zero values, and zero observations on the dependent can be either attributed to corner solution or non-participation when employing a probit estimator to stimulate the participation decision.

Additionally, the decision to engage in certified chickpea seed purchase decisions and the quantity of certified seed purchase activities can be estimated using the double Hurdle model. Therefore, a separate stochastic process can be used to model the probability of participation and the level of participation. As mentioned, the first stage of the double-hurdle model or the first hurdle in the current analysis represents the decision by chickpea farmers whether or not to purchase certified seeds of chickpea for planting during the 2021 cropping calendar, while the second stage represents the desired quantity of certified seed purchase. In the double-hurdle model, zero observations have the potential for a positive level of certified seed purchase. Therefore, in this research, a double-hurdle model is presented as an empirical framework to examine the effects of various factors on both participation in certified chickpea seed purchase during the production season and the intensity of certified seed purchase. Following Yen and Jones (1996) the specification of the double-hurdle model can be expressed as:

Stage 1: certified seed purchase decision.

D_i ^* = X_1i β₁ + ϵ_i ; ϵ_i ~N (0,1) (1)

$D_i=\left\{\begin{array}{c}1\text{if}{D}_i^{*}>0\\ 0\text{otherwise}\end{array}\right.$ (2)

Where D_i^* is a latent variable indicating whether or not the individual participates in certified seed purchase during the production season, β₁ is a vector of parameters to be estimated, X_1i is a vector of observed independent covariates that explain an individual’s decision, D_i is a binary variable that takes the value 1 for a farm household i of certified seed purchase and 0 otherwise, and ϵ_i is an unobserved error term capturing all other factors.

Stage 2: The intensity of certified seed purchase

Z_i ^* = X_2i β₂ + ω_i ~N (0,σ²) (3)

$Z_i=\left\{\begin{array}{c}1\text{if}{D}_i=1Z_i^{*}>0\\ 0\text{if}{D}_i=0\end{array}\right.$ (4)

Finally, the observed quantity of certified seed purchased is determined as

Y_i = D_i ∙ Z_i (5)

In this specification, a positive quantity of certified seed purchase Y_i is observed if D_i^* and Z_i^* > 0. This illustrates the double-hurdle element of the model. D_i^* is a latent endogenous variable representing the decision to participate in the purchase of certified chickpea by farmers i, Z_i^* is a latent variable representing the quantity purchased by farmer i, Y_i is the observed quantity of certified chickpea purchased by farmer i, X_1i is a set of farmers characteristics that influence the farmers’ decision to participate in the certified chickpea seeds purchase decision, X_2i is a vector of socioeconomic and institutional characteristics of farmers that affect the quantity of certified chickpea. β₁ and β₂ are vectors of the estimable parameters. In this formulation, (X_1i ; X_2i) may contain the same common explanatory variables, although their corresponding effects on the two hurdle equations might be quite different. ϵ_i is normalized to 1 since the outcome of the first hurdle is binary. Both error terms, ϵ_i and ω_i are assumed to be normally and independently distributed, which implies that there is no relationship between the two stages of a decision and can be written as:

$\left(\genfrac{}{}{0ex}{}{{ϵ}_i}{{\omega }_i}\right)~N\left[\left(\genfrac{}{}{0ex}{}{0}{0}\right),\left(\begin{array}{cc}1& 0\\ 0& {\sigma }^2\end{array}\right)\right]$ (6)

As in the Tobit and Heck model cases, the independent double-hurdle model is estimated using maximum likelihood techniques with the log-likelihood given as follows,

$\ln L=\sum _0\ln ⌈1-\phi \left(X_{1i}{\beta }_1\right)\phi \left(\frac{X_{2i}{\beta }_2}{\sigma }\right)⌉+\sum _{+}\ln ⌈\phi \left(X_{1i}{\beta }_1\right)\phi \frac{1}{\sigma }\left(\frac{Z_i-X_{2i}{\beta }_2}{\sigma }\right)⌉$ (7)

The first term on the right-hand side denotes the summation over the zero observations in the sample. It indicates that zero observations are affected by both participation and level of participation decisions. The additional term in equation (7) contributes to the effect of possible zero values in the second stage decision in the double-hurdle model. The first term captures the possibility of observing zero values in the second stage decision and thus indicating the second stage is represented like a Tobit model. The second term on the right-hand side indicates summation over the positive observations; this term expresses the conditional probability distribution and density function coming from the censoring rule and observed positive values (Fabiosa 2006).

Results show that the average ages of the respondents in Gimbichu, Becho Lume, and Adea districts are 45.3, 44, 46, and 42.7 years respectively. The overall average age across all study districts is 44.5 years, with the youngest participant being 21 years old and the oldest being 79 years old.This suggests that a significant portion of the households involved in certified chickpea seed purchase are still in their productive years, possessing extensive experience in certified chickpea procurement. These findings align with the research conducted by Tebeka et al. (2020) which indicates that the age of the household head has a notable influence on the development of knowledge and experience related to seed quality. This knowledge is crucial for farmers who are involved in seed production or rely on formal sources for purchasing seeds.

Female-headed households in rural Ethiopia face numerous challenges in agricultural production and marketing compared to their male-headed counterparts. This is due to the multiple responsibilities that female household heads have, such as collecting firewood, fetching water, childbearing, and managing household tasks. Additionally, they also have to manage farm-related responsibilities, which further adds to their burden. These multiple tasks, combined with limited access to resources and land ownership, as well as economic and social shocks like poverty and food insecurity, contribute to the difficulties faced by female-headed households in the country (NPC 2018). Similarly, the current research findings reveal that the majority of farm household heads (94.6%) are male, with only a small percentage being female. Additionally, it is evident that male farm household heads dominate the purchase of certified chickpea seeds. Therefore, it is crucial to make efforts to ensure that improved inputs, especially certified chickpea seeds, are accessible to both female and male-headed farm households. This will help to secure the seed supply for farming households in the study areas.

Experience shows that a knowledgeable farmer has the ability to use advanced agricultural technologies such as improved seeds and fertilizers. They are also capable of engaging in certified seed farming activities according to the cropping calendar and effectively managing available resources. The research findings reveal that approximately 61.9% of the household heads have completed elementary school and possess basic literacy skills. On the other hand, 23.8% of the surveyed farmers are illiterate, while 14.3% have completed high school in the Gimbichu district. Moreover, in the Becho district, 23.4% are illiterate, 2.1% can read and write, 40.5% have completed elementary school, and 34% have completed junior and high school. Similarly, in the Lume district, 30.2% are illiterate, 24.5% can read and write, and 34% have completed elementary school. In the Adea district, 25%,11.3%, 36.4%, and 27.3% of the sampled farmers are illiterate, can read and write, have completed elementary school, and have completed junior and high school, respectively. The results indicate that the education level of the household head has a significant influence on certified chickpea seed purchases, with a statistical significance of less than (P < 0.01). Moreover, it positively influences the decision-making process regarding the purchase and quantity of certified chickpea seeds. Therefore, the higher the education level of the household head, the more likely the farmer is to make informed decisions regarding certified chickpea seed purchases. This finding is consistent with (Teressa 2019).

The study findings indicate that the mean number of household size per household varied across different districts. Specifically, in Gimbichu, Becho, Lume, and Adea districts, the average household sizes were 5.9, 7, 6.4, and 5.4 persons respectively. Moreover, the average household size for certified chickpea seed purchasers was approximately 6.5 individuals, whereas for non-certified chickpea purchasers, it was 5.5 persons. Presumably, households with larger sizes are more inclined to engage in purchasing certified seeds in order to achieve higher productivity and meet the family’s food consumption and selling needs. Moreover, the presence of a larger household enables the execution of various farming operations that are necessary when utilizing certified seeds. According to the t-test, there is a significant (p < 0.05) disparity in household size between individuals who purchase certified chickpea seeds and those who do not. Consequently, the likelihood of a farmer participating in certified seed purchase is influenced by the size of their household.

The land is the primary factor of production. In Gimbichu, Becho, Lume, and Adea districts, the average landholding of the sampled farmers was approximately 1.3, 1.6, 1.6, and 1.06 hectares respectively, with a minimum size of zero and a maximum size of 4.8 hectares. Farmers who purchase certified chickpea seeds hold an average of 1.5 hectares of farmland, while those who purchase non-certified chickpea seeds hold an average of 0.65 hectares. The t-test analysis reveals a significant difference in landholding between certified and non-certified chickpea seed purchasers at a significance level of (P < 0.05). This indicates that the size of the landholding plays a crucial role in determining the demand for certified chickpea seeds. To address the challenge of land scarcity, young farmers often share land with their parents and relatives after marriage or gain access to farmland through sharecropping and renting. During the cropping season, the study districts recorded rented-in and rented-out land areas of 1.2 hectares and 0.03 hectares respectively.

Based on the research results, it has come to light that 77% of farmers engage in growing different improved seeds acquired from a range of sources including model farmers, farmers cooperatives,the local market, and private seed producers. Nevertheless, it is important to highlight that a mere 38.7% of the farmers surveyed choose certified chickpea seeds with a well-established provenance. Chickpea farmers opt to replace their seeds with new ones due to the superior yield advantage and disease resistance offered by the various chickpea varieties. The study findings indicate that the majority of chickpea grower farmers (70%) acquired information regarding certified chickpea seeds from seed producer cooperatives. On the other hand, the remaining farmers obtained this information from farmers cooperatives, neighboring farmers, development agents, and research centers .

The findings reveal that farmers producing certified chickpea in Gimbichu, Becho, Lume, and Adea districts had an average farm experience of 11.4, 9.4, 9.8, and 9.6 years respectively. Around 39.3% of the farmers sourced the certified seeds from seed cooperatives, and approximately 17.2% of the chickpea farmers engaged in cluster farming to produce the seeds. The farmers expressed that they obtained their certified chickpea seeds from these sources due to their consistent availability, superior quality, and limited availability from other sources. Hence, it is imperative for the government and practitioners to enhance the capabilities of seed producer cooperatives, enabling them to produce high-quality chickpea seeds and expand their involvement in the market. Moreover, 39.3% of the farmers surveyed opted for certified chickpea seed purchase, while the remaining percentage did not engage in such purchases. The certified chickpea seed grown includes Arerti, Ejire, and Habru chickpea varieties. The results indicate that 99% of chickpea farmers who participated in the study purchased supplementary inputs like Karate, Roundup, and Gallant chemicals for effective management of insect pests and weeds in chickpea production. These insecticides and herbicides have proven to be highly efficient in controlling biting, chewing, and sucking insect pests, as well as managing weed growth in the chickpea farmland.

The farm households purchased an average of 20 kg of certified chickpea seed per household. In Gimbichu Becho, Lume, and Adea districts, the quantities of certified chickpea seed purchased were approximately 17.3 kg, 11 kg, 51.8 kg, and 13.8 kg, respectively. According to the one-way ANOVA test, there is a significant difference in the quantity of certified chickpea seed purchased per household among the study districts at a significance level of (P < 0.001). This difference may be attributed to the agroecological variations between the districts. Overall, the results indicate that there are significant socioeconomic differences between certified chickpea seed purchasers and non-certified chickpea seed purchasers, as shown in Table 1.

According to the survey findings, it was found that 69.6% of the farm households included in the sample opted to buy certified chickpea seeds for the 2021/2022 cropping season. Additionally, the survey revealed that a significant majority of respondents (75%) sold their chickpea harvest to traders during the peak harvesting season in January, primarily to meet their household expenses. These results highlight the dual significance of chickpea production, as it not only contributes to food security but also serves as a valuable cash crop in the study districts.

The study findings indicate that a significant proportion of the participants (75%) expressed worries regarding the limited accessibility of certified chickpea seeds. Moreover, an even greater majority of farmers (85.9%) voiced their dissatisfaction with the exorbitant prices of these seeds. The farmers who were surveyed also highlighted that agricultural extension agents primarily prioritized assisting model farmers, thereby neglecting the requirements of farmers with limited resources. Moreover, the findings indicate that the availability of certified chickpea seeds is extremely limited, and unfortunately, the quality of the seeds supplied to farmers was subpar. These seeds exhibited low adaptability,were susceptible to diseases, and were occasionally delivered late. As a result, it is of utmost importance to strengthen the capabilities of seed suppliers in addressing quality concerns related to certified seeds through continuous monitoring and evaluation.

Additionally, the supply of certified seeds should be determined based on environmental suitability and consultation with farmers, taking into consideration disease-resistant varieties. The research findings also revealed that a significant percentage of respondents (66.4%) did not receive any extension services or advice regarding the significance of purchasing certified chickpea seeds. This indicates that more than half of the chickpea producers lacked proper guidance on acquiring certified seeds. This underscores the necessity for the government and development practitioners to tackle the scarcity of certified chickpea seeds by offering support to certified chickpea seed producers and granting subsidies to certified chickpea seed suppliers. Such assistance would enable farmers with limited resources to afford and acquire these seeds and it is vital to prioritize inclusive certified seed extension services in order to enhance chickpea production and productivity.

The majority of smallholder farmers who purchase certified chickpea seeds (75.2%) are members of farmers’ cooperatives/unions. However, a significant portion of these farmers (65.3%) have not received formal training on chickpea production. This highlights the oversight of agricultural cooperatives in recognizing the importance of training for improving farmers’ knowledge and skills in certified chickpea production. Such training is crucial for enhancing the overall production and productivity of this crop. The study also reveals that 60.4% of the sampled farmers obtained certified seed from the office of agriculture, while 19.3% purchased from seed producer cooperatives. Additionally, 17.7% of farmers exchanged seeds with their neighboring farmer. In addition, nearly all farmers (99.5%) purchase supplementary inputs such as pesticides and herbicides in conjunction with certified chickpea seeds. Nevertheless, chickpea remains classified as a subsistence crop, and most farmers cultivate it without the use of fertilizers. This could be attributed to the historical generalization and assumption that the crop can meet all its nutritional needs through natural nitrogen fixation. The research findings are consistent with a previous study conducted by (Dawit and Zewdie 2019).

Farmers had reported that the chickpea varieties that were highly preferred a year prior to the survey period were Arerti, Ejire, and Habru. This preference could be attributed to market demand, availability of improved seeds, and the productivity of these varieties. Approximately 14% of the sampled farmers cultivated the certified Arerti chickpea seed variety, while around 7% grew the Ejire variety and 4% grew the Habru variety. The remaining farmers cultivated local cultivars. The results of the one-way ANOVA test indicated a significant difference in variety preference among the study districts at a level of (P < 0.001). This difference in preference could be attributed to variations in weather and soil conditions across the districts. Consequently, it is crucial to target specific areas for the expansion of new chickpea varieties.

During the survey period, it was observed that the average productivity of Arerti, Ejire, and Habru chickpea varieties on farmers’ farmland was around 23.2, 20.8, and 21.6 quintals per hectare, respectively. These findings show that the production of local chickpea varieties by the majority of farmers in Ethiopia leads to a lower crop productivity compared to its potential yield. The estimated crop yield at the research station is 2.5–3.0 t/ha. Previous findings show that the current average farm yield of chickpea in Ethiopia stands at approximately 55% of the crop’s yield potential. This indicates that there is still significant room for farmers to enhance their yield by adopting certified chickpea seeds and accompanying agricultural practices. Table 2 provides a visual representation of the preferred chickpea varieties as indicated by farmers.

The study’s findings reveal that farmers choose to purchase certified chickpea seeds for various reasons. Approximately 22%, 36%, and 25% of the surveyed farmers consistently opted for certified seeds consistently over the past three years. This suggests a growing trend in the demand for certified chickpea seeds starting from 2020. However, there was a subsequent decrease in their procurement of certified chickpea seed between 2021 and 2022. The process of replacing chickpea seeds is a lengthy one, taking up to 8 years, mainly because there are only a few superior varieties available compared to the ones currently being used. This choice is driven by the desire to increase yield, meet market demands, and ensure resistance against pests. However, a considerable number of farmers show limited interest in periodically replacing chickpea seeds owing to the exorbitant price of certified chickpea seeds. To gain insights into the trends of certified chickpea seed purchases for seed replacement, Table 3 provides a visual representation of the data over the past three years.