Impact of Farmers Training Centers Based Training on Major Crops Productivity and Households Welfare: the case of Gurage Zone, Central Ethiopia

The purpose of this paper is to identify the impact of Farmers Training Centers Based Training on Major Crops Productivity and Households Welfare in southern Ethiopia. To select the respondent households applied a multi-stage stratifying sampling method. First, 3 districts Were selected purposively. Then, 3 Rural Kebeles from Fully functional Farmers Training Centers and 2 Rural Kebeles from non-functional Farmers Training Centers Were randomly selected. In the third stage, a total of 360 sample household heads (151) modular training graduates and 209 non-graduates) were selected. The interview schedule, focus group discussions, and key informant interviews were data collection tools. The result indicates the frequency of extension contact and farmers Condence in extension service signicantly affects both trained and not trained farmer’s crop productivity. I nd that trained farmers have 124.27% more net cereal crop income per cultivated land size as compared to the counterfactual scenario of non-trained. I nd that it increases the consumption expenditure per adult equivalent of trained farmers by about 204.83% compared to the counterfactual scenario of non-trained. The constraint which is considered by the respondents as a rst-rank and important constraint was the fence problem which accounts for 21.63% of the total respondents. Farmer training center-based modular training could be considered as a key pathway that contributes to the improvement of agricultural production and Welfare of the rural community. The output of this study will give concrete information on intervention strategies to enhance the role of FTC-based training intervention strategies to improve farmers’ livelihood in surrounding rural areas. improved farming and living practices. This shows the role of FTC-based training in improving farmers’ productivity and agricultural transformation from subsistence to market-oriented one, which contributes to the development of the sector and the economy. Particularly, GZBARD (2017) reported FTCs in the study area to provide different training services on compost preparation, manure preparation, use of the improved seed, tillage practice, row planting, irrigation water management practices, cropping calendar, soil water conservation practices, and use of credit, forage husbandry, animal housing, zero-grazing, and fattening. Using and Welfare impacts of modular training in the Gurage of central The study used net cereal crop's income per area covered by cereals to capture productivity and used consumption expenditure per adult equivalent as an indicator of Welfare of households. The study nding indicated that cooperative membership, sex of the household head, family size, farm size, condence in extension service, social responsibility, and Livestock ownership encourage households participation in modular training, while age squared of the household head, education, access to credit, distance from the market center, distance from FTC and access to off and/or non-farm activity negatively affect households participation decision.

and Dercon et al. (2009) on the importance of further inquiries, in this research I investigated the causal effect of FTC-based training on smallholder farmers major crops productivity and Welfare in the study area. In doing so, this study provides insight into determinants of participation to evaluate the inclusiveness of FTCs; evaluate the impact of modular training on the productivity and Welfare of farmers quantitatively, and examine if modular training graduation has a heterogeneous impact on Well-being among its participant that allows understanding how training can be more effective in improving the well-being of the rural community. Thus, this study is lling the information gaps on the issue to recommend further intervention.

Description of the study area
Gurage Zone is one of the zones found in the Ethiopian Southern Nations, Nationalities, and Peoples Region (SNNPR). It is located in the Eastern part of central Ethiopia; it is bordered on the south by Hadiya on the Ist, North, and East by Oromia Region, Yem on the southeast, and the Southeast by Site Zone (Fig. 1).
Its land area is estimated at 593,200 hectares. The zone is divided into 13 woredas and two town administrations.

Sampling technique and sample size
The study was applied a multi-stage stratifying sampling method to select sample households. In the rst stage, three districts, namely Abeshge, Cheha, and Eza of Gurage zone Were purposively selected based on their potential in the production of crops, availability of established and functional FTCs. In the second stage, teff & wheat-producing Pas in the selected districts Were strati ed into two: Pas with FTCs delivering modular training and Pas without FTCs delivering modular training. Then, 3 PAs from Fully functional FTCs and 2 PAs from PAs without functional FTCs Were randomly selected. In the third stage, a total of 360 sample household heads (151modular training graduates and 209 non-graduates) Were selected.

Data Set and Methods of Collection
For this study, both primary and secondary data that are quantitative and qualitative Were used. Primary data on Welfare status indicators, major crop yield, types, and quantity of inputs used in the production of major crops, and other demographic, social, economic, institutional, and ecological factors will be collected. This data was collected using an interview schedule through enumerators and the researchers. The enumerators Were trained on how to conduct the interview questions and how to approach household heads during the interview. To revise and modify the questions Were for the nal data collection, a pre-test was conducted on randomly selected respondents living in the sample Kebeles. In addition to this, FGD, and key informant interviews Were employed to supplement the research nding with qualitative information. Secondary data on aggregate major crops production, aggregate inputs used, the status of improved variety use and major constraints in operating FTCs in the study area, etc., Were obtained from various sources such as records, reports, etc, of both governmental and non-governmental organizations such as both Gurage zone and sample districts o ce of agriculture and rural development, nance and economic o ce, regional o ces, etc.

Methods of Data Analysis
To analyze the data to be collected, the study was employed descriptive statistics, inferential statistics, and econometric models.
Descriptive and Inferential statistics The study was employed descriptive statistics such as mean, percentage, frequency, and standard deviation to describe farmers' cereal crops production status, Welfare indicator, input usage, constraints facing FTCs, and different demographic, social, economic, institutional, and ecological factors of the farm households in the study area. Besides, inferential statistics such as independent t-test Were used to compare train farm households and nontrain households in terms of different characteristics.

Econometric models
To address the objectives, this study was employed stochastic production frontier models (to construct productivity outcome variables) and impact evaluation models (Endogenous switching regression models). Estimation of the impact of program intervention on agricultural productivity and Welfare of the farm households based on nonexperimental observations is not trivial because of the need to identify the counterfactual situation had they not had to get involved in the program (Lajqi et al, 2017). In experimental studies, this problem is addressed by randomly assigning farmers to the treatment and control group, where the outcome variable observed on the control farmers is statistically representative of what would have occurred without participation for treated farmers.
However, farmers are not randomly distributed to the two groups (treatment and control), but rather farmers make their own choices or are systematically selected based on their tendency to participate in the program. Thus, in the absence of random assignments, selection bias may persist as observed and unobserved characteristics of individuals may affect the likelihood of receiving treatments as Well as outcome indicators (Wossen et al, 2017). Failure to account for this potential selection bias could lead to inconsistent estimates of the impact of program intervention on the outcome variables (Lajqiet al, 2017).
In this study, ESR econometric models Are estimated to address the objectives and to control selection and endogeneity bias in the estimation of the impact of the modular training on farmers' productivity and Well-being.
The study also employed endogenous switching regression that accounts for both observed and unobserved sources of bias (Lokshin and Sajaia, 2004;Shiferaw et al., 2014;Ma &Abdulai, 2016;. The ESR approach addresses this endogeneity problem by estimating the selection and outcome equations simultaneously using the full information maximum likelihood (FIML) (Lokshin and Sajaia, 2004;Ma &Abdulai, 2016;. But, it has its limitations and the limitations of switching regression (e.g. tri-variety normal distribution and lack of exclusion restriction) are acknowledged.

Speci cation of the models
Page 5/22 The ESR framework follows two stages. The rst stage is an estimation of the selection equation, the decision to participate in the program, which is estimated using a probit model (Equation 1), and in the second stage, an Ordinary Least Squares (OLS) regression (Equation 2) with selectivity correction is used to examine the relationship between the outcome variable and a set of explanatory variables conditional on the participation decision (Shiferaw et al, 2014;. ESR is speci ed as: where T is a binary 0 or 1 dummy variable; T= 1if the household is graduated (participant) and T=0 otherwise, is a vector of parameters to be estimated, Z is a vector that represents household-and farm-level characteristics, and ε is the random error term.
Where 1 and 2 represent Welfare outcome variables such as major crop e ciency and consumption per capita, 1 and 2 are vectors of exogenous covariates, 1 and 2 are vectors of parameters; and 1 and 2 are random disturbance terms.

De nition of Variables
Outcome variables Cereal crops productivity: To measure the impact of modular training on the cereal crops productivity; the study considered the net income from cereal crops per cultivated land as our outcome variable. Following the approaches used by Tesfaye et al (2019) and Matsumoto and Yamano (2010), this study used household's net income from crops per land covered by cereal crops.
Households Welfare: to represent the Welfare of the households the study used consumption expenditure per adult equivalent unit (AEU) which is regarded as a good proxy variable for the outcome variable.

Selection variable
Modular training Graduation: it is a dummy variable that takes 1 if the farmer is graduated with a green certi cate (trained) and 0 otherwise.

Result And Discussion
Demographic and socio-economic characteristics of households Tables 2 and 3 presents descriptive statistics result on different farmers demographic, social, economic and institutional characteristics.   *, **, and *** indicates signi cant at 10, 5, and 1% probability level respectively.
The result revealed that 82.22% of the respondents Were male-headed farmers, implying that the majority of the sampled farmers Were male-headed. More speci cally, 77.99% and 88.07% of not trained and trained, respectively, Were male-headed households. Besides, 63.88% of the sampled household heads had leadership responsibility in the community, and the proportion of not trained farmers engagement in social responsibility was loIr (60.7%) compared to trained farmers proportion, which is 68.2%.
Further, the proportion of trained farmers' membership in cooperatives (43.7%) was higher than the proportion of total sampled farmers (39.7%) and not trained (36.8%) who Were members of cooperatives. Overall, it implies that the majority of the sampled households Were not a member of agricultural cooperatives. Furthermore, 37.22% of the sampled households had access to credit services. Speci cally, 37.74% and 36.84% of trained and not trained farmers, respectively, had access to credit services. It also revealed that 44.16 % of the households had access to offfarm and/or non-farm income-earning opportunities, which implies that more than half of the farmers had no access to off-farm and/or non-farm opportunities. In this scenario, the off/non-farm activities participation status of not trained (45.9%) was higher than the trained 41.7% farmer's participation status (Table 2).
Moreover, the study con rmed that 90.27% of the households reported that they had con dence in the extension service provided to them, and 86.6% and 95.3% of not trained and trained farmers, respectively, reported that they had con dence in the extension service. These imply that the majority of the sampled households had con dence in the extension service provided in the study area.
Regarding continuous variables, the result should that the average age of the sample farmers was 45.76 years. More speci cally, trained respondents and not trained respondents have an average age of 46.1 and 45.5 respectively. Besides, the average year of schooling of the sampled farmers was 6.9 years, with mean years of schooling of 7.1 and 6.7 for not-trained and trained farmers, respectively. It also indicates that the mean family size of the trained households' was 5. 8 members, which was higher than the average family size of sampled farmers (5.7 members) and not-trained farmers (5.6 members). Moreover, the mean experience of households in FTC experience of sampled farmers, not trained and trained was 7.4, 7.3, and 7.5 years, respectively.
It also should that the average farm size of the trained households was 3.23 hectares, and it was higher than the mean farm size of not trained farmers (2.97hectare) and sampled households (3.09 hectares). The mean distance of households from the nearest market and farmers' training center was 60.1 and 25.43 kilometers, respectively. The result also pointed out that the average frequency of extension contact of sampled households with extension agents in the study area was 2.186 times per month. It was loIr than the mean frequency of extension contact trained farmers (2.264 visits) and higher than not trained farmers average contacts (2.129 visits). It also revealed that the mean livestock holding of the sampled households, not trained and trained was 7.166, 7.339, and 6.925 tropical livestock units, respectively. The average amount of total income of respondents was 35029.72 birr. This average was higher than the mean amount of total income of not trained 34925.31and loIr than the mean applied by trained farmers 35174.24 birrs (Table 3).
In general, the Descriptive and inferential statistics result in continuous explanatory variables of trained and nottrained households. The result shows that statistically there is no signi cant difference between the two groups in terms of age, education level of household, spouse education level, family size, farm size, FTC experience, total income, and frequency of extension contacts. Compared to not-trained households, trained households had larger income from crops. Similarly, compared to not trained households, trained households encountered far distance to market.
Determinants of farmer's participation in FTC modular training in the study area The Probit model was estimated the log-likelihood of the explanatory variables that in uence the farmer's participation in FTC modular training, the level of signi cance and true relationship of this in uence was also appropriately estimated and indicated by the model.
The nding reveals that an increase in the level of any of the explanatory variables with positive sign Cooperative membership, Sex of the household head, family size, farm size, Con dence in extension service, Social responsibility, and Livestock ownership, in this case, has a positive effect on the farmer's participation in FTC modular training, whereas those explanatory variables with a negative sign, Age squared of the household head, Education, Access to credit, Distance from the market center, Distance from FTC and Access to off and/or non-farm activity exert a negative relationship on farmer's participation in FTC modular training. Among these variables Household size in AE, Cooperative membership, and Con dence in extension service signi cant at 5 percent and 1 percent signi cant level. This indicates that it is a strong factor considered for farmer participation in FTC modular Training although its coe cient being positive is contrary to a priori expectation because it is expected to be contributing negatively to participation, the positive sign could be attributed to more emphasis being placed on FTC modular Training (Table 4).  Tables 5 and 6. The Wald tests presented in Tables 7 and 8 con rm the joint signi cance of the error correlation coe cients in the selection and outcome equations.

Factors Affecting Crop Productivity and Household Welfare
The nding of the study indicates that Access to credit, household head education, distance from the market center, distance from FTC, livestock ownership, Experience in FTC, Soil fertility status, Spouse education, farm size, signi cantly affect Crop Productivity (Table 5).
The nding of the study indicates that cooperative membership signi cantly affects crop productivity of trained and not trained farmers (Table 5). Membership in cooperatives increased trained and not trained farmer's crop productivity. This could be due to cooperative organizations engage in the provision of inputs and collecting nal outputs produced, which improves farmers' e ciency in production.
Crop productivity, of both trained and not trained farmers, was loIr for households in the loIr and higher age of years, whereas it was higher for households in middle years of age. This shows that as the age of the farmers increases rst crop productivity increases and then starts declining. This nding supports the ndings of (Chauke P et al., 2014), who argued that young households earn more income from crop production than older ones.
The nding of the study indicates that Frequency of extension contact and farmers Con dence in extension service signi cantly affect both trained and not trained farmer's crop productivity.
It also should that owning more TLU of livestock increases Trained and not trained farmers' crop productivity. This nding is in line with (Aba t, J. and Kim K., 2014) and suggested that more TLU of livestock ownership helps to cultivate more land through the provision of traction post, manure, and provides transport service to market the produce. Also, the nding indicated that household size positively affects adopter farmers' income from crop production. This could be due to households with larger family size Besides, the nding indicated that household size positively affects both trained and not trained farmers' crop productivity. This could be due to households with larger family sizes accept high risks in experimenting with technologies to attain the maximum possible production from their farm to meet family demands .
Education of the farmers positively affected crop productivity of modular trained farmers. This is because educated farmers access information (both print and electronic) on the bene ts and potential risks of improved varieties much easier (Chauke P et al., 2014). Furthermore, the nding of the study pointed out that more experience in FTC enhanced crop productivity of trained farmers. This is due to experience in FTC enables farmers to better understand and exploit the means of increasing production and obtaining more yield from crop production. The estimated ESR model result of the impact of modular training on household Welfare pointed out that cooperative membership, frequency of extension agent contact, and con dence in extension service Were signi cantly affected the Welfare of both trained and not trained farmers. Besides, it gured out that age squared, farm size, distance from the market center, access to credit, access to off and/non-farm activity, frequency of extension contact, social responsibility, and spouse education signi cantly affected modular trained farmers Welfare (Table 6). The welfare of trained farmers was loIr for households in the loIr and higher age of years, whereas it was higher for households in middle years of age. This shows that as the age of the farmers increases rst Welfare increases and then starts declining. The nding of the study also pointed out that more experience in modular training enhanced the Welfare of trained and not trained farmers. The Welfare of female-headed households was loIr than male-headed households among not-trained farmers.
It is similar to (Baten and Khan's, 2010) nding, which should that female-headed households have less access to valuable resources and earn less farm income than males. Farm size positively affected the Welfare trained farmers in the study area. This is because farmers with large farm sizes are more likely to have more opportunities to adopt modern technologies and earn more income to have Welfare (Ahmem M. et al., 2017. and, Chandio, A. A. andYuansheng J., 2018). It also indicated that distance from the market center positively determined Welfare among trained farmers. Besides, the result of the study revealed that access to off-farm and/or non-farm activities increases trained farmers Welfare.

Impact of Modular Training on Productivity in the Study Area
The Wald test presented in Table 5 con rms the joint signi cance of the error correlation coe cient in the selection and outcome equation. A signi cant correlation coe cient of the selection equation and some of the outcome equation indicates the presence of self-selection in the modular training participation. This also suggests that access to modular FTC training had a signi cant impact on net income from cereal crops per cultivated land (Table 7). I nd that modular training has a positive impact on net income from cereal crops per cultivated land at p < 0.01. The nding indicates that participation in modular training increases net cereal crop income per cultivated land size by about birr 795.33 per hectare for training compared to a counterfactual scenario of non-trained. I nd that it increases the net cereal crop income per cultivated land size of trained farmers by about 124.27% compared to the counterfactual scenario of non-trained. Furthermore, modular training had increased non-trained households' net cereal crop income per cultivated land size by birr 80.65 had they decided to train.
The positive base heterogeneity effect implies that trained households have higher net cereal crop income per cultivated land size not possible due to their decision to participate in the training, but possibly due to unobservable. Adjusting for the potential heterogeneity in the sample, there is evidence that households who decided to participate in the training tend to have bene ts higher than the average if they participate (Di Falco et al., 2011). The positive transitional heterogeneity effect also indicates that the effect is loIr for non-trained households had they decided to use it.

Impact of modular training on households Welfare in the study area
The full information maximum likelihood (FIML) estimates of the ESR model are presented in Tables 6 and 8. The overall model is signi cant at 1%. The Wald tests con rm the joint signi cance of the error correlation coe cients in the selection and outcome equations (Table 6). Signi cant correlation coe cients of the selection equation and the outcome equation for households with modular training indicate the presence of self-selection in the participation of the training. This also suggests that participation in modular FTC training had a signi cant impact on the corresponding outcome among trained, and trained households would have gained greater Welfare from the training than non-trained had non-trained chosen to participate in the training (Wordofa and Sassi, 2018). I nd that modular training has a positive impact on consumption expenditure per adult equivalent at p < 0.01. The nding indicates that modular training increases consumption expenditure by about birr 61.60042per adult equivalent for training compared to a counterfactual scenario of not to train. I nd that it increases the consumption expenditure per adult equivalent of trained farmers by about 204.83% compared to the counterfactual scenario of non-trained. Furthermore, modular training would have increased nontrained farmers' consumption expenditure per adult equivalent by 242.8% had they decided to train. A positive effect on consumption expenditure is expected since FTC modular training would help households increase their production and productivity through capacity building and provision of technical information related to different agronomic, animal health and husbandry, and natural resource conservation. Besides, relaxing risk-aversion to postharvest loss and encouraging farmer's production of diverse crops, enhances households' agricultural production (Oluwatoba et al, 2016). The result is in agreement with Wordofa and Sassi (2018) that report a positive link between modular training and household Welfare.
The positive base heterogeneity effect implies that trained households have higher Welfare not possible due to their decision to participate in the training, but possibly due to unobservables. Adjusting for the potential heterogeneity in the sample, there is evidence that households who decided to participate in the training tend to have bene ts higher than the average irrespective of participation, but they are better off participating than not participating (Di Falco et al, 2011). Table 8 presents the expected values of the various outcomes under the actual and counterfactual conditions and the resulting treatment effects.
Major challenges facing the effective functioning of FTCs in the study area To identify the lists of constraints that impede the effective functioning of FTCs extension agents', supervisors', experts' and team leaders' suggestions, formal and informal discussions with farmers and extension personnel, and the SWOT analysis Were considered.
The constraints list included ten items (Fig. 2) and among these, the constraint which is considered by the respondents as a rst-rank and important constraint was the fence problem which accounts for 21.63%, for the second more than two problems which account for 15.73%, for the third lack of awareness about FTC by farmers accounts 12.08% and the last rank was a budget problem which accounts 5.05%.
Fence problem the respondents stated the FTC demonstration site didn't have a fence. Due to this all demonstrated activity was damaged by wild and tame animals. FTC which has enough demonstration elds (3-5ha) will have a better status and good output. If not the result will be the opposite. It is measured in terms of effective functioning (Fisseha T., 2009).

Budget problem
It has been reported by numerous Respondents that due to inadequate nancial investment to extension, the recruitment and retention of competent extension personnel, plus adequate provision for in-service training of staff and training of farmers, transportation, housing, and the conduct of extension programs cannot properly be carried out (Bahal, 2004). Therefore, an FTC which has a better status in budget allocation per year will have better status and good performance otherwise, it will be the reverse. The budget is measured in terms of Birr allocated for the FTCs per annum.

Lack of awareness about FTC by farmers
This re ects the involvement of the community from inception up to the evaluation of FTCs. In an FTC where the community participation is high the status of FTCs will be high (Fisseha T., 2009). In poor community participation, the status of FTCs will be poor. The participation is measured on an ordinal scale as poor, good, very good, excellent.
More than two problems(Infrastructure facilities, Equipment/material): the study FGD group member result revealed that the FTC has a shortage of different materials such as seats for trainees, chairs, tables, shelves, eld equipment, and other facilities that are necessary for the teaching-learning process. An FTC which has ful lled different internal facilities and eld equipment will have better status. Infrastructure includes different buildings and services found in the FTC like classrooms, o ces, residence, exhibition center, workshop, clinic, telecenter, etc. An FTC which has different buildings and services will have a better status and output. It is measured by the number of these services.

Conclusion
FTC modular training could be considered as a key pathway that contributes to the improvement of agricultural production and Welfare of the rural community. Using data collected from 356 households, the study analyzed cereal crops productivity and Welfare impacts of modular training in the Gurage Zone of central Ethiopia. The study used net cereal crop's income per area covered by cereals to capture productivity and used consumption expenditure per adult equivalent as an indicator of Welfare of households. The study nding indicated that cooperative membership, sex of the household head, family size, farm size, con dence in extension service, social responsibility, and Livestock ownership encourage households participation in modular training, while age squared of the household head, education, access to credit, distance from the market center, distance from FTC and access to off and/or non-farm activity negatively affect households participation decision.
Regarding impact evaluation, modular training has a positive impact on both productivities as Well as the Welfare of farmers. The positive impact implies that the modular training provides important technical and practical knowledge on agronomic practices that improve farmers' e ciency of production and enable them to earn more income from the allotted land. The more income farmers earn per cultivated land size, the more they spend on consumable items. Thus, it improves the Welfare of households by improving consumption expenditure per adult equivalent. Besides, the study evidence con rmed that fence problems, poor infrastructure, constraint in improved technology, and lack of awareness about FTC by farmers Are the major barriers to the effective functioning of FTCs.
Based on the evidence obtained from this nding, there is a need for urgent action aimed at addressing the need for improving household's access to FTC modular training to enhance their productivity and well-being in the study area. Policymakers need to consider the role of access to public infrastructure (cooperative membership and market center), extension (con dence on extension service), and education of the household head, household size, and livestock ownership in enabling the household's decision to participate in modular training. Improving middle age household's awareness, enhancing awareness of off and/or non-farm activity participating households should be a major part of any effort aimed at promoting household participation in modular training.
In terms of the impact of training on productivity, although a positive and signi cant impact was found, there is a need to strengthen and support farmer's participation in the training programs. The FTC-based modular training needs to encourage farmers to produce more cereal crops by providing training and production inputs. Moreover, there should be an increased effort towards the promotion of the provision of modular training as it improves the overall Welfare of cereal crop producer households.
Wossen T, Abdoulaye T, Alene A, Haile MG, Feleke S, Olanrewaju A, Manyong V (2017) Impacts of extension access and cooperative membership on technology adoption and household Welfare. J Rural Stud 54:223-233. Major constraints of FTC around the study area Source: own survey result, 2021