Impact of Agricultural Enterprises on Employment Creation and Income Generation in Hadiya Zone, Ethiopia

This study seeks to examine the impact of agricultural enterprises on employment creation and income generation in Hadiya zone, Ethiopia. Data were collected from 383 randomly selected household heads, eighteen FGDs and five key informants’ interviews. Data were analyzed using descriptive statistics and econometric model called propensity score matching. The survey results show that out of 383 sampled household heads 125(32.5%) were members and the remaining 258 (67.5%) were non-members. The propensity score matching result shows that being member in micro and small agricultural enterprises (MSAEs) had significant positive impact on members’ employment creation and income generation. The result revealed that the membership resulted in average increment of household’s annually employment creation by about 4 people (33.98%) and income generation by Birr 12,339.00 (32.57%). The result showed that membership in MSAEs had a significant and positive impact on employment creation and income generation and the impact estimates were found to be insensitive to unobserved selection bias. It is, therefore, essential to expand and strengthen development of agricultural enterprises and the membership of households.

access to modern technology (Admasu, 2012). Similarly, Mulu (2007) found that the average annual growth of the surveyed six major towns in Ethiopia was 9% since start-up and 69% of these enterprises did not grow due to the inadequate formal sources of credit.
In addition, a study by Hailay et al. (2014) reported that most enterprises did not grow and remain at their initial level due to different internal and external factors like, gender of manager, type of sector, managerial skills, amount of initial capital, access to training, access to market, consultancy service, access to premises, access to infrastructures, access to credit, insufficient technology and high cost of input. Enterprises are the main source of rapid economic growth and the basic transformer of the structure of economic system from agriculture to industrialization. These makes enterprises a major area of concern for government and NGOs with the objectives of investing in human capital, employment creation, saving promotion, asset building, income generation and income inequality reduction, import substitution, innovation etc. However, the intense studies in both academic and policy making circles about the impacts of agricultural enterprises on employment creation and income generation were not addressed in the country in general and the study area in particular. Due to this, the study gives high emphasis on the impacts of agricultural enterprises on employment creation and income generation on the basis of annual cross sectional data collected from sample households. Based on the enterprises development strategy 1997 division of enterprises by sector, this study deals with the agricultural sector enterprises engaged in milk production in the study area. The impact of these agricultural enterprises on employment creation and income generation in Hadiya zone is therefore the focus of this study and there are no similar studies in the study area on this topic.

Study Area
The study area Hadiya zone is found in the Southern Nation's Nationalities and People's Region (SNNPR) of Ethiopia. It is located at a distance of 232 km away from the Addis Ababa, capital city of the country, to south and 180 km away from regional capital city, Hawassa to North West. The estimated total area of the zone is 346,958.5 hectares. It is characterized by temperate type of climate with daily temperature ranging from 18 0 c to 27 0 c, and it is located 1900 meters above sea level. It has low to high rainy season for seven months from February to August and for the remaining five months from September to January have dry air condition throughout the year. The total population of the zone as per the national census of 2007 was estimated to be male 769,584 (49.7%) and female 778,262 (50.3%) the total of 1,547,846 hard-working, peace-full, multi-ethnic and religious people are found. It is divided into 10 woredas and two town administrations. Hosanna town is the head quarter of the zone.
Mixed farming, business activities public and private sectors employments are the dominant economic activities in the zone. Farmers in the study area practice mixed farming system, which is mainly characterized by the rearing of different types of livestock like cattle, sheep, and goat and production of multiple agricultural products such as cereals (wheat, teff, maize, barley and bean), fruits and vegetables. The area is specialized in wheat production and its productivity is about 65 quintals per hectare. The area is known as "the basket of wheat /smaller Canada" by the Great Leader Late Prime Minister Meles Zenawi. In addition, some cash crops like khat and coffee are also produced. The geographical location of the study area is given in Figure 1 below. European Journal of Business and Management www.iiste.org ISSN 2222-1905(Paper) ISSN 2222-2839(Online) Vol.12, No.4, 2020 73 Figure 1: Location map of the study area Source: Clipped from EthioGIS I. www.walris.wlrc-eth.org

Sampling Procedure
To evaluate the impact of agricultural enterprises (MSAEs) development on employment creation and income generation, the study used data collected at household level. For evaluation of the impacts of MSAEs development on employment creation and income generation, member and non-member household heads were included in the study based on the following sampling work frame. Accordingly, to select the representative samples from the population, this study were employed multi-stage and combination of different sampling procedures.
In the first step three woredas namely, Lemmo, Analemmo and Misha were randomly selected from ten woredas in the zone. In the second step, a total of two kebeles from Lemmo, two kebeles from Analemmo and two kebeles from Misha were randomly selected. Following this, six kebeles were randomly selected and household heads within the six kebeles were stratified into member and non-member groups. From those six kebeles 125 member household heads and 258 non-member household heads a total of 383 households were selected by taking household head lists of their names from the kebeles.  (2017) To determine the sample size for impact evaluation of MSAEs development on employment creation and income generation, this study used simplified formula provided by Kothari (2004) to determine the required sample size at 95% confidence level, estimated variance in the population 50% and margin of error 5%.  ISSN 2222-1905(Paper) ISSN 2222-2839(Online) Vol.12, No.4, 2020 Where, n is desired sample size; Z is values of standard variant at 95% confidence interval (Z = 1.96). e is desired level of precision; P is the estimated proportion of an attribute present in the population with the value of 0.5 as suggested by Israel (1992) to get the desired minimum sample size of households at 95% confidence level. Accordingly, a sample of 383 household heads were selected from six kebeles using random sampling with probability proportional to size as shown in Table 2.1 above. Qualitative data were also collected by using five key informants' interviews and eighteen focus group discussions (FGDs).

Types of Data and Data Collection Methods
The study used both primary and secondary data collected from various sources. The primary data were collected from the sample member and non-member household heads through structured questionnaire using interview supported by key informants' interview, focus group discussions and personal observation using checklists which are pre-tested prior to their use. Secondary data were obtained from published books and journal articles, as well as unpublished annual reports and records from government offices and other relevant organizations. All data collection processes are completed under close supervision of the researcher.

Data Analysis
The study employed both descriptive statistics and econometric model. The econometric model called Propensity Score Matching (PSM) was employed to evaluate the impacts of participation in MSAEs on employment creation and income generation. Based on Rosenbaum and Rubin (1983), propensity score can be defined as the conditional probability of receiving a treatment given pre-treatment characteristics. Let Yi T and Yi C are the outcome variable for participant (members in MSAEs) and non-participant (non-members in MSAEs) respectively. The difference in outcome between treated and control groups can be seen from the following mathematical equation: Where ATE is Average treatment effect, which is the effect of treatment on the outcome variables is Average outcome of untreated, when he/she would non-participant, or absence of The Average Effect of Treatment on the Treated (ATT) for the sample households is given: The fundamental evaluation problem in evaluation of impact is that it is impossible to observe a person's outcome for with and without treatment at the same time. While the post-intervention outcome   is possible to observe, however, the counterfactual outcome of the th i household when she/he does not use the treatment is not observable in the data. The effectiveness of matching estimators as a feasible estimator for impact evaluation depends on conditional independence assumption (CIA) and assumption of common support (ACS). There are different matching estimators were used in the estimation process of ATT in order to make sure that the results obtained are robust. However, the most commonly applied steps of implementation of propensity score match are discussed below. Nearest neighbor matching: An individual from a comparison group is chosen as a matching partner for a treated individual that is closest in terms of propensity score (Caliendo and Kopeinig, 2008). That is, each person in the treatment group chooses individual(s) with the closest propensity score to them. Nearest neighbor European Journal of Business and Management www.iiste.org ISSN 2222-1905 (Paper) ISSN 2222-2839 (Online) Vol.12, No.4, 2020 75 matching can be done with or without replacement. In the case of with replacement, an untreated individual can serve more than once as a match, where as it is considered only once in the case of without replacement. Nearest neighbor matching with replacement increases the average quality of matching and decreases precision of estimation while the reverse is true in the case of nearest nearest-neighbor without replacement (Caliendo and Kopeinig, 2008). Nearest neighbor with replacement is preferred to without when there are big differences between treated and untreated groups to reduce the risk of bad matching. Kernel matching: In kernel matching, each person in the treatment group is matched to a weighted sum of individuals who have similar propensity score with greatest weight being given to people with closer scores. All treated units are matched with a weighted average of all controls with weights which are inversely proportional to the distance between the propensity scores of treated and controls. The most common approach is to use the normal distribution (with a mean of zero) as a kernel, where the weight attached to a particular comparator is proportional to the frequency of the distribution for the difference in scores observed (Caliendo and Kopeinig, 2008). Caliper matching: If the closest neighbor is far away, the nearest neighbor matching produces bad matches, and in such situation caliper matching algorithm is used. In caliper matching an individual from the comparison group is chosen as a matching partner for a treated individual that lies within a given caliper (propensity score range) and is closest in terms of propensity score (Caliendo and Kopeinig, 2008). If the dimension of the neighbor is very small, it is possible that some treated units are not matched because the neighbor does not contain a control unit. But, the smaller the size of the neighbor the better is the quality of the matches. Radius matching: Each treated unit is matched only with the control units whose propensity score falls into a predefined neighborhood of the propensity score of the treated unit. If the dimension of the neighborhood (i.e., the radius) is set to be very small, it is possible that some treated units are not matched because the neighborhood does not contain control units. On the other hand the smaller the size of the neighborhood the better the quality of the matches (Caliendo and Kopeinig, 2005).
In this study, logit model was used to predict the probability of each household being member in the MSAEs as a function of observed household characteristics used sample of the MSAEs members and nonmembers. In the logit model the dependent variable was membership, which assumed the value of 1, if a household head is member in MSAEs and 0 otherwise. The explanatory variables used in the model are age of household head, educational level of household heads, farming experience of household heads, access to training, distance to market, soil fertility status, household size, sex of household head, access to formal credit, landholding size, market information, off/non-farm income, livestock holding and membership to association. The outcome variables in PSM model were specified as total annual employment created by household's measured in terms of number and the amount of household's total annual income in Birr was taken as the outcome variables. Testing the matching quality: This is checking whether the matching procedure can balance the distribution of different variables or not. Since we are conditioning on propensity score estimation is not to obtain a precise prediction of selection into treatment, but rather to balance the distributions of relevant variables in both groups. While there are different procedures available to check, the basic aim of all of them is to compare before and after matching and if there still exists any difference after conditioning on propensity score. If the differences exist, there is an indication of incomplete matching and suggests remedial for actions (Caliendo and Kopeinig, 2008). Sensitivity analysis: The final step in the implementation of PSM is checking the sensitivity of the estimated results (Caliendo and Kopeining, 2005). Matching method is based on the CIA, which states that the evaluator should observe all variables that are simultaneously influencing the participation decision and outcome variables. The estimation of treatment effects with matching estimators is based on the selection at observables assumption. However, a hidden bias might arise if there are unobserved variables which affect assignment into treatment and the outcome variable simultaneously which abolish the CIA. To check the sensitivity of the estimated ATT with respect to deviation from the CIA, it is suggested that the use of Rosenbaum bounding approach is appropriate (Rosenbaum, 2002).

Descriptive Results
The survey results showed that out of 383 sampled household heads, 125 (32.5%) were members and the remaining 258 (67.5%) were non-members. From total sample household heads, the majority of them (74%) were male-headed (the corresponding figures are 85% and 69% for members and non-members respectively). The figures show that enterprises membership is dominated by male headed households. This could be attributed to various reasons, which could be due to the weak economic position of female headed households including limited access and use of information due to cultural barriers in the social position. This might prevent female headed households from being member in MSAEs. The mean age of the household heads was 32 years with a minimum of 25 years and a maximum of 71 years. The mean age of the member household heads was 31 years, whereas, that of the non-member household heads was 34 years. The result depicts that the household heads found both in member and non-member category were found to be in active working age. Education level of household heads varies among the member and nonmember households. About 21.3% member household heads were illiterate and 78.7% were literate whereas, 26.5% of non-member household heads were illiterate and 73.5% were literate. The land holding size among both members and non-members group indicates that members has small land size than non-members. From the member group 48.29% household heads have land size below one hectare whereas, for non-member group 27.2% household heads have land size below one hectare. Based on the land size possessed by non-member household heads respondent they are better than member of MSAEs in the study area.
The average household size of the sample household heads in the study area had large average household size of six persons with high mean total dependency ratio of 1.34. The average household size for member household heads was 7.34 persons while it was 5.06 persons for non-member household heads. When we compared the average household size between members and non-members, household heads the member had higher household size than household heads that did not member in MSAEs. From this we can conclude that, household size is the determinant factor of membership. The mean livestock holding in Tropical Livestock Unit (TLU) of the sample household heads was 6.45, where the minimum was 0 and the maximum was 32. The mean livestock holding for member households was 4.62 TLU and it was 8.29 TLU for non-member household heads. The member household heads have less livestock holding than non-member household heads. This means household heads with low livestock holding encouraged being member in MSAEs than the household heads owed large number of livestock.

Econometric Results
PSM method was used to evaluate the impact of membership to MSAEs on employment creation and income generation. Prior to the matching analysis, members significantly differed from non-members in most characteristics. According to the results, the imbalance between the members and non-members samples in propensity score, reduced much below 10% after matching; and the no case was significantly different from zero (P<0.01). This indicated that all differences in means between members and non-members had been removed through matching in the initial period (before member in MSAEs). The process of matching thus creates a high degree of covariate balance between the members and non-members samples that were used in the estimation procedure.
There are some important tasks to be carried out before conducting the matching exercise. First, we estimate predicted values of treatment participation (propensity score) for all the sample household heads (both member and non-member groups). Second, a common support condition should be imposed on the propensity score distributions of member household heads and non-member household heads which helps to identify common support area both members and non-members that their propensity score fall inside the interval. Thirdly, discard observations whose predicted propensity scores fall outside the range of the common support region. Fourth, conducting a sensitivity analysis to check the robustness of the estimation (whether the hidden bias affects the estimated ATT or not).
As shown below in Table 3.1, among members, the predicted propensity scores range from 0.0136 to 0.9644, with a mean of 0.6454. Among non-members, they range from 0.0007 to 0.8844, with a mean score of 0.1718. Thus, the common support assumption is satisfied in the region [0.0136 to 0.8844]. In other words, households whose estimated propensity scores are less than 0.0136 and larger than 0.8844 were not considered for the matching exercise. As a result of this restriction 14 member households were excluded from the analysis.  (2017) The choice of matching estimator was conducted based on three different criteria as proposed by Dehejia and Wahba (2002). After obtaining the predicted values conditional on the observable covariates (p-score) from logit estimation, matching was done using matching algorithm which is selected among the commonly used matching methods (kernel matching, nearest neighbor matching, calliper matching and radius matching) based on the best fit matching algorithm selection criteria: estimator which have low pseudo-R2, large matched sample size, large number of insignificant variables after matching and mean standard bias between three and five. Thereafter, the estimation results and discussion are the direct outcomes of the nearest neighbor (4) matching algorithm.
European Journal of Business and Management www.iiste.org ISSN 2222-1905(Paper) ISSN 2222-2839(Online) Vol.12, No.4, 2020 (2017), nearest neighbor (4) matching was the best matching. * Number of explanatory variables with no statistically significant mean differences between the matched groups of member and non-member households.
After selecting the best performing matching algorithm, the next task is to check the balancing of propensity score and covariate using different procedures by applying the selected matching algorithm [nearest neighbor (4) matching in my own case]. The values of pseudo-R 2 and LR chi-square, before and after matching, can be used as indices for the fulfillment of the balancing requirement. A low pseudo-R2 value shows that member households do not have much distinct characteristics overall and as such finding a good match between member and non-member households becomes simple. The low value of pseudo-R 2 and the insignificant LR chi-square after matching supported the hypothesis that both groups have the same distribution in covariates after matching. This implies that we have found a comparable group of members with non-members on employment creation and income generation based on similar covariates. The main intention of estimating propensity score is to balance the distributions of relevant variables in both groups. The balancing powers of the estimations are ensured by following testing methods (Table 3.3).  (2017) The standardized bias before match and after match, the total bias reduction obtained by the matching procedure as shown at Table 3.3, standardized difference in covariates before matching is in between of 1% and 19.3% in absolute value whereas the remaining standardized difference of covariates for almost all covariates lies between 0.5% and 18.4% after matching. This is fairly below the critical level of 20% suggested by Rosenbaum and Rubin (1985). Therefore, the process of matching creates a high degree of covariate balance between the member and non-member samples that are ready to use in the estimation procedure. The same to that, t-values also reveal that before matching two of chosen variables exhibited statistically significant differences while after matching all of the covariates are balanced and become statistically insignificant, suggesting that matching helps to reduce the bias associated with observable characteristics.
These results indicate that the matching procedure is able to balance the characteristics in the treated (members) and the matched untreated (non-members) groups. Hence, these results can be used to assess the impact of MSAEs on employment creation and income generation. This enables researcher to compare observed outcomes for treatments (members) with those of an untreated (non-members) groups sharing a common support region.

Impact of MSAEs on employment creation
The PSM results confirm that there were difference between member and the non-member household's in terms of total annual employment created by households in number. A comparative analysis shows that member households were better than non-member households by creating employment opportunities for about 4 people annually (33.98% higher than non-member households) and this gain was statistically significant at 1% probability level. In this case, the mean employment created by the member households was about 6 people annually and that of the non-member households was about 2 people annually. These results confirm that membership in MSAEs increased the employment creation of the households, which is the good remedy for number of jobless individuals in the study area. European Journal of Business and Management www.iiste.org ISSN 2222-1905(Paper) ISSN 2222-2839(Online) Vol.12, No.4, 2020

Impact of MSAEs on income generation
According to the best matching algorithm estimates (i.e. nearest neighbor (4) matching algorithm) showed that being member in MSAEs had a positive and significant impact on the amount of household's total annual income received in Birr (Table 3.4). As the result indicated that the impact of being member in MSAEs was significant at 1%. As expressed below in Table 3.4, the mean household's total income was about Birr 18,302.00 for member household's, while the corresponding figure for the non-member household's was Birr 5,962.00 and the average income difference between member and non-member households were Birr 12,339.00. It implies that the total annual income of the member household's was almost 32.57% higher than that of total annual income of the non-member household's. The results indicate that membership in MSAEs increased the total annual income of the household's, which is the god source of income for the household's to carve the problem of income scarcity.

Sensitivity Analysis
In observation studies, program intervention treat are not randomly assigned experiment units, randomization tests are not generally applicable. Thus to compensate for the lack of randomization; member and non-member units are matched on the basis of observed covariates. However, in most case possibilities remain of bias due to residual imbalance in unobservable covariates. Therefore, sensitivity analysis were carried out to check quality of comparison matching among member and non-member group with observed covariates and mainly to check robustness for unobserved covariates and the result indicates the ATT estimate result are insensitive.
In case the CIA fails in PSM it can be easily solved the pitfall using the comparison between the simulated and baseline ATTs estimates. Thus, for any given configuration of the parameters the sensitivity analysis retrieves a point estimate of the ATT which is robust to the failure of the CIA implied by that particular configuration. Using a given set of values of the sensitivity parameters, the matching estimation is repeated many times and a simulated estimate of the ATT is retrieved as an average of the ATTs over the distribution of U. As it can be seen from the Table 3.5 provided below, though U is associated with a large outcome effects ( >1) and selection effects ( >1) for the nearest neighbor (4) matching algorithms, the overall simulated ATTs of each member of MSAEs are still too much closer to the baseline ATTs. Hence, both the values of outcome effect and selection effects are larger than unity each, and also the difference in percentage between the baseline ATTs and simulated ATTs are below 10% make it stronger in the credibility of our estimated ATTs as well. The simulated ATT of each of the employment creation and income generation are too close to the baseline estimate. Obviously, this implies that it is only when U is simulated to provide incredibly large outcome effect; the ATT can be driven far from the baseline estimates or even closer to zero. Thus, we can conclude that our impact estimates ATT are insensitive to unobserved selection bias and are a pure effect of membership in MSAEs. On the whole, all the results estimated support and strengthen the robustness of the matching analysis is the reliable conclusion.

Conclusion and Recommendations
The survey results showed that out of 383 sampled household heads 125 (32.5%) were members and the remaining 258 (67.5%) was non-members. From total sample household heads, the majority of them (74%) were male-headed (the corresponding figures are 85% and 69% for members and non-members respectively). The mean age of the household heads was 32 years (it was 31 years for member and 34 years for non-member) household heads. The result depicts that the household heads found both in member and non-member category were found to be in active working age. The land holding size among both members and non-members group indicates that MSAEs members has small land size than non-members. From the member group 48.29% household heads have land size below one hectare whereas, for non-member group 27.2% household heads have land size below one hectare. Based on the land size possessed by non-member household heads respondent they are better than member of MSAEs in the study area. The average household size of the sample household heads in the study area had large average household size of six persons with high mean total dependency ratio of 1.34. The average household size for member household heads was 7.34 persons while it was 5.06 persons for non-member household heads. From this we can conclude that, household size is the determinant factor of membership in MSAEs in the study area. The mean livestock holding in Tropical Livestock Unit (TLU) of the sample household heads was 6.45, where the minimum was 0 and the maximum was 32. The mean livestock holding for member households was 4.62 TLU and it was 8.29 TLU for non-member household heads. This means household heads with low livestock holding encouraged being member in MSAEs than the household heads owed large number of livestock.
The impact of MSAEs on employment creation and income generation between members and non-members was seen in this study clearly. A comparative analysis shows that member households were better than nonmember households by creating employment opportunities for about 4 people annually (33.98% higher than nonmember households) and this gain was statistically significant at 1% probability level. In this case, the mean employment created by the member households about 6 people annually and that of the non-member households was about 2 people annually. These results confirm that membership in MSAEs increased the employment creation of the households, which is the good remedy for number of jobless individuals in the study area. The impact of being member in MSAEs was significant at 1% probability level. Based on the result of PSM, the mean household's total income was about Birr 18,302.00 for member household's, while the corresponding figure for the non-member household's was Birr 5,962.00 and the average income difference between member and non-member households were Birr 12,339.00. It implies that the total annual income of member households was almost 32.57% higher than the total annual income of the non-member households.
The result showed that membership in MSAEs had a significant positive impact on employment creation and income generation and the impact estimates were found to be insensitive to unobserved selection bias. It is, therefore, essential to expand and strengthen development of MSAEs and the membership of households by giving special attention and necessary support to the sector in the study area. Hence, the findings of this study recommend the need for implementing different policies and strategies that separately target and address the MSAEs' that improves their impact on employment creation and income generation.