Determinants of the Market Outlet Choice of Bamboo Culms ( Yushania alpine ) Producers in Banja District, Western Ethiopia

. Bamboo meets a growing and various bamboo product demand and generates revenue. Te study is aimed at analyzing the critical variables that afect the choice of alternative bamboo market outlets. By using two stages sampling procedures, 114 bamboo producers were randomly and proportionally selected. Te determinants that afect the choice of bamboo market outlet were analyzed by multivariate probit model. Based on the model result, the probability of bamboo producers to select wholesaler, retailer, processor, and local traders’ outlet were 16.8%, 46.2%, 60.8%, and 54.3%, respectively. Te probability of success and failure to select four market outlets were 2.5% and 2.2%, respectively. Te result of MVP revealed that family size, total land holding size, amount of culm production, farming experience, distance to the market, and silviculture management practice afected the probability of farmers’ market outlet choice. Terefore, improving the producers’ knowledge and skills through capacity building and creating framers’ organization for collective action can help producers select the right market to sell bamboo products.


Introduction
Bamboo is one of the most important vegetation resources in the highlands of Ethiopia, with diverse local and national importance in terms of flling subsistence needs and cash income [1]. Te trading of bamboo products is limited to the domestic market. Te demand and supply of bamboo culms are minimal, and the trading activity is small regarding product to the locations, distributions, and fnal usage [2]. Te production, marketing, and processing systems remain under-developed and informal [3]. Te highland bamboo can grow between 2200 and 3200 m.a.s.l. and average yearly temperatures ranging from 10 to 200°C with an annual rainfall of 1700-2200 mm [4]. Bamboo is the most economically useful nontimber forest products, with its renewable nature and accessibility to rural poor people. It also has great potential for commercialization and can drive rural development. It can be used at all levels of industrial activity from small crafts-based industries to modern integrated plants [4]. Yushania alpina and Oxytenanthera abyssinica A.Rich Munro are distributed in South, Southwest, and in the central parts of Ethiopia, which includes Benishangule Gumuz, Oromia region, southern nations nationalities, and Amhara region [5].
Te bamboo value chain in the Awi zone is characterized by raw material distribution followed by marketing and product reaching to fnal users. Bamboo producers sell their bamboo culms to local traders who in turn sell the product to processors [6]. Te choice of marketing channel is an important farm-level decisions which have a great impact on the income of households' [7]. Te choices of marketing outlets are mostly household-specifc decisions, and they require the consideration of demographic, socioeconomic and market-related factors [7,8]. Understanding the factors that afect the choice of market outlet selection strategies is imperative since the exploitation of such strategies has the potential to increase crop production, investment, and farm income [9]. Muricho et al. [10] argue that understanding the relationship between the market outlets and the factors that afect the selection of each market outlet is fundamental to profling the markets as well as establishing policy interventions that are carefully designed to beneft farmers. Each market outlet is characterized diferent market return, risk, cost structure, and other requirements; understanding these characteristics is benefcial to producers who aim to access market outlets [9]. In order to get the maximum return, producers can select diferent market outlets. But, diferent factors afect the selection decisions of households. Identifying the factors help pinpoint the possible areas of intervention that may help bamboo producers in maximizing benefts. Tis study aimed to identify the factors that afect the market outlet choices for bamboo culm producers. Terefore, understanding diferent factors afecting the market outlet choice of bamboo culms helps to design sound policies related to the marketing of bamboo products and the overall contribution of the bamboo sector to the development of the nation.

Area Description.
Banja district was selected for this study ( Figure 1). It has 25 rural and 1 urban kebeles. From the total kebeles highland bamboo is produced in 20 kebeles. From Northwest of Addis Ababa and Bahirdar city, it is around 442 km and 116 km, respectively. Te total population of the Woreda is estimated at about 315,271, with 5% in towns and 95% in rural areas. From the above population, 187,213 are females and 128,058 are males, respectively, and average family size is 7 people per HH [11].

Sampling
Procedures. Two stage sampling procedure was used to select sample producers. Initially, 20 potential kebeles were identifed, and four kebeles were selected purposively. Te sampled kebeles include Kessa, Gashena, Ledeta, and Surta. Ten, populations were listed out and 114 sample producers were selected randomly. Yamane [12] is used to determine the sample size. In the selected kebeles, highland bamboo is produced by 1,449 farmers. Based on this formula, 114 sample respondents were selected from each kebele based on the proportional probability of the size of the population (Table 1).
where N = the number of producers in selected kebeles, n = the sample size, and e = the level of precision which is 9%. (2)

Method of Data Collection.
Primary data were collected through conducting semistructured interviews, and secondary data were collected from diferent published articles. Primary data were collected from bamboo producers, local traders, wholesalers, and processors. During the household survey, the following issues were included annual bamboo culm production, bamboo market supply, price of bamboo culms, and amount of income gained from diferent bamboo products. To support the primary data, secondary data were collected from diferent ofce reports and published materials.

Data Analysis.
Descriptive statistics was used to analyze the socioeconomic characteristics of bamboo producers and multivariate probit model (MVP) was employed to analyze the determinant factors. Econometric models such as multinomial probit/logit and multivariate probit models are used for the analysis of categorical choice dependent variables. Te multinomial probit/logit model assumes independence across the choices and does not allow correlation between alternative choices. MVP considers the interdependence and correlations among the outlets. It is an extension of the probit model and is used to estimate several correlated binary dependent variables jointly [13]. Te decision whether to select the market outlet or not is considered in proft maximization [14]. It considered interdependence among the choices of alternative market outlets. Assume i th bamboo producer (i � 1, 2, 3, 4, . . ., N) facing a problem related to the selection of alternative market outlets. Assume U * K denote proft of producer to select wholesalers (Y1), retailers (Y2), processors (Y3), and local traders (Y4). Te bamboo producer decides to select represent the utility derived from the selected K th market outlet if selected by i th farmer and U 0 is utility if the market k is not selected. Te net beneft (Y * ik ) that producers obtained by selecting a market outlet which is a latent variable depends on the observed independent variables (Xi) and error terms (ϵ i ) as given in the following equation: Terefore, based on the indicator function, the abovementioned equation can be translated into observed binary outcome equation as follows: Te probabilities that all alternative market outlets selected by a producer can be entered into the likelihood function are specifed as follows: 2 International Journal of Forestry Research In MVP, the selection of combinations of the market outlet is possible, and the error terms jointly follow a normal distribution with a zero mean, normalized to unity, and symmetric to the covariance Ω given as follows: Te function of log-likelihood associated with the sample outcome is described as follows: where ⍵ is an optional weight for observations i. . .N and Φi with arguments µi and Ω are standard normal distributions.

Hypotheses and Defnition
and end users of bamboo culm. Local traders are the traders who purchase and collect bamboo culm from the producers and sold it to wholesalers and retailers. Tey sold bamboo culm to wholesalers and processors in the district. Each market outlet is a binary indicator which takes one if the producer chooses the given alternative outlet and zerootherwise. Te hypothesis of independent variables were presented in Table 2.

Characteristics of Bamboo Producers.
Te age of producers ranged from 26 to 75 years. 53.11 years is the mean age of households in the study area. It indicated most of the producers are categorized under the productive age of the population. Te family size of respondents ranged from 2 to 9 members. Land is the main production factor which afects the production and supply of bamboo culm. In Banja district, producers used their land for bamboo production. Te minimum and maximum land sizes of the producer was 0.125 and 2.75 ha, respectively, and the mean land size is 1.21 ha of land. Te average land size allocated for bamboo plantation was 0.15 ha (Table 3). It indicates that the land size allocated to bamboo production is limited.

Bamboo Marketing Outlets.
In Banja district, bamboo producers used wholesalers, retailers, processors, and local traders to sell bamboo culms. However, the selection of right market outlet is not easy because it is afected by diferent factors. In this section, based on the output of the multivariate probit model signifcant independent variables were discussed.
Te Wald X 2 (102.52, p � 0.001) indicated that the test is signifcant at 1%. It indicated the subsets of coefcients are jointly signifcant, and explanatory variables explained the dependent variables. Terefore, multivariate probit model is highly signifcant. Te likelihood ratio test (LRX 2 (6) = 29.68 (prob > X 2 = 0.001)) which is statistically signifcant at 1%. It revealed the rejection of the null hypothesis, i.e., all rho values equal to zero are rejected based on the test result. It also indicated the ftness of the model and the decision to select alternative market outlets were interdependent with each other, and it supports the use of the multivariate probit model.
Te ρ values (ρij) in Table 4 indicate the correlation between each pair of market outlets. Te value of ρ31 (the degree of correlation between processor and wholesaler) is negatively correlated and statistically signifcant at 1%. Te value of ρ42 (degree of correlation between local traders and retailer) is negatively correlated and signifcant at 1% and the value of ρ43 (correlation between local traders and processor) indicated the existence of negative correlation and which is statistically signifcant at 1%. Tese correlations indicate that each pair of market outlets competes with each other.
Te maximum likelihood estimation indicated the marginal probability of for each market outlets. As indicated in Table 4, the likelihood of choosing wholesale market (16.8%) was small as compared to the likelihood of selecting (46.2%), processor (60.8%), and local trader (54.3%). Te joint probability of success and failure was 2.5% and 2.2%, respectively. It indicated that bamboo producers are more likely to select four market outlets jointly (Table 4).
Tree variables afect the wholesale market; two variables afect the retailer market; four variables the processor market; and two variables afect signifcantly the local market outlet at 1%, 5%, and 10% levels of signifcance (Table 5).

Family Size of the Household.
It was positively and negatively correlated and afected the probability of selecting wholesale market and processor market outlets at a 1% and 5% signifcance level, respectively. Household heads who have more active labour forces have high probability of selling bamboo culms to wholesale market because it helps the producer to harvest a large number of bamboo culm which they can deliver to the wholesale market. Tis fnding  is supported by the fndings of Honja et al. [15], who found that the active labour force in the household afected the probability of selecting wholesale market outlet positively. It is also consistent with Melekamu [16] who reported that households with a large family size were positively related to the probability of choosing a wholesaler outlet because having a large family size helps to supply output to wholesale market rather than selling to local traders. In the other way, households who have more number of family members have less probability to choose processor outlet rather they want to produce diferent value-added product, in this case, the probability of supplying to the processor become decreased. Tis fnding is consistent with Kuma et al. [8], who reported that family size has a relationship with the probability of selecting processor market outlets.

Total Land Size of the Household Head.
It was positively correlated with wholesaler, and it was signifcant at a 10% level of signifcance. Tis fnding revealed that households who have a better land size are more likely to choose wholesaler outlet than households having smaller area of land because it enables to produce more number of bamboo culm and wholesalers need to buy a larger number of bamboo culms in bulk form. Total land size was positively correlated with assembler market outlets because when bamboo producers have large total land size, the probability of allocating land for bamboo plantation increases. It enables the producer to produce large volumes of agricultural products for wholesalers who can purchase the product in large volume [17]. Similar results were reported by Dessie et al. [18] and Kumar [19], who indicated that farm size positively afected the choice of assembler market outlet.

Farming Experience of the Household Head.
It was negatively correlated with retailers and local traders, and it was statistically signifcant at 1%. It indicated that more experienced bamboo producers have less probability of selling bamboo culms to the retailers' market outlets and local traders than less experienced bamboo producers. Because, through time, experienced producers make market linkage with processors. Tis fnding is in line with Wosene [20], who found that experience afected negatively the Log-likelihood ratio: .68 * * * probχ 2 � 0.001 * * * , * * , and * indicate the statistical signifcance at 1%, 5%, and 10%.  probability of selecting retailer market outlet and it is also supported by the report of Emana et al. [21], who indicated farming experience is negatively correlated and afected negatively the probability of selecting collectors. On the other way, it afected the probability of choosing the processor positively at a 1% signifcance level. It means through time, experienced bamboo producers form relationship with bamboo processors. A study conducted by Kiplangat and Kiprop [22], who indicated that pineapple farming experience has a positive relation with the probability of selecting urban market outlet.

Distance to the Market.
As hypothesized, it was negatively and positively correlated with processors and local traders, respectively, and it was signifcant at a 1% signifcance level. Households who walked long distances are less likely to sell bamboo culms to processors because bamboo processors found on towns. Tis result also supported by Tilahun [23], who found that long market distance has a negative relation with the probability of choosing processor market outlets. On the other way, producers who are distant from the district market are more likely to choose local traders to sell bamboo culms because the local traders collect the bamboo culm at the farm gate; in this case, producers did not expend transportation cost. Households located far away from the nearest market face difculty in delivering their output to retail markets. As a result, they prefer to sell products to local collectors because collectors can purchase products at farm gate [24]. Te result is consistent with Mmbando [25], who revealed that producers far from the market incurred a high transaction cost, and they opt for the brokers to sell the product at a production place rather than selling to the wholesale market. Tis fnding is also consistent with the fnding of Emana et al. [21] who indicated that distance from the nearest rural market is negatively related to the retail market because they prefer to sell their products to local traders.

Silviculture Management.
It was found to be positively correlated with processors and statistically signifcant at 5%. It indicated that households who practice good silviculture management are more likely to select processor outlet as compared to other bamboo producers who cannot practice silviculture management in bamboo farming. It implies that when producers apply good silviculture management activities, they can produce a number of bamboo culms with the required quality. In this case, processors require a bamboo culm which has a required size, type, and thickness with the required quality. Terefore, the probability of selling to the processor increases.

Number/Quantity of Culm Harvested.
It was positively correlated with wholesalers and retailers and signifcantly afected wholesalers and retailers at a 10% level of signifcance. Te result indicated that households who can produce large number of bamboo culm mostly preferred to select combinations of two or more market outlets found in the district. It implies that when bamboo producers can produce enough bamboo culms, they can sell culms to wholesalers and retailers. It is supported by Mirie et al. [17], who reported the probability of selecting wholesale and retail markets were positively afected by the quantity of tef production. A study conducted by Tilahun [23] also indicated that the probability of selecting wholesalers and retailers outlets were afected by the quantity of potato produced. Again, it is in line with Medekesa [26], who indicated that the quantity of cofee sold to the market agents increases the probability of farm households choosing trader market outlets.

Conclusions and Policy Implications
Te study was aimed at analyzing determinants of the market outlet choices of bamboo producers. Te joint probability of success and failure was 2.5% and 2.2%, respectively. Te result of a multivariate probit model also revealed that the three ρ values are statistically signifcant and negatively correlated. It indicated the existence of competitive relationship between the processor with wholesalers, local traders with retailers, and local traders with processors. Te result of the MVP model indicated that bamboo producers in the area made their market outlet choice based on the bamboo farming experience, area of total land holding size, number of bamboo culm harvested, family size, distance, and silviculture management practice. To increase the beneft of bamboo producers, micro, and small enterprise ofce and development agents of the district should link the processors with bamboo producers. If producers can link with bamboo processors, the market margin for processors and producers become increased because unnecessary costs can be eliminated through market linkage. Terefore, it is the appropriate market outlet for bamboo producers to increase their marketing margin. Awareness creation about ways of obtaining market information, ways of selecting appropriate market outlets, and ways to get fair price for bamboo products should be provided by development agents and market experts for bamboo producers. Te development agents and other concerned bodies should improve the producers' knowledge and skills through capacity building and creating framers' organization for collective action help to increase the market supply of bamboo, the share of producers, and the ability of producers to choose appropriate market outlets to sell bamboo culms.

Data Availability
Te data used to support the fnding of this study are available from the corresponding author upon reasonable request.

Conflicts of Interest
Te authors declare that they have no conficts of interest.