Assessment of Factors Causing Coffee Yield Gap Among Smallholder Farmers in Mbinga and Mbozi Districts

This study assesses the factors causing coffee yield gap among smallholder in the study area. The aim of this study is to increase coffee productivity from the current level. The primary data were collected from 218 adopters and 102 non-adopters of improved coffee varieties using a structured questionnaire. The descriptive statistics was used to assess yield gap and linear regression model was used to determine factors causing yield gap among smallholder farmers in the study area. The findings showed that the yield potential (3000 kg/ha and 1000 kg/ha) for improved and traditional coffee varieties respectively has not yet been realized by farmers and there is a large gap between the average coffee yield (1141 kg/ha and 384 kg/ha) gained by smallholder farmers growing improved coffee varieties and farmer growing traditional coffee varieties respectively. The yield gap from smallholder farmers with improved coffee varieties was 2000 kg/ha and 646 kg/ha from traditional coffee varieties. The main factors causing coffee yield gap were lack of access to extension services (p<0.000), plant population (0.007), low use of fertilizer (p< 0.002), coffee diseases (p<0.008). To minimize coffee yield gap in Tanzania promotion of the use of improved coffee varieties, fertilizer and agro-inputs is important.


Introduction
Coffee is among of the important tropical valuable and traded commodities produced in 70 countries in the world (Sänger, 2018). The average total coffee production by all exporting countries between 2008 and 2018 was 146 997 thousand 60kg bags of which Brazil contributed about 35% of all coffee produced, Vietnam 17%, Colombia 6%, Indonesia 7%, Ethiopia 4%, Uganda 3%, Côte d'Ivoire 2%, Kenya 0.9% and Tanzania 0.7% (ICO, 2018a). The productivity of coffee in Tanzania is ranging from 250 to 300 kg/ha for Arabica coffee produced by smallholder farmers . The productivity levels of other coffee growing countries such as Kenya 302kg/ha (ICO, 2019a), Ethiopia 802 kg/ha (Bickford, 2019), Rwanda 880 kg/ha (Nzeyimana, 2018) and Uganda 2100 kg/ha (ICO, 2019b) are relative higher than Tanzania. In general, it can be argued that coffee yield from smallholder farmers in Tanzania is low despite of Tanzania having coffee varieties with potential of producing 3000 kg/ha and 1000 kg/ha for improved and traditional coffee varieties respectively the coffee yield from smallholder farmers is lower compared to the potential yield levels (Kilambo et al., 2015). Progressive promotion of high yielding coffee varieties and replacement of old coffee trees with improved coffee varieties coupled with farmer training on the implementation of good agricultural practices (GAPs) were expected to increase coffee yield and close the yield gap. Yield gap analyses method was used by this study to understand the factors causing coffee yield gap among smallholder farmers in the study area. The findings from this study will help to inform policy makers, researchers and coffee producers on the strategies to be used to minimize yield gap among smallholder farmers in the study area.

1.1
Theoretical, Empirical and Conceptual Frameworks

Theoretical framework
Agricultural production economics is concerned primarily with economic theory as it relates to the producer of agricultural commodities (Debertin, 1986). The major concerns in agricultural production economics among other goals and objectives of the farm manager, choice of outputs to be produced, allocation of resources among outputs. The theory of utility maximization has been used extensively to explain the preference of inputs by farmers (Muellbauer, 1974). This theory predicts that farm productivity, measured by marginal factor products, will differ over farms using different levels of inputs. This theory considers a simplified view of the economy in which production output is determined by the amount of labour involved and the amount of capital invested. The strategies of increasing the coffee production in Tanzania include encouraging investment on man power through training, dissemination of improved seedlings, remove of taxes on agricultural related inputs such as fertilizer, pesticides and herbicides (TCB, 2012a).
Therefore, the Cobb Douglas production function is useful as the basic framework for understanding the causes of gap resulted from utilizing the major factors coffee production in the study area.

The empirical framework
Different approaches such as field experiment (van Ittersum et al., 2013), crop growth simulation models (Lu & Fan, 2013;and van Bussel et al., 2015), socio-economic survey (Tamene et al., 2016), and precision agriculture (Schulthess et al., 2013;Tittonell & Giller, 2013) have been used to assess farmers' yields and yield gaps between households. Field experiments approach is used to compare farmers yield 'control' and research yield within the experimental plots (van Ittersum et al., 2013). This approach does not set the optimum yields to the level attainable by farmers and the limited number of test locations makes it difficult to upscale to larger areas (Tittonell & Giller, 2013). Crop growth simulation models (Lu & Fan, 2013;and van Bussel et al., 2015), compare the potential yield with actual yield but requires intensive data for model input, calibration and validation (van Ittersum et al., 2013). Precision agriculture which is fast and accurate is used to measure yield of plots in real time but it is more of technology intensive (Tittonell & Giller, 2013).
Socio-economic survey capture yield harvested by smallholder farmers and corresponding agronomic/management practices depends much on farmers memory and it is time saving (Tamene et al., 2016). This study therefore opts to use the socio-economic survey because of the nature of data collected from the study area depended on farmers records.

The conceptual framework
The conceptual framework developed in this study was mainly intended to determine a set of one or more factors of coffee production that will bring about optimal levels of coffee output for smallholder farmers in Mbinga and Mbozi districts. To estimate the yield gap among smallholder farmers, the analysis of input preference and usage is grounded in the theory of utility maximization (Huffman, 2011 andMuellbauer, 1974). Following that theory, it is hypothesized that farmers would prefer the use of an input as long as the utility derived from the application is greater than that of other inputs. An important conceptualization of this study is that preference of an input usage is not an end in itself.
That is the utility of the farmer is not only maximized by application of an input, but also by the achievement of set objectives or outcomes. Farmers use inputs in order to achieve some objectives, some are subjective and others are objective. Some objectives for using some inputs include improved quality of coffee berries, increased productivity and profitability of coffee. This study therefore provides an opportunity to examine which factors play a significant role in both coffee productivity ( fig. 1).

Description of the study area
This study was conducted in two coffee producing Districts in Tanzania

Research design
The present study employed a cross-sectional research design. The cross-sectional design was used because it is suitable for description purposes as well as for the determination of relationship between variables and it is cost effective and saves time over longitudinal and panel data. This design has been recommended by several scholars including (Levin, 2006;Omair, 2015;Setia, 2016).

Sampling frame
Sampling frame included households engaged in coffee production in four villages. Coffee growers were selected from family household's viewpoint as a unit of assessment. Also, similar check-list of questions was used in Focus Group Discussions (FGD) and interview with Key Informants' (KIIs) so as to validate information obtained from households and FGD respectively.

Sampling procedure
To collect the representative sample, purposive and random sampling procedure were used in selecting districts, wards, villages and the household. At the first stage, purposive sampling has been applied in selecting the two districts intended by this study. Secondly, random sampling was applied in selecting wards from the long list of wards where coffee is grown. The third stage involved random sampling of villages with adopter of improved coffee varieties and lastly, random sampling was applied in selecting coffee households growing improved and traditional coffee varieties in the villages with at least 100 number of coffee trees which is the minimum number of improved coffee trees farmer a farmer can own to break even. All these stages involved collaboration with districts and wards extension officers. Finally, from the list of coffee growers developed in the second and third stage, a required sample size of respondents was proportionally selected from each village.

Sample size determination
This study used the formula in equation 1, to determine sample size of smallholder farmers growing improved and traditional coffee varieties from Mbozi and Mbinga districts (Krejcie, 1970a). Therefore, a sample size of 320 was collected from six randomly selected wards: Igamba, Isansa and Ihanda in Mbozi district Kilimani, Utiri and Luwaita in Mbinga district as shown in (Table 1). (1) Where: S=Required sample size, X =z value (assumed to be 1.96 for 95% confidence level), N = Population size, P = Population proportion (assumed to be 0.5 since this would provide the maximum sample size), d = degree of accuracy (5%), expressed as a proportion (0.05). Accordingly, Mbozi district consists 930 households and Mbinga district consist of 990 households, making a total of 1920 target households.

Focus group discussion (FGDs)
Focus Group Discussions (FGDs) were used to collect primary data. About 42 participants were involved in making two groups from each district; one for those with improved varieties and the other for those with traditional varieties making a total of four groups.
Each group had that comprised 7 -8 participants (including 1 to 2 females) who were purposively selected among coffee producers. Participants in FGDs were different from those involved in questionnaire interviews. The rationale for the choice of focus group discussion method was that it helped to capture in-depth information on factors affecting coffee yield among adopters and non-adopters of improved coffee varieties and to validate some information gathered during primary data collection from the households in the study area.

Key informant interviews (KIIs)
Key Informants' Interview (KIIs) was used to collect primary data. Key informants included ward extension staff, local leaders one from each ward in the study area respectively, District Coffee Subject Matter Specialist (DCSMS) and TaCRI extension officer to make a total of 9 KIIs in the discussions from each district in order to obtain their opinion on factors affecting coffee yield among adopters and non-adopters of improved coffee varieties and to validate some information gathered during primary data collection from the households and focus group discussions.

Coffee yield estimation among smallholder farmers
Data collected through farmer interviews were coded and analysed using the SPSS whereby descriptive statistics (i.e. frequencies, percentages, means, minimum and maximum values of variables) were determined. The coffee yield reported by farmers recall were converted to standard unit conversion factors of which for this study was kg/ha and findings were compared using t-test of difference in means to establish if there is any statistically significant difference between yield obtained by smallholder farmers with improved coffee varieties and those with traditional coffee varieties.

Yield gap estimation among smallholder farmers adopter and non-adopters
Yield gap (Yg) is a quantitative difference between average research yield (generally reported from research trials) and average farmers yield (generally: obtained from acceptable farmer management practices) over some specified spatial and temporal scale holding other crop attributes remain constant (Sadras, 2015). Yield gap as the difference between the maximum farmer yields (attainable yield) and the average farmer yields (Lobell et al., 2009;Tamene et al., 2016). Yield gaps were analyzed by comparing farmers yield with yield data from research studies carried out by TaCRI. The research yields data was collected from the recently published reports from TaCRI. The average farmer yield can be affected by outliers, then for purpose of this study the median yield was used as a reference attainable yield versus potential on-farm experimental yield. Therefore, coffee yield gap Yg (kg/ha) is defined as the difference between the potential on-farm experimental yield (Yr) and the attainable yield obtained from farmers management practices (Yf). = − …………………………………………………………………. (2) Where: =Yield gap, = research yield and = the average farmers' yield.
The t-test analysis for the two categories of respondents with improved coffee varieties and respondents with traditional coffee varieties were done to compared difference in means to establish if there is any statistically significant difference between yield gap among smallholder farmers with improved coffee varieties and those with traditional coffee varieties.
The average of these yield gaps was also calculated as the expression of overall yield gap index as follows:

Estimation of factors affecting coffee yield
Linear regression analysis and correlation methods were used to estimate predictors of yield gap among smallholder farmers in the study area. These methods have been widely used in yield gap studies to show specific factors influencing crop yields However, the heterogeneity in smallholder farms is likely to result in high spatial variability in yield gaps and their causes. To obtain a spatial view of the causes of yield gaps on smallholder farms, multivariate regression analysis was done. Some important variables were taken on the bases of perceptions. The production function regression model (Yusi, 2016) as cited in (Echevarria, 1998) was chosen and hypothesis of the study were taken into consideration. The following regression equation was estimated, = 0 + 1 1 + 2 2 + 3 3 + 4 4 + ⋯ + 9 9 + ……………(4) Where Y = coffee yield gap of i th crop (kg/ha), X1 = gender (female, male), X2 = age of the farmers (years), X3 = household size, X4 = level of education (years spent in school), X5 = household size (total number in the household), X6 = Extension contacts, X7 = farm size (ha), X8 = Agronomic practices (weeding, fertilizer application, pruning and control of coffee pests and diseases), X9 = Technology adoption index and e = error term, β0 = Intercept, β1… β = Regression Coefficient.

Land under coffee production and type of coffee varieties grow
The findings from the study area showed that, the average land size under improved coffee varieties is 1.6 ha and 2.19 ha is under traditional coffee varieties. The findings imply that, the area under improved coffee varieties is small than area under traditional coffee varieties. The findings in Table 2 (Mhando, 2017 andTaCRI, 2017).

Coffee yield estimation
The findings in Table 3 show that, the average coffee yield from smallholder farmers with improved coffee varieties is 1141 kg/ha, median yield was 1000 kg/ha, maximum was 3842 kg/ha and minimum was 327 kg/ha. Respondents with traditional coffee varieties gain an average yield of 384 kg/ha, median yield was 354 kg/ha, maximum yield was 850 kg/ha and minimum yield was 126 kg/ha. The findings imply that, farmers with improved varieties have relative higher yield than farmers with traditional varieties. Similarly, the maximum yield recorded from smallholder farmers is an indication of possibilities to rise current coffee production from the current average yield to relative maximum yield. In addition, the finding implies that, at the farm level yield can be realized through proper input use and farm management. This findings support (Diro & Erko, 2019;Wu, 2005) who reported that, farmers who adopted improved varieties gain higher yield than those with traditional varieties. The study conducted by TCB, 2012a) recoded the average yield of 555 kg/ha, maximum yield of 1554 kg/ha and minimum yield 51 kg/ha in Mbinga district and average yield of 422 kg/ha, maximum yield of 1745 kg/ha and minimum yield 253 kg/ha in Mbozi district. 0.000

The estimation of coffee yield gap
The yield gap was computed by subtracting the average research coffee yield reported from research trials and the median yield obtained from farmers in the study area. The findings in Table 4 indicates that, the median yield gap from smallholder farmers with improved coffee varieties was 2000 kg/ha and 646 kg/ha from traditional coffee varieties.
The possible reason for yield gap include low use of fertilizer (Tamene et al., 2016).

The factors causing coffee yield gap among smallholder farmers
To identify the determinants of coffee yield gap among adopters and non-adopters of improved coffee varieties, linear regression analysis was done. The findings indicated that the variables specified in the model fitted well because the F(14, 198) and statistically significant at P= value 0.000 with adjusted R-squire 0.93 which show the strength of the model.
The findings in Table 5 show that, visits by extension officers have positive coefficient (129.645) and statistically significant at 0.000 implying that, lack of extension services to farmers contribute to increase coffee yield gap. Different scholars (Ghimire et al., 2015;Lugandu, 2013;Teferi et al., 2015) documented that, farmer access to extension services help in improving farm management practices hence increase the chance of reducing yield gap. Therefore, the transfer of the recommended agricultural practices through extension officers could effectively help farmers minimize yield gaps.
The findings also indicated that plant population per ha have positive coefficient and statistically significant at (0.007). The finding implies that plant population with inadequate number of productive primary branches could lead to increase yield gap.
According to TaCRI (2020), the average number of productive primary branches per tree are 30 to 38 trees.
The findings also indicated that, infestation of coffee diseases due to poor control of CBD and CLR have positive coefficient (4.684) and statistically significant at P<=0.000.
Fungicides are used to control major coffee diseases which include CBD and CLR but the price of fungicides were reported to be high and farmer fail to purchase and apply the right dosage at the recommended rate and time. According to (Kilambo et al., 2015) without control of these diseases crop yield will be reduced by 50 to 0% which is the big crop loss.
Pest and diseases of coffee reduce yields and sometimes killing trees (Magina, 2011).
Fungicide application is required to control major coffee diseases and contribute to increase coffee yield hence reduce the yield gap (TaCRI, 2011).
The findings also indicated that, the coefficient of the amount of fertilizer applied have was positive (0.073) and statistically significant at 0.008 level increase coffee yield gap.
The findings imply that, smallholder farmers do not apply sufficient amount of fertilizer or either they don't apply at all hence cause the increased in yield gap. According to (Maro, 2014 andTaCRI, 2011), application of recommended fertilizer would contribute to increase coffee yield hence reduce the yield gap. The reason for low fertilizer application among smallholder farmers is due to high price of fertilizer. The possible alternative that farmer can afford and adopt is integrated soil fertility management (ISFM) proposed by (Maro, 2014 andTaCRI, 2011). Meanwhile the government should take the issue of input availability to farmers for reasonable price a s important agenda and put strategies to ensure that resource poor smallholder farmers who are productive get adequate amounts of quality inputs at the right time and reasonable price to obtain high yields. Technology adoption index (20.695), improved varieties (1210.367) and traditional varieties (1267.929) were positive and statistically significant at 1 % causing increase in coffee yield gap.

Conclusion
The aim of this study is to understand factors causing yield gap among smallholder farmers in Mbinga and Mbozi districts. The study show, smallholder farmers with improved coffee varieties gain relative high coffee yield than farmers with traditional coffee varieties. The findings also show that, the yield gap among smallholder farmers is caused by number of factors including, lack of visits by extension officers to farmers, lack or insufficient application of fertilizer, poor or lack of control of coffee pests and diseases, type of coffee varieties planted by farmers and the extent of technology adoption by smallholder farmers in the study area.

Recommendations
In order to minimize coffee yield gap among smallholder famers and increase coffee production in Tanzania there is a need to strengthen extension services to help in dissemination of appropriate knowledge and skill to farmers to achieve high yield.
Subsidized coffee input such as fertilizer, pesticides and fungicide required to smallholder farmer so as they can afford to apply the recommended rates of fertilizer, fungicides and pesticide to attain high yield further more progressive adoption of improved varieties and implementation of good agricultural practices is recommended to farmers.

Authors' contributions
The first author handled the data analysis and discussion of results. Other authors supervised the writing of the manuscript, proofread to ensure quality of the research as well as contributed to the revision of the study for publication. All authors read and approved the final manuscript. Sokoine University of Agriculture (SUA), Tanzania. She has majored in economics of food safety standards in agro-food exports for developing countries. She has worked and done consultancy and assists SMEs in product development and processing specifically on dairy products, cooking oil, fish and vegetables.