Meta-Analysis on Farmers’ Adoption of Agricultural Technologies in East Africa: Evidence from Chinese Agricultural Technology Demonstration Centers

Abstract

Farmers’ low agricultural productivity in East African countries is attributed to among other factors, the low application of modern agricultural technologies. Using meta-analysis this study intended to examine the determinants of adoption of agricultural technologies in East African countries recipients of Chinese Agricultural Technology Demonstration Centers (ATDCs). A comprehensive review employed to gather 22 studies addressing the determinants of adoption of agricultural technologies such as improved varieties, agricultural mechanization, soil conservation, and animal breeding from Ethiopia, Tanzania, Uganda, and Rwanda. The collected data were analyzed using STATA 17 software. The study used a random-effect model to estimate the determinants of agricultural technologies adoption. The findings demonstrated that access to credit, age, education, farming experience, farmer group membership, household size, and off-farm income have a significant influence on farmers ‘adoption of improved varieties. Age, farm size, and education have a significant influence on the adoption of agricultural mechanization. Soil conservation practices are influenced by age, household size, and sex. Moreover, age, education, extension services, and farming experience have a significant impact on the adoption of animal breeding. It is recommended that East African governments, policymakers, and agricultural practitioners to focus on these common variables when planning for the intervention of new agricultural technologies.

1. Introduction

In the 1960s, as the Asian food crisis grew worse, the Green Revolution applied contemporary science to the problem. The Green Revolution’s exceptional speed and magnitude in addressing the food crisis contributed to a notable reduction in poverty levels and the start of wider economic growth in numerous Asian countries [1]. However, global food insecurity persists as a concern despite the success of the Green Revolution. A substantial portion of the rural population in developing nations, especially in South Asia and Sub-Saharan Africa still relies on unproductive agricultural practices and lives in poverty [2].

Moreover, the agricultural technologies emphasized during the Green Revolution are regarded as a solution to challenges of world hunger and malnutrition with the purpose of increasing food production [3]. During the Green Revolution era, technological development was primarily focused on enhancing food production, for example, fertilizers to enhance soil fertility, irrigation systems to facilitate the distribution of water essential for crop growth, pesticides to protect crops from pests and diseases, breeding programmers to produce high-yielding crop varieties, and agricultural machinery to increase the speed and scale of farming activities [3]. It is of paramount importance to enhance agricultural production in emerging nations in order to address the issues of population growth and food security [4]. Consequently, agricultural research and technology are crucial for increasing production, reducing poverty, and meeting food demands [5].

Since the establishment of the People’s Republic of China in 1949, the country has been engaged in providing foreign aid, thereby acquiring an abundance of experience [6]. The first initiatives in the field of agricultural cooperation between China and Africa were launched in the late 1950s. In the initial stages of its engagement with African nations, China deployed agricultural specialists and constructed agricultural infrastructure as part of its customary development assistance programs [7]. From the 1950s onwards, China has constructed several agricultural facilities including farms, rural and farmland water conservation infrastructure, agricultural technology experimental and extension stations, agricultural technology demonstration centers, and other projects in Africa [8].

In addition to supporting local construction projects in Africa, China provides annual agricultural aid to African countries including grain, improved seeds, fertilizers, pesticides, small and medium-sized agricultural machinery, and agricultural processing equipment. Undeniably, China’s efforts have had an impact on China’s and the world’s food security. Accordingly, China’s assistance to Africa in agricultural cooperation has contributed to the development of African countries and the advancement of China-Africa cooperation [9]. Similarly, Chen and Guo [10] reported that in the mid-1990s the Chinese government adopted an international cooperation strategy known as the “going out” strategy to support its outward economic development, and the agricultural sector is an important aspect of this strategy.

In this epoch, the Agricultural Technology Demonstration Centre (ATDC) has been proposed as an alternative model to achieve sustainability in Chinese foreign aid to African countries [11]. First introduced at the Third Forum on China-Africa Cooperation (FOCAC) Summit in Beijing in 2006, ATDCs aim to transfer Chinese agricultural technology and development experience to support agricultural development and address food security issues in Africa. The agricultural technology demonstration centers mark a change and evolution in China’s foreign aid strategy [12]. Also, ATDCs significantly increased the harvest and production levels of food crops, reduced commercial dependence, improved grain yields, increased local incomes, and improved producer prices and output values of demonstration grain crops in recipient countries [13]. Correspondingly, the highlighted objectives of the Agricultural Technology Demonstration Centers (ATDCs) were: (i) to contribute to the achievement of the country’s strategic diplomatic goals, (ii) to enhance the recipient’s agricultural development capacity and food security, (iii) to promote China’s agricultural “going out” and “bringing in”, and (iv) to achieve sustainable development [14].

It is prominent that the agricultural sector contributes significantly to the economies of developing countries, supporting their rural development, export earnings, and food security [15]. Specifically, in East Africa the agricultural sector employs 75% of the labor force, demonstrating the importance of the sector in reducing poverty and creating employment opportunities across countries [16]. However, agricultural productivity in East African countries remains low and the countries have fallen behind in terms of total and per capita food production [17].

To date, the East African countries recipients of Chinese agricultural technology demonstration centers (ATDCs) include Ethiopia, Tanzania, Uganda and Rwanda. The areas of cooperation of Chinese ATDCs in these East African countries include vegetables, and livestock (e.g., pigs, cows and chickens), cereals (e.g., rice, maize, soybeans), flowers, freshwater aquaculture, mulberry plantation and silkworm rearing, jun-cao cultivation and water conservation [14]. In the region, farmers are discouraged from adopting agricultural technological innovations due to low output prices and high risks associated with farming, which together have contributed to the persistence of rural poverty [18]. Similarly, the difficulties faced by East African farmers in sustaining their livelihoods have increased, and as agriculture is a source of food security and livelihoods for around 80% of the population, the impact could be disastrous [18,19].

Indeed, the impact of new agricultural technologies such as improved farming practices, inputs and related commodities such as crop protection, improved crop varieties, fertilizers, and irrigation on household food security will become apparent when modern technology is widely adopted [20]. Therefore, this paper aims to examine the determinants of agricultural technology adoption in East African countries that are recipients of Chinese Agricultural Technology Demonstration Centers (ATDCs) by compiling previous empirical studies to arrive at an average true effect size that represents the determinants of agricultural technology adoption.

2. Methodology

2.1. Data Source

This study is centered on the retrieval and selection of the previously published empirical studies from the English-language database Google Scholar and Web of Science by typing “adoption”, “technology”, “technology adoption”, “adoption of improved seed varieties”, “adoption of agricultural technology”, etc. Figure 1 illustrate a literature review process. At first, a total of 151 studies addressing the determinants of agricultural technology adoption were considered from the literature review, out of 151, 118 studies were excluded for not reporting results using the logistics regression model as finding an effect size measure that works for numerous methodologies (e.g., linear probability, logistic, probit model, etc.) at the same time is not feasible. Accordingly, 33 titles presenting results using the logistic regression model were screened. However, 11 of the 33 studies that reported results without odds ratios and standard errors were excluded. Subsequently, 22 studies that reported results using a logistic regression with odds ratios and standard error were included in this meta-analysis.

2.2. Data Coding

In meta-analysis the coding sheet’s design is the first step in the coding process [21]. According to Gujarati [22] the methods for estimating the qualitative response of dummy dependent variables are the linear probability model (LPM), the logit model, and the probit model. In this study, the effect size is taken from the studies that used the logistic regression model rather than the linear probability model, and the probit regression model. Data coding was conducted following the initial selection and screening of studies to be included in the meta-analysis. The coding included the column for the name of author(s), study area, sample size, study year, type of technology, and effect sizes. The outcome data taken from the primary research papers included odds ratios and standard errors. The characteristics of studies included in this meta-analysis are listed in

Table A1. Characteristics of studies included to estimate the determinants of agricultural technologies adoption in the selected East African countries.

NO.	Author(s)	Country	Study Year	Model Used	Sample Size	Categories of Technologies
1	Lugamara et al. [40]	Tanzania	2021	Logit	400	Improved variety
2	Kabanyoro et al. [36]	Uganda	2013	Logit	171	Improved variety
3	Gecho, Y. & Punjabi, N. K. [37]	Ethiopia	2011	Logit	150	Improved variety
4	Salum, A. K. [57]	Tanzania	2016	Logit	120	Improved variety
5	Hassen, N. A. [33]	Ethiopia	2019	Logit	180	Improved variety
6	Hagos et al. [32]	Ethiopia	2018	Logit	150	Improved variety
7	Milkias, D. K. [58]	Ethiopia	2020	Logit	154	Improved variety
8	Teshome et al. [38]	Ethiopia	2019	Logit	120	Improved variety
9	Negese, T. [59]	Ethiopia	2020	Logit	355	Improved variety
10	Mentire, W. & Gecho, Y. [42]	Ethiopia	2016	Logit	124	Improved variety
11	Ketema et al. [60]	Ethiopia	2021	Logit	129	Improved variety
12	Mmbando, F. E. & Baiyegunhi, L. J. S. [34]	Tanzania	2016	Logit	160	Improved variety
13	Mwatawala et al. [31]	Tanzania	2022	Logit	166	Improved variety
14	Mwanja et al. [28]	Uganda	2016	Logit	140	Improved variety
15	Rwebangira et al. [47]	Tanzania	2022	Logit	180	Agricultural mechanization
16	Mhango, S. S. [52]	Tanzania	2023	Logit	299	Agricultural mechanization
17	Dawud, T. & Jianjun, L. [61]	Ethiopia	2021	Logit	137	Agricultural mechanization
18	Mengie, B. [35]	Ethiopia	2023	Logit	330	Soil conservation
19	Nahayo et al. [39]	Rwanda	2016	Logit	712	Soil conservation
20	Hawas, L. D. & Degaga, D. T. [41]	Ethiopia	2023	Logit	400	Soil conservation
21	Ingabire, et al. [29]	Rwanda	2018	Logit	120	Animal breeding
22	Jebessa et al. [48]	Ethiopia	2023	Logit	266	Animal breeding

 Appendix A.

2.3. Data Analysis

Random-Effect Model

To examine the determinants of agricultural technology adoption, this study employed a random-effect model to obtain a mean of true effect sizes. The aim of the random-effect meta-analysis is to draw conclusions about the research population of from the sample of studies that were included in this study. Borenstein et al. [23] reported that when we choose to include a set of studies in a meta-analysis, we presume that the studies are sufficiently similar to warrant a synthesis of the data; however, there is typically no reason to assume that the studies are “identical” in the sense that the true effect size is precisely the same across all of the studies. Thus, according to Jain et al. [24], the random-effect model assumes that the study i’s treatment effect θi_R is an estimate of its own true treatment effect θiR, with variance viR.

θi_R~N[0, ViR] and θi_R~N[0, δR²]

(1)

where δR² represents the between study variation; hence the random-effect model is further written as:

θi_R = θ_{R +} µ_iR + e_iR

(2)

where θi is the observed effect in the study i; θ_R is the common estimated true effect; and eiR~N[0, ViR] and µiR~N[0, δR²] describe the within and between the study variation, re-spectively. Guo et al. [25] reported that within-study and between-study variance is the cause of the observed variation in the random-effects model. On the other hand, Xie et al. [26] pointed out that by giving small weights to studies with large sample sizes and large weights to studies with small sample sizes, the random-effects model eliminates some of the consequences of heterogeneity by using the inverse of the sum of the within-study variance and the between-study variance as weights.

According to Borenstein et al. [27] under the random-effect model the weight assigned to each study is:

$$W_i^{*}={\displaystyle \frac{1}{V_{Yi}^{*}}},\mathrm{and}{V}_{Yi}^{*}=V_{Yi}+T^2$$

(3)

where V*_Yi is the within-study variance for study i plus the between studies variance T².

Then, the weighted mean, $M^{*}$, is calculated as

$$M^{*}=\frac{{\sum }_{i=1}^K{W}_i^{*}{Y}_i}{{\sum }_{i=1}^K{W}_i^{*}}$$

(4)

where k is the number of studies, $W_i^{*}$ is the weight related to the study i and Y_i is the effect size collected from the study i.

The description and summary of common variables taken from empirical studies to be used in this meta-analysis are presented in

Table 1. Descriptions and summary of common variables used in empirical studies.

Variables	Descriptions	References
Acces to credit	1 if farmers accessed agricultural credit; 0 if otherwise	Mwanja et al. [28]; Ingabire et al. [29]
Age	Age of household head (years)	Mwalongo et al. [30]; Mwatawala et al. [31]
Distance to market	Distance to the market (continuous)	Hagos et al. [32]; Hassen [33]
Education level	Years spent in school	Mbando & Baiyegunhi [34]; Mengie [35]
Extension services	1 if farmer accessed extension services; 0 if otherwise	Mwanja et al. [28]; Kabanyoro et al. [36]
Farming experience	Number of years engaged in farming	Gecho & Punjabi [37]; Teshome et al. [38]
Farmer group membership	1 if farmer is a member of farmer group; 0 if otherwise	Mwalongo et al. [30]; Nahayo et al. [39]
Farm size	Total area planted (continuous)	Mbando & Baiyegunhi [34]; Mengie [35]
Household size	Number of individuals in a household	Lugamara et al. [40]; Hawas & Degaga [41]
Off-farm income	1 if farmer engaged in off-farm activities; 0 if otherwise	Hagos et al. [32]; Hassen [33]
Sex	1 if famer is Male; 0 if otherwise	Mwanja et al. [28]; Mwalongo et al. [30]
Agricultural training	1 if the farmer participated in agricultural technology training; 0 if otherwise	Hagos et al. [32]; Mentire & Gecho [42]

and Figure 2 demonstrates descriptive statistics of the categories of agricultural technology examined in the study.

3. Results and Discussion

Estimated Overall Effect Size Results

Table 2 presents results for the mean effects size (ES), 95% CI, and Z statistics. The overall result of the Egger test was not significant (p > 0.05), indicating no evidence of publication bias. Regarding the determinants of agricultural technology adoption, the current study demonstrated that access to credit has a significant influence on the adoption of improved seed varieties. Consistently, Mmbando and Baiyegunhi [34] reported that access to credit is significantly associated with farmers’ adoption of improved maize varieties in Hai District, Tanzania. The possible explanation is that credit enables farmers to increase the production of their land as it provides them with access to agricultural inputs allowing them to receive more benefits. This study also indicated that age has a significant effect on the adoption of all categories of agricultural technologies. This variable is linked to the learning process of the family and the management of their agricultural practices as a whole. Although this finding concurs with previous empirical studies by Quaye et al. [43], Fadeyi et al. [44], and Addison et al. [45]; it differs from the argument of Nguyen-Van and To-The [46], which reported that older farmers are less likely than younger farmers to adopt new technologies.

Moreover, this study indicated a significant association between farmer’s education level and the adoption of improved seed varieties, agricultural mechanization, and animal breeding. This finding is supported by the previous studies by Mwalongo et al. [30], Men-tire and Gecho [42], Rwebangira et al. [47], and Jebessa et al. [48]. Similarly, Kassie et al. [49] conveyed that farmers with some form of formal or informal education adopt new technology more quickly than illiterate farmers. The present study also confirmed that access to extension services has a significant influence on the adoption of animal breeding. This result corresponds with the previous study by Ingabire et al. [29]. However, it contradicts a prior meta-analysis by Ruzzante et al. [50], which reported that access to extension services is significantly related to the adoption of more than one category of agricultural technology.

Furthermore, the current findings revealed that farming experience was significant in influencing the adoption of improved seed varieties, and animal breeding. This was evidenced in a study by Gecho and Punjabi [37]. Farming experience is anticipated to influence the adoption of agricultural technologies because experienced farmers may have additional skills and access to new information on improved technologies [51]. The current study also showed that farmer group membership has a significant effect on the adoption of improved seed varieties. This result is consistent with the previous studies by Mwalongo et al. [30], and Teshome et al. [38]. Farmers who are members of farmer’s groups and have been at the forefront in adopting certain practices and technologies may provide successful stories of their experience, encouraging others to adopt these particular practices and technologies.

Nevertheless, the present study indicated that farm size has a significant impact on farmers’ adoption of agricultural mechanization. Such results were also reported in the studies by Rwebangira et al. [47], and Mhango [52]. It is hypothesized that farmers with larger farms seek to adopt new technology in large quantities for efficiency reasons [53]. This study also demonstrated that household size was significantly associated with the adoption of improved seed varieties, and soil conservation practices. This result agrees with the previous studies by Mengie [35], and Gecho and Punjabi [37]. Likewise, Elemineh et al. [54] reported that farmers with larger household sizes are more likely to adopt new agricultural technologies especially when these innovations require a higher financial investment.

On the other hand, the current study demonstrated that off-farm income was significant in influencing the adoption of improved seed varieties. This result is consistent with that of Hagos et al. [32]. It is expected that the availability of off-farm income will have a positive effect on the decision to adopt technology, as households engaged in off-farm activities earn more money to purchase other important agricultural inputs [55]. The present results also indicated a significant relationship between sex and the adoption of soil conservation practices. A similar result was reported in a prior study by Mengie [35]. Moreover, according to Gebrea et al. [56] males are more open to new experiences than females.

Table 2. Estimated overall effect size using restricted maximum–likelihood model.

Categories of Technologies	Determinants Factors	ES	95% CI	Z
Improved varieties	Access to credit	2.260 ***	[0.687, 3.834]	2.82
	Age	1.023 ***	[1.000, 1.046]	86.33
	Distance to market	0.346	[−0.276, 0.968]	1.09
	Education level	2.645	[1.075, 4.215]	3.30
	Extension services	4.351	[−0.618, 9.320]	1.72
	Farming experience	0.961 ***	[0.846, 1.075]	16.42
	Farmer group membership	2.273 ***	[1.569, 2.976]	6.33
	Farm size	18.692	[−11.821, 49.205]	1.20
	Household size	1.116 ***	[0.834, 1.399]	7.75
	Off-farm income	1.817 ***	[0.558, 3.075]	2.83
	Sex	4.196	[−0.131, 8.524]	1.90
	Agricultural training	0.952	[−0.787, 2.690]	1.07
Agric mechanization	Age	0.980 ***	[0.952, 1.009]	67.15
	Farm size	0.697 ***	[0.429, 0.965]	5.10
	Education level	0.974 ***	[0.841, 1.107]	14.35
	Extension services	1.474	[−0.844, 3.792]	1.25
	Sex	4.916	[−2.325, 12.157]	1.33
Soil conservation	Access to credit	0.466	[−3.597, 4.528]	0.22
	Age	1.037 ***	[0.992, 1.082]	45.00
	Education level	6.409	[−4.884, 17.702]	1.11
	Farm size	2.690	[−0.880, 6.260]	1.48
	Extension srvices	6.409	[−4.884, 17.702]	0.98
	Household size	1.103 ***	[1.064, 1.143]	54.86
	Sex	1.101 **	[0.185, 2.016]	2.35
Animal breeding	Access to credit	1.354	[−0.094, 2.801]	1.83
	Age	1.048 ***	[1.015, 1.081]	62.25
	Education level	1.046 ***	[0.603, 1.488]	4.63
	Extnsion services	1.042 ***	[1.007, 1.077]	58.24
	Farming experience	0.940 ***	[0.883, 0.997]	32.43
	Off-farm income	0.534	[−0.248, 1.316]	1.34

Notes: ***, and ** indicate effect size is significant at 1% and 5% significant levels, respectively.

Table 2. Estimated overall effect size using restricted maximum–likelihood model.

Categories of Technologies	Determinants Factors	ES	95% CI	Z
Improved varieties	Access to credit	2.260 ***	[0.687, 3.834]	2.82
	Age	1.023 ***	[1.000, 1.046]	86.33
	Distance to market	0.346	[−0.276, 0.968]	1.09
	Education level	2.645	[1.075, 4.215]	3.30
	Extension services	4.351	[−0.618, 9.320]	1.72
	Farming experience	0.961 ***	[0.846, 1.075]	16.42
	Farmer group membership	2.273 ***	[1.569, 2.976]	6.33
	Farm size	18.692	[−11.821, 49.205]	1.20
	Household size	1.116 ***	[0.834, 1.399]	7.75
	Off-farm income	1.817 ***	[0.558, 3.075]	2.83
	Sex	4.196	[−0.131, 8.524]	1.90
	Agricultural training	0.952	[−0.787, 2.690]	1.07
Agric mechanization	Age	0.980 ***	[0.952, 1.009]	67.15
	Farm size	0.697 ***	[0.429, 0.965]	5.10
	Education level	0.974 ***	[0.841, 1.107]	14.35
	Extension services	1.474	[−0.844, 3.792]	1.25
	Sex	4.916	[−2.325, 12.157]	1.33
Soil conservation	Access to credit	0.466	[−3.597, 4.528]	0.22
	Age	1.037 ***	[0.992, 1.082]	45.00
	Education level	6.409	[−4.884, 17.702]	1.11
	Farm size	2.690	[−0.880, 6.260]	1.48
	Extension srvices	6.409	[−4.884, 17.702]	0.98
	Household size	1.103 ***	[1.064, 1.143]	54.86
	Sex	1.101 **	[0.185, 2.016]	2.35
Animal breeding	Access to credit	1.354	[−0.094, 2.801]	1.83
	Age	1.048 ***	[1.015, 1.081]	62.25
	Education level	1.046 ***	[0.603, 1.488]	4.63
	Extnsion services	1.042 ***	[1.007, 1.077]	58.24
	Farming experience	0.940 ***	[0.883, 0.997]	32.43
	Off-farm income	0.534	[−0.248, 1.316]	1.34

Notes: ***, and ** indicate effect size is significant at 1% and 5% significant levels, respectively.

4. Conclusions and Policy Recommendations

To achieve the second target of zero hunger of the sustainable development goals, the adoption of modern agricultural technologies should not be underestimated. Thus, unlike individual studies, this study used a meta-analysis to examine factors influencing the adoption of agricultural technologies in East African countries that have received Chinese Agricultural Technology Demonstration Centers (ATDCs). Although the majority of average findings are limited by heterogeneity; however, the average results which are relatively large and significant are very compatible with the adoption theory [50]. Similarly, because of the non-random sampling of studies to be included in the meta-analysis, the results presented in this study are not rigorously generalized in all situations, rather, they represent the collective results of agricultural technology adoption articles that have been accessible and published in recent years. Consequently, the current research concludes that the explanatory variables access to credit, age, education, farming experience, farmer group membership, household size, and off-farm income have a significant influence on farmers’ adoption of improved varieties. The adoption of agricultural mechanization was significantly influenced by age, farm size, and education. Furthermore, soil conservation practices are influenced by age, household size, and sex; while age, education, extension services, and farming experience have a significant influence on the adoption of animal breeding.

In light of this study’s conclusions, the following policy recommendations can be put forth: First, East African governments, policymakers, and agricultural stakeholders should work together to create an enabling environment that will enhance the uptake of innovative agricultural practices established in their countries. Second, future agricultural technology transfer programs in East Africa should take into account the common variables influencing agricultural technology adoption discussed in this study. Third, there is a need to have a campaign to demonstrate success stories of the impact of ATDC such as an increased harvest and production levels of food crops, reduced commercial dependence, improved grain yields, increased local incomes, and improved producer prices and output values of demonstration grain crops in recipient countries [13]. Last but not least, this study recommends that the East African Community (EAC) should collaborate well with China to enhance agricultural technology transfer in East Africa.