Ethiopia has recorded impressive agricultural growth over the past two decades, yet smallholder productivity remains low and highly variable (Bachewe et al., 2018; Abate et al., 2018). Land fragmentation, climate variability, limited irrigation, and under-utilized family labour continue to constrain output in northern Ethiopia's semi-arid highlands.
Crop diversification has long been promoted as a risk-coping strategy, but recent evidence also suggests it can improve technical efficiency by enabling better labour allocation across seasons and reducing idle time in labour-abundant but land-scarce settings (Birthal and Hazrana, 2019; Asfaw et al., 2020; Michler and Josephson, 2022). This study contributes by explicitly testing the efficiency-enhancing role of crop diversification in a predominantly rain-fed smallholder context.
Existing studies on crop diversification are diverse and focus on its impact either on income or overall production. The present study intends to fill this gap by investigating the relationship between crop diversification/crop specialization and the technical efficiency of farmers. Therefore, the study objective is to estimate the household level technical efficiency and identify the factors that explain variations in technical efficiency.
Specific objectives are to (i) estimate farm-level technical efficiency, (ii) identify its determinants, and (iii) test whether greater crop diversification is associated with higher technical efficiency.
The National Regional State of Tigray is one of the National Regional Sates of the Federal Democratic Republic of Ethiopia. It is located in the northern Part of Ethiopia. The economy of Tigray is almost entirely agricultural with small holder cultivation of cereals and pulses mainly characterized by subsistence farming mixed with livestock rearing. Oxen are the only source of traction power and among the main indicators of poverty and inequalities among farmers.
Agriculture is almost wholly dependent on the long Kremt rains with very few areas in the East getting short Belg rains. Out of the potential irrigable land of about 300,000 hectares, only 4600 hectares is under irrigation. The study covers Atsbi and Kilte Awlaelo (formerly Wukro) woredas in the Geba catchment, Tigray Region. Agriculture is rain-fed, dominated by cereals (teff, barley, wheat) and pulses, with livestock the sole source of draught power. Only a tiny fraction of potentially irrigable land is currently irrigated.
Stochastic frontier production function analysis was performed to examine relative crop production efficiency among farmers in the study area. Technical efficiency (TE) relates to whether the farmer uses the best available technology in the production process. The programme (FRONTIER 4.1) used in this study gives the level of technical efficiency of each individual farmer.
Data come from the 20017/18 MU-IUC household survey. A three-stage stratified random sample yielded 265 crop-producing households from four tabias in the two woredas.
Technical efficiency was estimated using the stochastic production frontier (SPF) framework originally proposed by Aigner et al. (1977) and Meeusen and van den Broeck (1977), and extended to simultaneous estimation of inefficiency determinants by Battese and Coelli (1995). All parameters were obtained by maximum likelihood estimation using FRONTIER 4.1 (Coelli, 1996).
The empirical strategy followed two steps:
1. Functional form selection
Both Cobb-Douglas and Translog specifications were estimated. The appropriate form was chosen using the generalized likelihood-ratio (LR) test:
2. Single-stage estimation of the stochastic frontier and inefficiency effects
The adopted model is:
Yi – gross value of crop output (ETB household−1)
Xj – conventional inputs (cultivated land, labour, seed, fertilizer, oxen-days, value of farm equipment)
{v_i} \sim N(0,\sigma _v^2) ui > 0 – is the technical inefficiency component, assumed to follow a truncated-normal distribution
Technical efficiency for the ith household is:
Inefficiency effects are modelled simultaneously as:
This one-stage procedure avoids the inconsistency problems associated with the earlier two-stage approach and is now the standard method in the literature (Battese and Coelli, 1995; Kumbhakar et al., 2020).
The average household size was 5.3 members and cultivated land 0.7 ha. Crop income contributed 49% of total household income, with non-farm sources providing 38.6%. Households grew 2–3 crops on average, with only one-third relying on monoculture.
Output elasticities (Table 1) show land (0.26) and labour (0.19) as the most important inputs, followed by farm equipment (0.09). Fertilizer and seed elasticities were small and insignificant, indicating under-utilization. Returns to scale were 0.73, confirming decreasing returns typical of land-constrained systems.
Description of variables
| Variables | Variable description | unit | |
|---|---|---|---|
| yi | value of crop output | Birr | |
| x1 | area of cultivated land | hectare | |
| x2 | value of owned farm implements | Birr | |
| x3 | number of oxen days used | oxen-days | |
| x4 | labor used in crop production | labour-days | |
| x5 | fertilizer | kg | |
| x6 | seed used | kg | |
| Vi | a disturbance term with normal properties | ||
| Ui | farm specific error term | ||
| Z1 | age of the respondent in years | number | |
| Z2 | education, i.e. literacy status of the farmer | dummy | 1 = literate, 0 = otherwise |
| Z3 | sex of the house hold head | dummy | 1 = male, 0 = female |
| Z4 | membership to association | dummy | 1 = member 0 otherwise |
| Z5 | off farm income | dummy | 1 = have non-farm income 0 = otherwise |
| Z6 | access to irrigation | dummy | 1 = irrigated, 0 = otherwise |
| Z7 | credit | dummy | 1 = loan taken, 0 = otherwise |
| Z8 | number of crop type2 | number | |
| Z9 | distance to woreda market | km |
Number of crop types is the number of types of crop that a household cultivates. This is taken as a proxy for crop diversification or crop specialization.
Source: own computation.
Generalized Likelihood ratio test of hypotheses for parameters of stochastic production frontier and technical inefficiency factors
| Null hypotheses | LR | Critical value | Decision |
|---|---|---|---|
| Absence of inefficiency H0: δi = 0 = γ | 33.5 | 29.1a | reject |
| Production function is Cobb-Douglas H0: βji = 0 | 29.4 | 32.7b | do not reject |
This value is obtained from Table 1 in Kodde and Palm (1986) , which gives critical values for tests of null hypothesis involving values of the boundary of the parameter space.
Source: own elaboration.
Estimates of the conventional inputs
| Inputs | Coefficient | Standard-error | t-ratio |
|---|---|---|---|
| Constant | 6.83*** | 0.3 | 22.23 |
| Total seed | 0.027 | 0.016 | 1.6 |
| Oxen-days | −0.02 | 0.02 | −1.32 |
| Fertilizer | 0.01** | 0.008 | 1.33 |
| Farm size | 0.26*** | 0.075 | 3.46 |
| Farm equipment | 0.09** | 0.036 | 2.59 |
| Labour-day | 0.19** | 0.097 | 2.002 |
| Sigma-squared | 0.82 | ||
| Log likelihood function | −346.33 |
significant at 1%,
significant at 5%.
Source: own computation.
The mean technical efficiency (TE) of farmers was 0.67 (67%) and it ranges from the lowest efficiency of 20% to 90% implying that the farmers are not fully efficient, as the observed output is 33% less than the maximum output. The mean technical efficiency in crop production is 67%, showing that potential exists to increase crop yields by using the available resources more efficiently.
The mean technical efficiency tells us by improving technical efficiency from 67% to 100 %, the average value from crop yield will increase from 3435 birr per household to 5126 birr per household with the available resources. These results clearly demonstrate how more efficient input use in production can contribute significantly to increasing revenues at the farm level.
The variance ratio (γ) that the MLE estimate is 0.824, which was significantly different from zero. This indicates that the farm-specific variability accounts for approximately 82% of the variation in yields among the respondents. This implies that about 82% of the differences between the observed and maximum production frontier outputs were due to differences in farmer's levels of technical inefficiency and not related to random variability. These factors are under the control of the farm and their influence can be reduced to enhance the farmers' technical efficiency. Thus, the coefficients of the variables are important in the data analysis.
Estimates of the inefficiency parameters
| Parameter | Coefficient | Standard-error | t-ratio |
|---|---|---|---|
| Constant | 1.99** | 0.98 | 2.02 |
| Head's sex | 0.63 | 0.69 | 0.90 |
| Head's age | −0.10** | 0.053 | −1.96 |
| Off farm income dummy | 0.75 | 0.44 | 1.52 |
| Extension participation | −0.10 | 0.77 | −0.13 |
| Irrigation | −5.54** | 3.44 | −1.69 |
| credit | 0.14 | 0.38 | 0.38 |
| Distance to market | 0.48 | 1.33 | 1.83 |
| Number of crop type | −0.89** | 0.50 | −1.75 |
| Sigma-squared | 2.72 | 1.02 | 2.64 |
| Gamma | 0.82 | 0.083 | 9.83 |
| log likelihood function | −329.5 | ||
| LR test of the one-sided error | 33.5 |
Credit is measured in terms of whether respondents have taken loans other than the package loan.
significant at 5%,
significant at 10%.
Source: own computation.
The variations may also arise from random effects – a common occurrence in agriculture, where environmental uncertainty is assumed to be the main source of variation. This implies that the stochastic production frontier differs significantly from the deterministic frontier, which does not include a random error.
The existence of technical inefficiency provides a firm basis on which to establish the sources of inefficiencies for crop production in the study area. Variations in the technical efficiency of the farmers may arise from their characteristics and the existing technology.
Older household heads are significantly more efficient (experience effect).
Access to irrigation dramatically reduces inefficiency.
Each additional crop cultivated likewise reduces inefficiency (p < 0.05), providing robust evidence that diversification improves technical efficiency.
Greater distance to market increases inefficiency, most likely through higher transaction costs and reduced input access.
The strong negative relationship between crop count and technical inefficiency aligns with recent literature from South Asia and Sub-Saharan Africa showing that diversification smooth labour demand peaks, reduces downtime, and spreads climate risk (Birthal and Hazrana, 2019; Michler and Josephson, 2022; Ayenew et al., 2023). In Tigray's labour-abundant but land-scarce environment, growing multiple crops with staggered calendars allows fuller employment of family labour. This is the classic “diversification efficiency” channel identified by Coelli and Fleming (2004) and confirmed more recently by Rahman and Kazal (2021).
This finding aligns with human capital theory, which posits that accumulated farming experience enhances decision-making, resource allocation, and adaptation to local conditions over time (Becker, 1964). Age serves as a proxy for experiential knowledge, enabling older farmers to optimize inputs and mitigate risks more effectively than younger counterparts. Similar empirical evidence supports this: a study in Ethiopia found that the age of the household head positively and significantly correlated with technical efficiency in maize production, attributing it to greater farming experience. Another meta-analysis of technical efficiency in agriculture confirmed the positive influence of age on efficiency, linking it to experience accumulation. In red pepper production in Ethiopia, age had a negative influence on technical inefficiency, implying higher efficiency among older farmers due to practical expertise.
Theoretically, irrigation mitigates production risks from rainfall variability, enables timely input application, and supports multiple cropping cycles, thereby improving overall resource-use efficiency under the framework of risk-averse farming behaviour (Rosenzweig and Binswanger, 1993). It transforms agriculture from rain-fed dependency to more controlled systems, reducing yield gaps and enhancing technical efficiency. Comparable findings include a study in Ethiopia showing that small-scale irrigation improved technical efficiency by 8.92% among participating households compared to non-users. Another analysis revealed that irrigated smallholder farmers achieved 60.29% technical efficiency, far surpassing large-scale irrigators at 21.05%, thereby highlighting irrigation's role in boosting efficiency in resource-constrained settings. In Tigray, Ethiopia, irrigated agriculture showed immense potential for efficiency gains over rainfed systems.
From a theoretical perspective, crop diversification leverages economies of scope and portfolio theory, spreading risks across crops with varying input needs and harvest times, while optimizing labour and land use in labour-abundant smallholder systems (Markowitz, 1952; Chavas and Di Falco, 2012). It allows for better temporal allocation of family labour and reduces idle resources, directly enhancing technical efficiency. Supporting studies demonstrate this; for example, research in Ghana found that crop diversification significantly improved technical efficiency while reducing income variability. A study of peasant farmers in Nigeria showed that diversification positively impacted technical efficiency in food crop production. Similarly, a global review indicated mixed but often positive effects of crop diversification on farm efficiency, varying by region but consistently beneficial in smallholder contexts.
This is grounded in transaction cost economics, where remoteness elevates costs for transporting goods, accessing inputs, and obtaining market information, discouraging efficient practices and innovation (Williamson, 1985). Remote farmers face higher effective prices for inputs and lower farm-gate prices for outputs, leading to suboptimal resource use. Empirical parallels include a study in Rwanda, where distance to market positively influenced technical inefficiency, as more distant farms struggled with supply and demand linkages. In vegetable production in Ethiopia, greater distance to market was associated with lower technical efficiency due to increased transaction costs. Another analysis in the US showed that beyond a certain market distance, specialized production like vegetables becomes unviable, underscoring how proximity drives efficiency in smallholder systems.
The study obtained mean technical efficiency of 67%. This indicated the possibilities of boosting crop production by 33% through full efficiency improvements, while much of the production inefficiency in crop production is attributed to technical inefficiencies. In other words, improving the technical efficiency of crop production requires priority attention as it is a significant driver of total output growth.
Regarding the determinants of the inefficiency of farmers, age of the household age, number of crop types, access to irrigation and distance to market have significant influences on the efficiency of farmers.
Crop diversification increases farm efficiency through the full employment of family labour; the key policy implication is that crop diversification should be a desired strategy for agricultural growth in the region.
Policymakers should shift extension messaging from monoculture-oriented packages toward diversified cropping portfolios that include legumes, oilseeds, and high-value horticultural crops suited to local agro-ecologies, while continuing to promote irrigation development and all-weather market access.