The methodological heart of the study is a generalized double-hurdle model with correlated errors, an econometric framework designed precisely for situations where a decision unfolds in two sequential stages. In the first hurdle, a farmer decides whether to adopt an improved technology at all. In the second, the farmer decides how much land to allocate to it. What makes the generalized specification powerful is that it allows the errors from these two stages to be correlated, capturing unobserved factors, such as innate risk tolerance or motivation, that influence both decisions simultaneously. The data confirmed this suspicion emphatically: the correlation between the two hurdles was positive and statistically significant at ρ = 0.42 with a p-value below 0.01, revealing substantial selection bias. In plain terms, farmers who choose to adopt are not a random subset of the population, and analyzing their intensity of use without correcting for that self-selection would produce misleading conclusions about what drives how much land they commit.

The adoption rates the study documented vary dramatically by technology type, and this variation itself tells a story. Inorganic fertilizer, a technology that has been promoted in Ethiopia since the 1960s and is relatively easy to apply in a single season, had been adopted by 78 percent of surveyed households. Improved crop varieties, by contrast, reached only 23 percent, a striking gap given that improved seeds are often touted as the centerpiece of African agricultural transformation. The disparity likely reflects differences in the nature of the decisions involved. Fertilizer is an annual, reversible purchase, whereas switching varieties may require new knowledge, different agronomic practices, and trust in seed systems that farmers may not yet possess. The finding suggests that decades of dissemination have succeeded in normalizing some inputs while leaving the more knowledge-intensive technologies far behind.

When the researchers unpacked which factors drive the initial adoption decision, household-level characteristics emerged as the dominant forces. Education significantly increased the probability of adoption, as did years of agricultural experience, which likely reflects accumulated learning about how new technologies perform under local conditions. Participation in social networks, such as farmer associations and community groups, also played a meaningful role, consistent with the idea that information travels through social ties and that seeing neighbors succeed with a technology reduces the perceived risk of trying it. These findings paint adoption as, at its core, a behavioral and informational threshold: farmers need to know about a technology, understand it, and feel confident enough to cross the line into first use.

The drivers of adoption intensity, however, told a strikingly different story. Once a farmer had decided to use an improved technology, the amount of land devoted to it depended predominantly on resource endowments rather than on education or social connections. Livestock holdings, which serve as a store of wealth, a source of draft power, and a buffer against shocks, emerged as a key determinant of intensity, as did overall farm size. Farmers with larger plots and more assets can afford to experiment more broadly, absorb the risk of a poor season, and finance the input requirements that come with expanding the area under improved technologies. In other words, the binding constraint shifts from the mind to the wallet: information barriers govern entry, but resource barriers govern scale.

Between these two poles, institutional factors proved to be the connective tissue operating across both stages of the decision process. Access to agricultural extension services, the government-run advisory system that connects farmers with technical experts, significantly influenced both whether farmers adopted and how intensively they used the technologies. Access to credit showed the same dual importance, easing the cash constraints that accompany input purchases at the adoption stage and enabling larger allocations of land at the intensity stage. This dual role gives extension and credit systems a rare policy leverage: investments in these institutions pay off twice, once at the threshold of first use and again in the depth of subsequent engagement.

Perhaps the most policy-relevant contributions come from the study’s heterogeneity analyses, which disaggregated the sample to reveal how these relationships differ across farmer types. Credit constraints, the authors found, bind more tightly for small farms and for female-headed households, groups that typically possess fewer collateral assets and face greater difficulty accessing formal lending. This means that a uniform credit program, however well designed on paper, will systematically underperform for precisely the farmers who are most resource-constrained. Meanwhile, extension services appeared to substitute for formal education among less educated farmers, suggesting that hands-on advisory support can compensate for gaps in schooling. For female-headed households and smallholders, the implication is that targeted instruments, such as collateral-free credit schemes or intensive extension contact, may be necessary to level the playing field rather than simply expanding existing programs.

The researchers complemented the double-hurdle analysis with a Multivariate Probit model to account for the joint adoption of technology packages, recognizing that farmers rarely make decisions about fertilizer, seeds, and other inputs in isolation. These technologies are complements and substitutes in complex ways: a farmer investing in an improved variety may simultaneously demand more fertilizer to realize its yield potential, or may face a budget constraint that forces a choice between the two. Modeling these interdependencies matters because policies evaluated technology-by-technology can miss the synergies, or the trade-offs, that shape real farm decisions. Sub-sample analyses and robustness checks, including a test for omitted variable bias following Oster’s method, reinforced the credibility of the central findings.

The broader significance of the work lies in its challenge to how agricultural transformation is typically measured and targeted across sub-Saharan Africa. Much of the existing literature treats adoption as a binary outcome, a farmer either uses the technology or does not, and designs interventions accordingly. Molla and Tilahun’s results show that this framing conflates two distinct problems. A region where most farmers have tried fertilizer but few have embraced improved varieties faces an information and trust problem, not merely a resource problem, and vice versa. Policies that address only entry-level behavioral barriers, such as demonstration plots and farmer field days, will boost adoption rates but may leave intensity stagnant if livestock holdings and farm size remain the binding constraints. Conversely, credit expansions alone will do little for farmers who have not yet been convinced to try.

For a region characterized by diverse agro-ecologies and persistent food insecurity, and for the millions of smallholders across climate-vulnerable settings who face similar dilemmas, the message is one of tailored precision. Scaling up improved crop technologies, the authors argue, requires interventions that simultaneously address the behavioral barriers at the point of entry and the resource constraints that limit how deeply farmers can engage once they are in. The study received no external funding and was approved by the Department of Economics at Debre Markos University’s Burie Campus, with informed verbal consent obtained from all participants in Amharic. As Ethiopia and its neighbors confront mounting pressure to raise yields on shrinking land in a warming climate, the finding that adoption is a multi-phase process, with distinct levers at each phase, offers a roadmap that is more nuanced, and ultimately more actionable, than the one-size-fits-all dissemination campaigns of the past.

Subject of Research: Drivers of improved crop technology adoption and use intensity among smallholder farmers in Ethiopia's Amhara Region

Article Title: Unpacking the drivers of improved crop technology adoption and intensity among smallholder farmers in Ethiopia’s Amhara Region

Article References: Molla, M., & Tilahun, W. (2026). Unpacking the drivers of improved crop technology adoption and intensity among smallholder farmers in Ethiopia’s Amhara Region. Discover Sustainability. https://doi.org/10.1007/s43621-026-05011-w

Image Credits: AI Generated

DOI: 10.1007/s43621-026-05011-w

Keywords: technology adoption, smallholder agriculture, double-hurdle model, Ethiopia, Amhara Region, improved crop varieties, inorganic fertilizer, agricultural extension, credit constraints, adoption intensity, agricultural economics, food security

The analysis, published in Discover Agriculture by Ibrahim L. Kadigi of Mbeya University of Science and Technology and Lekumok Kironyi of Mwalimu Nyerere University of Agriculture and Technology, draws on the 2019/20 National Sample Census of Agriculture, Tanzania’s most comprehensive agricultural survey. The census, run by the National Bureau of Statistics, covered thousands of villages across the mainland and captured detailed records of household crop production, landholdings, and input use. From an initial pool of 2,830 sorghum-producing households, the researchers assembled an analytical dataset of 2,000 farms with complete harvest information, spanning all seven of the country’s agro-ecological zones.

What the data revealed about fertilizer adoption was stark. Only 91 farms, or 4.6 percent of the sample, reported applying organic fertilizers such as manure to their sorghum, and just 59 farms, about 3 percent, used inorganic mineral fertilizers. The remaining 92 percent of farms grew the crop on whatever nutrients the soil still held. Adoption was heavily concentrated in the Southern Highlands, where 48 of the 59 inorganic fertilizer users were found, a pattern the researchers attribute to better market integration and stronger extension support. In the Central Zone, the heart of Tanzanian sorghum production with more than 255,000 hectares under the crop, not a single sampled farm reported using mineral fertilizer.

To compare productivity across these three farming systems, the team went beyond simple averages. They employed a two-stage Monte Carlo simulation framework, a stochastic technique that converts observed farm-level yields into full probability distributions. In the first stage, yields for each fertilizer category were parameterized as empirical distributions and resampled 500 times using Latin Hypercube Sampling, an efficient method that preserves the statistical properties of the parent data with relatively few iterations. In the second stage, a multivariate empirical procedure introduced correlated randomness across the three systems, ensuring that simulated outcomes respected the real-world covariance structure of the data rather than producing implausible values such as negative yields.

The simulation was validated in two rounds against the observed census data, with goodness-of-fit tests and probability density function overlays confirming that the simulated distributions reproduced the means, spreads, and extremes of actual harvests. The researchers are careful to note what the method cannot do: because fertilizer use was not randomly assigned, the results describe associations between fertilizer categories and yield outcomes, not causal effects. Farms that choose to fertilize may differ in soil quality, management skill, or access to markets in ways the census cannot fully capture. Still, the scale of the differences is striking.

Farms using no fertilizer averaged 1.30 tonnes per hectare, with a minimum observed yield of just 0.49 tonnes. Organic fertilizer users averaged 1.38 tonnes per hectare, a marginal gain accompanied by greater variability, with a coefficient of variation of 56.4 percent compared with 47.2 percent among non-users. The researchers link this inconsistency to the uneven nutrient density and variable quality of manure and compost, which depend on storage, handling, and application rates that few farmers can control precisely. Inorganic fertilizer users, by contrast, averaged 1.94 tonnes per hectare, nearly 50 percent above non-users, with a higher minimum yield of 0.66 tonnes per hectare that suggests mineral nutrients also cushion farmers against the worst outcomes.

The probability analysis sharpened the picture further. Among non-users, 38 percent of farms fell below the national average yield of 1.0 tonne per hectare, and only 10 percent exceeded the 2.0-tonne benchmark widely regarded as the standard for economically viable sorghum production. Organic fertilizer users fared no better in the short term: 41 percent fell below 1.0 tonne, and just 8 percent cleared 2.0 tonnes, a result consistent with evidence from Kenya, Ethiopia, and Burkina Faso showing that organic inputs build soil structure and fertility over multiple seasons but rarely deliver immediate yield gains on their own. Inorganic fertilizer users told a different story. Only 19 percent fell below 1.0 tonne, while 30 percent exceeded 2.0 tonnes, a threefold increase over non-users.

When the bar was raised to the aspirational threshold of 3.0 tonnes per hectare, aligned with global productivity standards and Tanzania’s national goal of doubling cereal output by 2030, the gap widened further. Ninety percent of non-users and 92 percent of organic users remained below 2.0 tonnes, and only 1 percent and 4 percent respectively surpassed 3.0 tonnes. Among inorganic fertilizer users, however, 18 percent, nearly one farm in five, achieved yields above 3.0 tonnes, and a further 12 percent landed between 2.0 and 3.0 tonnes. The authors describe this as a tiered performance gradient: non-users trapped at subsistence levels, organic users marginally and inconsistently better, and mineral fertilizer users shifted decisively toward surplus production with commercial potential.

The findings arrive with important caveats that the authors themselves emphasize. The census does not record fertilizer formulations, nutrient composition, application rates, or timing, so the study cannot say which products or practices drive the gains. It also cannot identify farms applying organic and mineral fertilizers together, leaving integrated soil fertility management, the approach most agronomists consider optimal, empirically untested here. The small number of fertilizer users in the sample limits precision, and simulated iterations, however sophisticated, are not a substitute for additional field observations. Confounding factors such as improved seed varieties, rainfall, and soil type may contribute to the yield differences observed.

Even so, the policy implications are difficult to ignore. Only about one in twelve Tanzanian sorghum farms uses any fertilizer at all, and the yield gap between fertilized and unfertilized fields mirrors patterns documented across Ethiopia, Nigeria, Ghana, and Zimbabwe. The study recommends expanding access to inorganic fertilizers through targeted subsidies and improved distribution, strengthening extension services so that farmers apply nutrients correctly, and integrating mineral inputs with organic amendments to secure both immediate yields and long-term soil health. For a crop that anchors food security in Tanzania’s drought-prone center, where sorghum occupies nearly a quarter of cultivated land in regions like Singida, closing even part of this gap could transform a subsistence staple into a reliable source of food, income, and climate resilience.

Subject of Research: The contribution of organic and inorganic fertilizer use to sorghum productivity in Tanzania's rainfed smallholder farming systems

Article Title: Organic and inorganic fertilizer use and sorghum productivity in Tanzania’s rainfed farming systems using national sample census of agriculture data

Article References: Kadigi, I. L., & Kironyi, L. (2026). Organic and inorganic fertilizer use and sorghum productivity in Tanzania’s rainfed farming systems using national sample census of agriculture data. Discover Agriculture, 4(1), Article 312. https://doi.org/10.1007/s44279-026-00777-z

Image Credits: AI Generated

DOI: 10.1007/s44279-026-00777-z

Keywords: sorghum, Tanzania, fertilizer, inorganic fertilizer, organic fertilizer, crop yield, smallholder farming, rainfed agriculture, Monte Carlo simulation, food security, soil fertility, National Sample Census of Agriculture