Exploring the Impact of Improved Maize Seeds on Productivity of Tanzanian Family Farms: A Maize Seed Stochastic Simulation (MaizeSim) Approach

Abstract

Investment in modern agricultural practices (MAPs) is crucial for improving crop productivity and household food availability in developing countries like Tanzania, where agriculture forms the backbone of the economy. This study assesses the impact of improved maize seeds on productivity across Tanzania’s agroecological zones using data from the Tanzanian National Panel Survey (NPS) Wave 5. A stochastic simulation model (a non-parametric model, “MaizeSim”) was employed to account for the inherent variability and uncertainty considerations in maize yields, offering a more accurate representation of outcomes for both improved seed users and non-users. The results reveal that farmers who used improved seeds had a 33% probability of achieving yields above 2 t/ha, compared to only 11% for those using local varieties. Conversely, non-users faced a 65% probability of harvesting below 1 t/ha, while this probability dropped to 38% for improved seed users. Regionally, the highest productivity gains were observed in the Central, Southern Highlands, and Northern Highlands zones, whereas the Eastern Coastal, Southern, and Lake zones experienced minimal benefits. The findings underscore the critical importance of encouraging the adoption of improved seed varieties as a pathway to enhance maize productivity, particularly in regions with favorable agroecological conditions. This study provides valuable insights for the development of the Tanzanian Seed Sector Development Strategy 2030, advocating for policies that promote increased investment in improved maize seeds. The results suggest that sustained application of these seeds, alongside complementary interventions such as agronomic training and improved access to inputs, is essential for improving the productivity and food availability of Tanzanian smallholder farmers. By addressing regional disparities and promoting tailored seed varieties, this strategy could significantly enhance the resilience and productivity of the country’s maize sector.

1. Introduction

Food production in Sub-Saharan Africa (SSA) has exhibited stagnation or decline on a per capita basis for over four decades [1]. This persistent food insecurity is characterized by insufficient food availability, price volatility, high poverty rates, and susceptibility to multifaceted challenges like extreme weather events and climate change [2,3]. The situation is further exacerbated by a rapidly growing population, declining soil fertility, and increasing land limitations in densely populated rural areas (Khumalo, 2021 [2]). A growing consensus acknowledges the need for an SSA-specific green revolution to enhance food production and quality, particularly within the small-scale, low-productivity farms that support two-thirds of the region’s population [1,4]. Improved seeds are certified high-yielding seeds developed through breeding improvements to ensure genetic purity and performance consistency. Recycled seeds are uncertified improved seeds preserved by farmers themselves from their previous harvests for replanting in subsequent seasons replanting.

Despite the challenges faced by smallholder farmers, family farms are undeniably the cornerstone of African agriculture [5]. They provide employment for a large portion of the population, estimated at two-thirds, and cultivate a significant majority of the land, exceeding 60% [6]. These farms are typically small in size, with many in Sub-Saharan Africa being less than one hectare. Their production is diverse, encompassing food and cash crops, livestock, and catering to both subsistence and market needs [7]. Traditional practices are often employed, with limited use of modern advancements like irrigation, fertilizers, and improved seeds [8]. While family farms play a vital role in agricultural development and poverty reduction, their restricted access to resources like high-yielding seeds, essential inputs, markets, climate change, and financing hinders their ability to reach their full potential in terms of production and resilience [9,10,11].

In Tanzania and across SSA, maize remains a cornerstone of food availability and rural livelihoods. Yet, its production continues to face numerous challenges that hinder the realization of potential yields. Research has shown that maize farming in the region is often constrained by factors such as degraded soils, erratic rainfall patterns, limited access to improved seeds and fertilizers, and underdeveloped market infrastructure [12,13]. Social and economic challenges, including land tenure insecurity, high input costs, and poor extension services, further exacerbate low productivity among smallholder farmers [14,15]. Like many farms within SSA, smallholder family farms in Tanzania are dominated by maize cultivation [16]. This crop serves as the primary source of calories and is consumed by over half of the SSA population [17]. Tanzania ranks fourth in maize production within SSA, following South Africa, Nigeria, and Ethiopia, with an output of 5.3 million tons in 2013 [18]. Historically, Tanzania has boasted a national surplus and sees potential for exporting maize to neighboring regions [19]. This has led the Tanzanian Government to prioritize improvements in maize productivity, aiming to establish the country as a regional breadbasket [20].

Traditional farming methods, characterized by low external input use and reliance on outdated seed varieties, often leave farmers vulnerable to environmental shocks, pests, and diseases, thereby widening the yield gap [21,22,23]. Environmental stresses such as droughts and soil nutrient depletion are particularly acute in semi-arid regions, where maize yields are significantly below the regional potential. Furthermore, weather variability, particularly droughts, presents another substantial challenge, especially for smallholder livelihoods [24]. Despite efforts to introduce improved technologies, adoption rates for modern inputs remain modest, reflecting persistent socio-economic and institutional barriers. Therefore, closing the maize yield gap necessitates a greater understanding of how modern interventions, specifically the use of improved seeds and appropriate fertilizer management, can be effectively scaled within diverse agroecological and socio-economic settings. This study contributes to this growing body of research by assessing the probability of achieving potential maize yields under different farming setups in Tanzania.

The influence of improved maize seeds on Tanzanian family farms is a multifaceted issue. While research has largely focused on the potential for increased yield and food availability through adoption [25,26,27], recent studies highlight the importance of socio-economic and environmental factors that can act as barriers to widespread use [28,29]. Kadigi et al. [30] further emphasize the need for risk analysis and management strategies when adopting improved farming practices, which can also be applied to maize seed adoption. However, there is a lack of robust empirical evidence on the intricate interplay between these factors and their impact on adoption and effectiveness. This necessitates exploring stochastic simulation techniques to address the inherent complexity of this issue.

This study employs a comprehensive stochastic simulation analysis to investigate the potential farm and household-level impacts of adopting improved maize seeds in Tanzania (keeping other factors constant). It specifically addresses critical seed policy questions: first, whether improved seeds demonstrably influence yield, and second, the magnitude of their productivity impact in terms of probability distribution. Agriculture is inherently stochastic, meaning factors like weather, pest outbreaks, and market prices can vary unpredictably and directly impact farm productivity. A stochastic model can account for this variability, providing a more realistic picture of potential outcomes [31]. Contributing to the existing impact assessment literature, this study utilizes a rigorous approach, incorporating the inherently stochastic nature of Tanzania’s agricultural sector, to examine the productivity gap between those using improved seeds and those without. Likewise, this study’s findings are more applicable to policy perspectives, as Tanzania is developing the Tanzania Seed Sector Development Strategy 2030 (TSSDS 2030). Among others, the TSSDS aims to streamline the government’s actions to foster food production and availability and improve livelihoods in the country and the region [32].

In order to account for spatial variability in environmental and production conditions that significantly influence maize performance, this study groups areas in Tanzania into agroecological zones, bearing in mind that these zones differ in terms of soil fertility, rainfall patterns, temperature, and elevation, all of which affect the effectiveness of agricultural inputs, including improved seeds. By analyzing impacts across these zones, this study generated more accurate, context-specific insights into where improved seeds perform best and where additional interventions may be needed. This disaggregation allows for targeted policy recommendations, resource allocation, and scaling strategies that consider each zone’s unique challenges and opportunities, ultimately leading to more effective seed sector development and improved food availability outcomes. The subsequent sections of this paper are structured as follows: Section 2 outlines the materials and methods employed to assess the impact of improved seeds on maize productivity, Section 3 presents the key findings (results) of the study, and Section 4 discusses the major results, while Section 5 provides the conclusions.

2. Materials and Methods

2.1. Conceptual Framework

This study delves into the influence of improved maize seeds on Tanzanian family farms. These seeds, often hybrids, hold the potential for increased crop yields, disease resistance, and various other advantages. The diffusion of innovation theory provides a framework to understand how adopting these improved seeds unfolds among Tanzanian farmers (Figure 1). The theory considers factors that influence the spread of new ideas and technologies. It is conceptualized that the introduction of improved maize seeds with specific characteristics tailored to Tanzanian conditions is the independent variable. These could include drought-resistant varieties for arid regions or high-yielding hybrids for areas with better water availability. The impact on Tanzanian family farms is the dependent variable. This encompasses a range of potential outcomes. On the other hand, it is conceptualized that the impact is being mediated by several factors that can moderate the impact of improved seeds on Tanzanian family farms.

2.2. Study Area

This study included all regions of Tanzania, and to capture the yield variabilities, this study categorized Tanzania into seven major agroecological zones, namely: Northern Highlands Zone (NHZ), Eastern Coastal Zone (ECZ), Central Zone (CZ), Southern Highlands Zone (SHZ), Lake Zone (LZ), Plateau Zone (PL). NHZ included three regions, namely: Arusha, Kilimanjaro, and Manyara. ECZ included six regions: Morogoro, Dar es Salaam, Pwani (Coastal Region), Lindi, Mtwara, Zanzibar, and Tanga regions, with SZ having Lindi and Mtwara regions. Central Zone comprises Dodoma and Singida regions, while SHZ comprises Iringa, Mbeya, Songwe, Njombe, and Ruvuma regions. LZ encompasses Geita, Shinyanga, Simiyu, Mwanza, Mara, and Kagera regions, and PL takes Katavi, Kigoma, Tabora, and Rukwa. The agroecological zones used in this study are summarized in

Table 1. Agroecological zones of Tanzania.

Agroecological Zones	Specific Regions
Northern Highlands Zone (NHZ)	Arusha, Kilimanjaro, and Manyara
Eastern and Coastal Zone (ECZ)	Morogoro, Dar es Salaam, Pwani (Coastal Region), Zanzibar and Tanga, Lindi, and Mtwara regions
Central Zone (CZ)	Dodoma and Singida regions
Southern Highlands Zone (SHZ)	Iringa, Mbeya, Songwe, Njombe, and Ruvuma regions
Lake Zone (LZ)	Geita, Shinyanga, Simiyu, Mwanza, and Kagera
Plateau Zone (PL)	Katavi, Kigoma, Mara, Tabora, and Rukwa

.

2.3. Data

2.3.1. National Maize Production and Yield Data

This study utilized data from the Tanzania National Panel Survey Wave 5 (NPS-5) that were collected from December 2020 to January 2022. The NPS is a nationally representative longitudinal survey designed to provide data from the same households over time in an attempt to better track national and international development indicators, understand poverty dynamics, recognise the linkages between smallholder agriculture and welfare, and evaluate policy impacts in the country. The main financiers of the fifth wave of the NPS included the Ministry of Finance and Planning, the European Union (EU), the World Bank/Gates Foundation, and UNICEF. The National Bureau of Statistics has implemented the NPS in collaboration with the Office of the Chief Government Statistician—Zanzibar since its inception in 2008/09.

2.3.2. Farm Survey

Data extraction from the NPS started by sorting appropriate data for simulation. Variable [ag3a_07], which provides data on the main crop cultivated on a particular plot in the long rainy season of 2020, was picked, and only maize was selected. Variable [hh_a01_1] was used to identify and locate the region code and where the farm is located. Variable [ag4a_08] collected information on the type of seed used (whether improved, improved but recycled, or local). Variable [ag4a_27], which provided data on the quantity harvested (production per farm), was also picked to provide the yield per ha. In this way, we found 441 households that used local varieties, 71 used improved but recycled varieties, and 302 used purely improved varieties, as shown in

Table 2. Number of observations for local seed, improved but recycled, and improved seeds.

Regions/Agroecological Zones	Local Seed Users (_0)	Improved but Recycled Users (_1)	Improved Seeds Users (_2)	Total
NHZ	21	5	59	85
ECZ	98	7	19	124
CZ	44	4	22	70
SHZ	69	10	63	142
LZ	88	20	83	191
PL	121	25	56	202
All Regions	441	71	302	814

. Farms that were planted with local or traditional seeds were denoted by _0: this includes All Regions_0, NHZ_0, ECZ_0, CZ_0, SHZ_0, LZ_0, PL_0, while farmers that used improved but recycled seeds were denoted by _1: this includes All Regions_1, NHZ_1, ECSZ_1, CZ_1, SHZ_1, LZ_1, PL_1. Farmers that used improved seed varieties were denoted by _2, and this includes All Regions_2, NHZ_2, ECSZ_2, CZ_2, SHZ_2, LZ_2, PL_2.

2.3.3. A Stochastic Simulation Approach

Non-parametric Monte Carlo simulation procedures outlined by Kadigi et al. [30,33] and Richardson et al. [34] were used to evaluate the yield distributions for each farming system. Following the Monte Carlo procedures, the first step was to develop a maize seed simulation model (MaizeSim) using Simetar software version 5 (Simetar 5) available at www.simetar.com. Since we have 441 representative farms that used local varieties, 71 improved but recycled, and 302 that used purely improved varieties across all regions, the first step was to define, parameterize, simulate, and validate the risky (stochastic) yield for each seed type. After setting the first model, we sorted the data again agroecologically to reconsider the variabilities across the agroecological zones. In this regard, this model had a total of 21 random models (three for all regions under three farming practices and 3 × 6 = 18 for all agroecological zones for all three farming practices). Figure 2 provides the conceptual flow of the MaizeSim Simulation Process.

The MaizeSim simulation process begins by collecting and organizing real-world yield data categorized by farming system and agroecological zone. Using observed historical deviations, stochastic models are built to capture variability. A Monte Carlo simulation, supported by Latin Hypercube Sampling, runs 500 iterations for each farming setup, generating a wide range of possible yield outcomes. The simulation results are analyzed through Stoplight Charts to determine the probability of farms achieving yields above, below, or within target thresholds. The model’s outputs are then validated against historical observations to ensure reliability. This simulation approach enables a robust analysis of farming system performance under uncertainty, informing recommendations for climate-resilient maize production strategies across Tanzania.

The model was built using a multivariate empirical (MVE) distribution described by Richardson et al. [35] and Kadigi et al. [36] to incorporate three farming systems (local seeds, recycled, and improved seeds) and six agroecological zones. The MVE was used because of its ability to account for many variables at once. The residuals (deviations from the observed mean as percent deviations from the mean) from observed/historical yields for each yield of maize varieties were used to estimate the parameters for the MVE yield distribution. An MVE distribution is not only an appropriate tool to account for many variables at once, but it can also eliminate the possibility of values exceeding reasonable values, like negatives in surveyed data [37]. The equation below (Equation (1)) was used to develop the MaizeSim model and

Table 3. Definition of symbols used in the model.

Symbols	Definitions
~	A tilde represents a stochastic variable.
i	Type of maize yield varieties used (Local seeds, Improved recycled, Improved seed).
$\omega$	Represents agroecological zones.
ai	Hectares (ha) allocated for farming system i.
${\tilde{y}}_i$	Stochastic mean yield per ha for farming system i.
${\overline{y}}_i$	Deterministic (mean) yield per ha for farming system i.
Sy	Fraction deviations from the mean or sorted array of random yields for farming system i.
P(Sy)	Cumulative probability function for the Sy values.
CUSDy	Simetar function to simulate correlated uniform standard deviates of random variables.
EMP()	Simetar function used to simulate a stochastic variable (yield).

defines the symbols and signs used in the model.

$${\tilde{y}}_{i,\omega }={\overline{y}}_{i,\omega }\ast \left(1+EMP\left(S_{y,i,\omega },P\left(S_{y,i,\omega }\right),CUS{D}_{y,i,\omega }\right)\right)$$

(1)

The second step was to simulate the developed MaizeSim model in Equation (1) for at least 500 iterations using the Latin Hypercube (LHC) sampling procedures defined by Richardson et al. [38]. The LHC procedure ensures that a sample of only 500 iterations is necessary to reproduce the parent distributions. Since we have 21 random variables with different observations (Table 1) for each agroecological zone and seed type, a simulation of 500 iterations for each random variable was needed to have an adequate sample to capture the differences (inherent risk) in the yields. Instead of only 814 observed yields, the model simulated 10,500 for easier comparison. The third step was to validate the simulated distribution against the historical distribution. The validation results are presented in

Table 4. Observed yields (t/ha) per seed variety used (local, recycled, improved).

Yield Distribution (t/ha)	Seed Variety
	Local	Improved Recycled	Improved
Mean	0.996	1.388	1.813
STD	0.884	1.536	1.665
CV	88.99	109.22	91.93
Min	0.010	0.062	0.092
Max	4.74	5.67	6.75
Simulated Data	441	71	302

Source: Observed yields from the Tanzanian National Panel Survey Wave 5 (NPS-5).

and

Table 5. Simulated yields (t/ha) per seed variety used (local, recycled, improved).

Yield Distribution (t/ha)	Seed Variety
	Local	Improved Recycled	Improved
Mean	0.997	1.385	1.815
STD	0.887	1.516	1.668
CV	88.99	109.22	91.93
Min	0.010	0.062	0.092
Max	4.76	5.68	6.74
Simulated Data	500	500	500

Source: Simulated from MaizeSim.

, respectively.

2.3.4. Assumptions Underpinning the MaizeSim Model

The MaizeSim model employed in this study relies on a non-parametric Monte Carlo simulation framework designed to estimate the probability distribution of maize yields under various input scenarios. As with any simulation-based model, several key assumptions guide its application and interpretation. First, the model assumes that the historical yield data, drawn from the NPS Wave 5, accurately reflect the actual yield distributions across different agroecological zones and farming practices. These data are assumed to capture the inherent variability in farming systems, including differences due to management, environment, and seed and fertilizer use. Consequently, the simulation is grounded in the belief that these historical patterns are representative of current and near-future conditions, thus enabling a business-as-usual projection of yield outcomes [35].

Second, the model presumes the stationarity of environmental and agronomic conditions. This implies that the relationships observed in the historical data, such as the yield effect of using improved seeds, will remain consistent in the future unless otherwise modeled. It does not explicitly incorporate potential shifts in climate, policy, or technology adoption that may alter yield trajectories. Third, the simulation assumes that the baseline scenario (in this case, farms that use traditional seeds) or alternative scenarios (improved seed users) apply seed type uniformly and systematically across farms within each farming setup category. These practices are treated as fixed conditions during simulation, allowing for a structured comparison of probabilistic yield outcomes.

Fourth, MaizeSim adopts a non-parametric approach, assuming that yield outcomes can be modeled using empirical distributions rather than fitting to parametric forms like the normal or lognormal distributions. This method allows the model to preserve the actual shape and spread of observed data, avoiding distortions that may arise from incorrect distributional assumptions. Fifth, the simulation assumes that a sample size of 500 iterations, generated using Latin Hypercube Sampling, is sufficient to replicate the parent distributions of yield outcomes [33,35,36,38]. This approach provides a robust estimate of target probabilities, such as the likelihood of yields falling below 1.5 t/ha or exceeding 3.5 t/ha, for each farming system within each agroecological zone.

Moreover, the model does not require the explicit inclusion of biophysical variables such as rainfall, temperature, or soil type. Instead, it assumes that these factors are implicitly embedded in the historical yield data. This differentiates the model from biophysical crop models (like DSSAT or APSIM), which depend heavily on site-specific environmental and agronomic parameters. Lastly, the model treats each agroecological zone as an independent analytical unit. It assumes no spatial or policy-related interaction effects across zones, which simplifies the modeling structure but may limit the understanding of inter-regional dynamics and spillovers.

2.3.5. Limitation of the MaizeSim Model

While the MaizeSim simulation model offers a valuable framework for understanding maize yield variability under different farming setups in Tanzania, several limitations should be acknowledged. First, MaizeSim follows a non-parametric Monte Carlo simulation approach, which relies on historical yield data rather than agronomic or biophysical inputs such as rainfall, soil type, or temperature [35,39]. As a result, unlike mechanistic crop models like DSSAT or APSIM, MaizeSim does not simulate physiological growth processes under specific environmental conditions but instead provides probabilistic yield outcomes under “business-as-usual” farming practices. This distinction limits the model’s use for forecasting yields under future climate or input change scenarios.

Second, although the simulation outcomes were visually validated using statistical tests and probability density function (PDF) comparisons between observed and simulated data, parameter calibration using multi-year field trials was not conducted. This may reduce the granularity of localized adaptation insights, particularly in regions with highly variable agroecological and socio-economic conditions. Future research could enhance model robustness by incorporating longitudinal yield observations from representative zones (e.g., Southern Highlands, Central Zone) and integrating adaptive practices such as intercropping and conservation agriculture. Also, the model does not explicitly capture heterogeneity in labor inputs, household adoption behavior, or the cost structure of inputs, which could influence the economic feasibility and willingness of smallholders to adopt improved seeds and fertilizers. These factors are critical to understanding barriers to technology adoption and could be explored through integrated economic-simulation frameworks or behavioral models in future studies.

Despite these limitations, the model provides credible evidence on the yield probabilities associated with different input-use strategies and offers a practical decision-support tool for policymakers under current agricultural conditions.

2.3.6. Target Probabilities for Ranking Improved Seed Varieties Against Local

The stoplight chart criteria were used to compare the target probabilities for the three maize varieties. For comparison, we specify the stoplight’s two probability targets (lower target and upper target). The stoplight function calculates the probabilities of (a) exceeding the upper target (green), (b) being less than the lower target (red), and (c) observing values between the targets (yellow ). Since the average yield per ha for maize in Tanzania is nearly 1 t/ha, we set our lower target to be 1 t/ha and the potential yield to be 2 t/ha (twice the lower threshold). Doubling the upper limit was in line with the desire for the government to double the yield by 2030.

3. Results and Discussion

3.1. Model Validation

We conducted a rigorous model validation process to ensure the accuracy and reliability of our simulated random variables (specifically yield). This involved comparing the statistical properties of our simulated data to the observed or parent data. Two validation techniques were used. Two validation methods were used: Statistical Tests (Student-t Test and Chi-Squared Test) and the Graphical Tool (probability density functions (PDFs)).

3.1.1. Student-t Test and Chi-Squared Statistical Test

Student’s t-test and Chi-Squared statistical test were used to compare the simulated data’s means and standard deviations with the observed (original means). The null and alternative hypotheses were:

$${\mathrm{H}}_{\mathrm{o}}:{\overline{\mathrm{X}}}_{\mathrm{s}}={\overline{\mathrm{X}}}_{\mathrm{h}}$$

$${\mathrm{H}}_{\mathrm{A}}:{\overline{\mathrm{X}}}_{\mathrm{s}}\ne {\overline{\mathrm{X}}}_{\mathrm{h}}$$

Since the calculated t-statistics for all three scenarios do not exceed the critical values, we fail to reject the null hypothesis (H₀) that the means are equal. By comparing measures like mean, median, standard deviation, and skewness, we verified that the simulated data accurately captured the central tendency and dispersion of the observed yield values. The results of these tests, summarized in

Table 6. Student-t test and Chi-Squared Statistical test for simulated distribution vs. observed local seeds yield.

Test of Hypothesis (H0) for Parameters for All Farms Using Traditional Maize Seed (All_Regions_0)
Confidence level		95.00%
	Given Value	Test Value	Critical Value	p-Value
t-Test	1.00	0.03	2.25	0.98	Fail to Reject the H0 that the Mean is Equal to 0.997
Chi-Square Test	0.88	503.24	LB: 439.00	0.88	Fail to Reject the H0 that the Standard Deviation is Equal to 0.887
			UB: 562.79

,

Table 7. Student-t test and Chi-Squared Statistical test for simulated distribution vs. observed recycled seeds yield.

Test of Hypothesis (H0) for Parameters for All Farms Using Recycled Maize Seed (All_Regions_1)
Confidence level		95.00%
	Given Value	Test Value	Critical Value	p-Value
t-Test	1.39	0.00	2.25	1.00	Fail to Reject the H0 that Mean is Equal to 1.388
Chi-Square Test	1.52	492.59	LB: 439.00	0.86	Fail to Reject the H0 that the Standard Deviation is Equal to 1.516
			UB: 562.79

and

Table 8. Student-t test and Chi-Squared Statistical test for simulated distribution vs. observed distribution of improved seeds.

Test of Hypothesis (H0) for Parameters for All Farms Using Improved Maize Seed Varieties (All_Regions_2)
Confidence level		95.00%
	Given Value	Test Value	Critical Value	p-Value
t-Test	1.81	0.02	2.25	0.98	Fail to Reject the H0 that Mean is Equal to 1.815
Chi-Square Test	1.662	502.52	LB: 439.00	0.89	Fail to Reject the H0 that the Standard Deviation is Equal to 1.668
			UB: 562.79

, provided further evidence of the model’s ability to replicate the key characteristics of the observed data.

3.1.2. Graphical Tool for Validation

The graphical tool used the PDFs to investigate the yield distributions of observed and simulated data. This technique was used to assess the similarity between the observed and simulated data, as depicted in Figure 3, Figure 4 and Figure 5. The PDFs revealed a striking resemblance between the observed distribution and the simulated data, suggesting a strong correlation between the two. This visual comparison provides compelling evidence that the simulation model effectively captures the underlying statistical properties of the real-world phenomenon being studied.

3.2. Yield Probabilities for Farms Using Local Maize Seeds

The summary statistics in

Table 9. Summary statistics of yield for farms using local maize seeds.

Statistics	All Regions_0	NHZ_0	ECZ_0	CZ_0	SHZ_0	LZ_0	PL_0
Mean	0.97	1.44	0.74	0.71	1.25	1.05	0.97
STD	0.79	1.23	0.67	0.57	0.92	0.78	0.68
CV	80.9	85.5	90.1	80.0	73.2	75.0	70.6
Min	0.01	0.25	0.01	0.10	0.06	0.10	0.02
Max	4.74	4.74	2.97	2.37	4.45	4.74	3.76

provide valuable insights into maize yield variations across different regions in Tanzania when using local maize seeds. The mean yield across all regions is 0.97 tons per hectare, indicating generally low productivity with significant regional variation. For instance, the Northern Highlands Zone (NHZ) has the highest mean yield of 1.44 tons/ha, while the Central Zone (CZ) exhibits the lowest at 0.71 tons/ha. The Southern Highlands Zone (SHZ) also shows a relatively high mean yield of 1.25 tons/ha, reflecting its favorable conditions for maize production. The Eastern Coastal Zone (ECZ) and Lake Zone (LZ) are moderate, with means of 0.74 and 1.05 tons/ha, respectively. These statistics underscore the importance of region-specific interventions to address the variability in maize productivity, particularly in low-yielding regions like CZ and ECZ.

The standard deviation (STD) and coefficient of variation (CV) further highlight the significant yield variability within regions. The highest variability is observed in NHZ and ECZ, with CVs of 85.5% and 90.1%, respectively, while SHZ has the lowest variability at 73.2%. This suggests that NHZ has the highest mean yield, but that it also experiences considerable fluctuations, likely due to environmental factors and inconsistent input use. The minimum yields in all regions are near zero, indicating that some farms experience near-total crop failure, which could be attributed to poor weather conditions, pests, or limited access to inputs. The maximum yields range from 2.37 tons/ha in CZ to 4.74 tons/ha in NHZ and LZ, further illustrating the wide disparities in productivity. Addressing these variations through targeted policies and improved agricultural practices will be key to enhancing overall maize productivity in Tanzania.

Figure 6 presents the stoplight chart illustrating the yield probabilities for farms that use local maize seeds in Tanzania. The results show that a significant proportion of farmers (65%) who use local seeds have a high probability of harvesting less than 1 t/ha, which is well below the national average. Only 11% of farmers are projected to achieve yields above 2 t/ha, highlighting the limited potential of local seed varieties in boosting productivity.

3.3. Yield Probabilities for Farms Using Improved but Recycled Maize Seeds

Table 10. Summary statistics of yield for farms using recycled maize seeds.

Statistics	All Regions_1	NHZ_1	ECZ_1	CZ_1	SHZ_1	LZ_1	PL_1
Mean	1.22	1.18	1.06	1.84	1.33	1.08	1.44
STD	1.02	0.97	1.11	1.69	0.85	0.67	1.37
CV	83.6	81.9	105.6	92.1	64.2	61.6	95.1
Min	0.06	0.06	0.08	0.29	0.16	0.07	0.15
Max	5.67	2.97	5.67	4.94	3.56	2.4	5.05

presents the summary statistics for maize yield using recycled seeds across different regions in Tanzania, offering insights into how recycled seeds impact productivity. The overall mean yield across all regions using recycled seeds is 1.22 tons per hectare, which is an improvement over the 0.97 tons/ha observed with local seeds (as shown in Table 4). The Central Zone shows the highest mean yield at 1.84 tons/ha, significantly surpassing all other regions, while the Eastern Coastal Zone has the lowest mean yield at 1.06 tons/ha. This suggests that recycled seeds are more effective in regions like the CZ, potentially due to better management practices or environmental conditions that favor seed reuse. On the other hand, the ECZ’s poor performance with recycled seeds highlights the need for interventions to improve seed quality and farming techniques in this region.

The CV reflects the variability in yields, with ECZ having the highest variation at 105.6%, indicating considerable yield inconsistencies across farms using recycled seeds. In contrast, the Southern Highlands Zone and Lake Zone exhibit relatively lower CVs of 64.2% and 61.6%, respectively, suggesting more stable yields in these regions. The standard deviation (STD) shows similar patterns, with higher yield fluctuations in CZ (1.69) and ECZ (1.11). The maximum yields achieved in ECZ and CZ are quite high, at 5.67 tons/ha and 4.94 tons/ha, respectively, demonstrating that some farms are achieving excellent productivity levels. However, the wide gap between minimum and maximum values across regions also highlights farming practices and input disparities. These variations underscore the need for tailored regional strategies to improve the consistency of maize yields when using recycled seeds.

Figure 7 demonstrates the yield probabilities for farms using improved but recycled maize seeds. The data reveal that the probability of harvesting below the 1 t/ha threshold decreases to 53% for these farmers, compared to 65% for those using local seeds. Additionally, 18% of farmers using improved but recycled seeds have a chance of achieving yields above 2 t/ha. This slight improvement suggests that even though the seeds have been recycled, they still provide better results than local varieties. However, the fact that more than half of the farmers remain below the 1 t/ha mark highlights that recycling seeds reduces the effectiveness of improved varieties.

3.4. Yield Probabilities for Farms Using Purely Improved Maize Seeds

Table 11. Summary statistics of yield for farms using improved maize seeds.

Statistics	All Regions_2	NHZ_2	ECZ_2	CZ_2	SHZ_2	LZ_2	PL_2
Mean	1.71	1.95	1.34	1.79	2.36	1.32	1.65
STD	1.28	1.39	1.33	1.11	1.70	0.96	1.21
CV	74.9	71.3	99.7	61.9	72.0	72.6	73.4
Min	0.09	0.15	0.09	0.74	0.12	0.20	0.25
Max	6.74	5.19	5.93	5.34	6.74	3.95	5.93

provides summary statistics for farms using purely improved maize seeds. The statistics provided offer insights into farm yield results using improved maize seeds across various regions. The mean yields indicate a range from 1.34 in the Eastern Central Zone to a high of 2.36 in the Southern Highlands Zone (SHZ_2), illustrating significant regional variations in agricultural productivity with these seeds. The Southern Highlands Zone’s higher yield suggests optimal farming conditions or practices that favor improved seed varieties. Conversely, the Eastern Central Zone reports the lowest mean yield, which might indicate less favorable conditions or challenges in leveraging improved seed benefits. The average yield across all regions stands at 1.71, with a standard deviation of 1.28, indicating moderate overall variability in yield outcomes.

The variability in yields is further detailed by the CV percentages, which are notably high across the regions, especially in the ECZ at 99.7%, indicating a wide spread of data around the mean, thus higher risk and unpredictability in yield. The minimum and maximum yield values highlight the disparities within regions; for example, SHZ has a range from 0.12 to 6.74, showing a vast difference between the lowest and highest yields, possibly due to micro-climatic and soil differences or varied management practices. Such statistics are crucial for understanding regional performance and could guide targeted interventions to enhance maize production efficiency using improved seeds in underperforming areas.

Figure 8 illustrates the yield probabilities for farms using purely improved maize seeds. The results show a substantial improvement in maize productivity, with only 38% of farmers likely to harvest below 1 t/ha and 33% having the potential to exceed 2 t/ha. This indicates that farmers who adopt improved maize seeds are three times more likely to achieve yields above the national average compared to those using local seeds. Furthermore, improved seeds consistently demonstrate higher productivity across all agroecological zones, with the Northern Highlands Zone, Southern Highlands Zone, and Central Zone showing the most favorable outcomes.

4. Discussion

The results for a farming system using local maize seeds are consistent across the agroecological zones, with the Eastern Coastal and Central Zones exhibiting the highest likelihood of poor yields. In these zones, the probability of farmers harvesting less than 1 t/ha exceeds 70%, indicating the severe challenges that smallholder farmers who rely on traditional seed varieties face [40].These results reflect broader trends observed in other studies, such as those by [25], who found that local maize varieties are more vulnerable to environmental stressors like drought and poor soil fertility. The reliance on local seeds, particularly in regions prone to climatic variability, perpetuates a cycle of low productivity and food insecurity [25,40]. This suggests an urgent need for intervention by introducing improved seed varieties and modern agricultural practices to enhance productivity. Without addressing these issues, farmers in Tanzania’s most vulnerable zones will continue to struggle with insufficient yields to meet subsistence and market needs.

In the case of farming systems using recycled maize seeds, the results are particularly telling for regions such as the Southern Highlands Zone and Central Zone, where using recycled seeds has led to moderate yield improvements. These findings align with research by Kihara et al. [26], highlighting that improved seeds’ full potential is often diminished when seeds are reused for multiple planting seasons. Seed recycling contributes to genetic degradation, thereby reducing the seeds’ resilience and yield potential [41]. Therefore, although farmers can experience some gains with recycled seeds, the practice of consistently reusing seeds limits the long-term benefits of improved maize varieties.

For farms using improved seeds demonstrated a positive impact which is corroborated by studies such as those by Gebre et al. [25], which highlight the role of improved seed varieties in enhancing resilience to environmental stressors and increasing crop yields. These results support the argument that improved seeds are a critical tool for addressing the productivity gap in Tanzania’s maize sector [42,43]. However, the success of improved seeds is also influenced by factors such as access to proper agronomic practices and inputs like fertilizers, as emphasized by Nguyen et al. [31]. Thus, while improved seeds offer substantial benefits, their full potential can only be realized through complementary interventions that address the broader agricultural ecosystem.

The results from this farming setup clearly indicate that using purely improved maize seeds substantially enhances productivity and presents a promising pathway for transforming maize production in Tanzania. As shown in Figure 7, farms using improved seeds consistently report higher probabilities of achieving yields greater than 2 t/ha, particularly in the Northern Highlands Zone (42%), Southern Highlands Zone (42%), and Central Zone (34%). These zones also report the lowest probabilities of sub-optimal yields (i.e., <1 t/ha), with SHZ_2 standing out at just 22%, compared to 54% in the Eastern Central and Lake Zones. When compared to local seed systems, where low yields dominate (often exceeding 70% probability), these findings underscore the transformative productivity potential of improved seed adoption [21,22].

The implications for food availability are significant. As maize is Tanzania’s staple crop, increasing the share of farms surpassing the national average yields directly contributes to household food availability and surplus production for markets [44]. With only 38% of farms under improved seeds falling below 1 t/ha compared to over 70% under traditional systems, the data affirm the potential of improved seeds in stabilizing and raising national production levels. This shift could reduce dependence on imports, enhance resilience to food crises, and support agribusiness growth across the maize value chain [45].

From a policy perspective, these findings provide critical evidence to inform the implementation of Tanzania’s Seed Sector Development Strategy (SSDS) 2030. The SSDS aims to increase smallholder access to certified improved seed varieties and promote sustainable seed systems [32]. Our results support this policy direction by demonstrating that improved seed use can significantly close yield gaps, especially when regionally tailored. Regions like the Eastern Central Zone (ECZ_2) and Lake Zone (LZ_2), which lag in performance despite improved seed use, highlight the need for complementary investments in agronomic support, input supply systems, and site-specific extension services. These findings echo the importance of bundling seed interventions with fertilizer access, training, and climate-smart practices to maximize impact [46].

Moreover, the MaizeSim simulation approach used here provides a practical decision-support tool for agricultural planners to test the probabilistic impact of seed systems under uncertainty. These insights can be instrumental for the Ministry of Agriculture and seed system stakeholders in designing adaptive seed distribution strategies, optimizing fertilizer-seed bundles, and monitoring SSDS 2030 implementation progress at the agroecological level [47,48]. This study reinforces the argument that improved maize seeds, when integrated into context-specific, input-enabled farming systems, are key to increasing productivity and pivotal in achieving broader goals of food availability, climate resilience, and agricultural transformation in Tanzania. Nevertheless, it should be noted that yield gains must be complemented by policies that promote dietary diversification, access to nutrient-rich foods, and agricultural investment in biofortified or diversified cropping systems (e.g., legumes, vegetables, and fruits) to ensure nutritional adequacy, as argued by Jones et al. [49] and Pinstrup-Andersen [50]. This can be achieved by conducting studies that integrate household-level dietary diversity indicators, nutrient adequacy data, and income or consumption metrics from household surveys such as the Tanzania Household Budget Survey or LSMS-ISA datasets.

5. Conclusions

This study examined the potential for maize farms across Tanzania’s agroecological zones to exceed the national yield benchmark of 3.5 t/ha, under different combinations of seeds (local, improved, or recycled). Using a Monte Carlo or stochastic simulation approach (MaizeSim), this study analyzed farming setups across seven agroecological zones to estimate the probabilities of high, moderate, and low productivity. The MaizeSim provided a robust analysis, accounting for the inherent variability in yield outcomes, and underscored the potential of improved seeds to enhance food availability and income generation for Tanzanian smallholder farmers. Maize was considered in this study because it is the staple food for most Tanzanians and a strategic/priority crop. This study found that farmers who adopted improved seed varieties had a much higher probability of achieving yields above the national average compared to those using local or recycled seeds. However, the results also revealed substantial regional disparities in the benefits of improved seed adoption. While zones such as the Southern Highlands, Northern Highlands, and Central Zones experienced significant yield increases, areas like the Eastern Coastal and Southern Zones saw more modest gains. These regional variations highlight the importance of tailoring agricultural interventions to specific environmental and agroecological conditions to maximize their effectiveness. Without addressing these regional challenges, the broader potential of improved seeds in boosting productivity across the country will remain unrealized.

Ultimately, this study emphasizes the critical role of improved maize seeds in closing the productivity gap and enhancing the resilience of Tanzania’s maize sector. However, adoption alone is not sufficient. Complementary support, including improved access to agricultural inputs, extension services, soil fertility management, crop rotation, integrated pest management, access to stable input markets and infrastructure, is vital to ensuring that smallholder farmers can fully realize the benefits of these improved varieties. These findings are especially relevant to national strategies aimed at improving maize productivity, food availability, and farmer resilience. The evidence supports ongoing efforts under Tanzania’s SSDS 2030 to promote improved seed adoption, and it also underscores the complementary role of integrated soil fertility management. By identifying region-specific success probabilities, this study equips policymakers with the information needed to design targeted and inclusive agricultural interventions that match local conditions.

Promoting the adoption of improved seeds and sustainable input use is not only critical for increasing yields, but also for building the climate resilience of smallholder maize farmers. Given that maize is a staple food crop with substantial socio-economic importance, scaling the best-performing farming setups can contribute significantly to rural poverty reduction, food self-sufficiency, and national economic development. Additionally, this study demonstrates how a simulation-based assessment can serve as an evidence-based decision-support tool for agricultural planning in Tanzania and beyond. Future research should build on this approach by integrating market access and other crucial inputs like fertilizers, gender dynamics, and climate projections to refine recommendations for long-term agricultural sustainability.

While this study focused on yield performance as a foundational indicator of profitability, it did not include a full-cycle economic analysis incorporating the costs of inputs such as seed premiums, increased fertilizer use, labor opportunity costs, or financing constraints. Given that improved seeds often cost 2–3 times more than local seeds and require higher input levels, future research should integrate these factors into a net present value (NPV) or stochastic cost-benefit framework. Additionally, considering that a significant proportion of Tanzanian farmers (approximately 52%) rely on informal credit systems, future models should explore financial feasibility and adoption dynamics under realistic budgetary and credit access scenarios.

Likewise, although this study uses historical yield data to simulate yield variability under existing agroecological conditions, it does not explicitly incorporate future climate change projections or biophysical stressors such as drought and heat waves. As a result, the model does not capture how seed performance might attenuate under projected climate scenarios, including increased rainfall variability or temperature extremes. Future research should also find a way to integrate CMIP6-based climate models and Shared Socio-economic Pathways (e.g., SSP2-4.5, SSP5-8.5) to evaluate the resilience of improved seed varieties under varying climatic conditions. Such integration would enable the quantification of income risks due to climate shocks and help assess the long-term value of stress-tolerant seed technologies in Tanzania’s maize sector. Further research is also needed to explore how yield gains in maize relate to broader dietary diversity and nutritional outcomes.