1 Maize seed demand, supply and varietal age dynamics in Ethiopia Jeleta Kebede1*, Abdu Mohammed2, Moti Jaleta2 1 Department of Economics, York University, Toronto, Canada. 2International Maize and Wheat Improvement Center, CIMMYT, Addis Ababa, Ethiopia Abstract Despite a continuous maize varietal development in Ethiopia, farmers are using relatively old varieties. Against this backdrop, using the quantity of maize seed supply and distribution for the period of 2009-2023, this study analyses dynamics in maize varietal age in the country. Result shows that while average varietal age (AVA) of maize seed demanded, supplied, and distributed in the country gradually decreased overtime for both public and privately sourced hybrids, it was increasing for improved open-pollinated varieties (OPVs). We find that varietal age below threshold increases demand, supply and distribution of maize seed varieties, but varietal age above threshold reduces demand, supply and distribution of the varieties. We observe similar results for the effects of seed specific varietal age on demand, supply and distribution of the varieties. We further demonstrate that while there is a bidirectional causality between maize seed supply and distribution, the relationships between maize seed demand and supply, and demand and distribution are unidirectional: previous year supply and distribution cause demand for specific variety, but supply and distribution are not responding to previous demand. This implies that farmers’ varietal knowledge is limited to the varieties supplied to them in the previous season, and supply is driving demand. The results further imply that maize seed production and marketing focused on more recently released varieties in general and that of OPV in particular can be used as an instrument to reduce AVA thereby increasing varietal turnover. Keywords: Average varietal age, maize, seed, Ethiopia. ______________________ *Corresponding author: email: jaalii2007@yahoo.com mailto:jaalii2007@yahoo.com 2 1. Introduction Increasing production and productivity of smallholder farmers is crucial to improve their income, food security and livelihood (Lunduka et al., 2019; Donovan et al., 2021; Nuthalapati et al., 2024). The role of improved crop variety adoption by smallholders is vital in this regard (Jaleta et al., 2018; Gatto et al., 2021; Rutsaert et al., 2021; Habte et al., 2023; Kalsa et al., 2024). To use improved varieties to address adverse effects of climate change thereby ensuring food security, economic stability, and poverty reduction by increasing production and productivity, speeding up varietal turnover and reducing varietal age are essential (Atlin et al., 2017; Lunduka et al., 2019; Rutsaert & Donovan, 2020; Rutsaert et al., 2021; Chivasa et al., 2022). Chivasa et al. (2022) argue that the cropping systems require not only improved varieties with tolerance to multiple stresses but also active varietal turnover. The authors posit that the genetic gains in grain yield can only translate into productivity if improved varieties are rapidly disseminated and old ones are replaced. De Groote and Omondi (2023) contend that while many new varieties of maize have been released and disseminated in East and Southern African countries, yield increases have been limited partly due to the slow adoption of the latest varieties or low varietal turnover. In such a context, Atlin et al. (2017) call for the need for rapid varietal turnover through active dissemination of new varieties, and withdrawal of obsolete ones. In Ethiopia, despite the importance of maize in the country’s crop production and food security and hence increased related investment in the last two decades, the growing importance of maize has been based on just a few market-dominant varieties (Bukero et al., 2019; Marenya et al., 2022). Although extensive literature has been undertaken related to the AVA of maize seed varieties (Rutsaert & Donovan, 2020; Chen et al., 2024), most of the studies deal with analyzing survey data collected at farmers level and as such provide insufficient insight at aggregate level. Moreover, the existing studies (Fisher et al., 2015; Lunduka et al., 2019; Rutsaert et al., 2021) by and large undertake varietal age analysis using area weighted approach. Against this backdrop, using data from 2009 to 2023 for 26 seed varieties, we examine the maize seed dynamics in Ethiopia. This study incrementally contributes to the literature related to maize seed system in several ways. First, we analyze the AVA of maize seed demand, supply and distribution using quantity weighted approach. Second, employing a panel dynamic model, we present a nonlinear effect of AVA and seed specific varietal age on supply and distribution of maize seed in the country. Third, we demonstrate that the results of our analyses vary between: (i) public and private seed sources, and (ii) hybrid and OPV sources. Last, to the best of our knowledge, this study is the first to present causality analysis of maize seed supply and distribution system in Ethiopia. Our results are robust to several sensitivity analyses. The remaining part of the paper is structured as follows. Section 2 presents data and methodology employed for the analysis. Section 3 presents the results and discussion. Section 4 concludes the discussion and draws policy implications. 3 2. Methodology 2.1. Data and sample Demand, supply, and distribution data of seed varieties was obtained from the Agricultural Inputs Department at the Ministry of Agriculture in Ethiopia (MoA), 2024. Data on cereal yield, fertilizer consumption, inflation and GDP per capita growth was obtained from World Development Indicators (WDI), World Bank, 2024 while data on agricultural research, extension technology transfer, and input subsidies came from MAFAP, FAO, 2024. The sample size— in terms of the maize seed varieties and years employed —was determined based on data availability. Table 1 presents the sample varieties of maize seeds used in the study, their source (public and private), types (hybrid and OPV), and year of release. Table 2 presents the variables used in the study, their definitions and data sources. Table 1. Maize varieties under consideration No. Variety Sources: Public or private Types: Hybrid or OPV Year of release 1 BH661 Public Hybrid 2011 2 BH660 Public Hybrid 1993 3 BH545Q Public Hybrid 2008 4 BH543 Public Hybrid 2005 5 BH542 Public Hybrid 2002 6 BH540 Public Hybrid 1995 7 BH140 Public Hybrid 1988 8 BH547 Public Hybrid 2013 9 BH546 Public Hybrid 2013 10 Melkassa 1Q Public OPV 2013 11 Melkassa 2 Public OPV 2004 12 Melkassa 4 Public Hybrid 2006 13 Melkassa 6Q Public OPV 2008 14 MH140 Public Hybrid 2013 15 MH138Q Public Hybrid 2012 16 Jabi (PHB3253) Private Hybrid 1995 17 Shone (30G19) Private Hybrid 2006 18 Shala (P2859W) Private Hybrid 2011 19 Limu (P3812W) Private Hybrid 2012 20 Welel (PHB30V53) Private Hybrid 2006 21 Tabor (PHB30H83) Private Hybrid 2001 22 Jibat (AMH 851) Public Hybrid 2009 23 Hawassa 1 Private OPV 2012 24 Guto Private Hybrid 2015 25 Gibe-1 Public OPV 2001 26 Gibe-2 Public OPV 2011 4 2.2. Econometric models 2.2.1. Dependent variables The dependent variables used in regression analyses are demand, supply and distribution of maize varieties. Supply measures the amount of maize seed production during the year while distribution represents the amount sold during the year under consideration. 2.2.2. Independent variables Varietal age and AVA are the main independent variables. Varietal age can impact the dependent variables through several channels. For instance, information dissemination about a new variety takes some time before farmers make the decision to adopt it. As the information about the seed gradually spreads, the adoption of the variety may increase over time. However, over time, as other relatively new varieties are released, demand for the existing variety may decrease and hence its supply and distribution likely fall over time. This implies that the effect of varietal age on the supply and distribution of varieties appears to be nonlinear. In a similar, we expect that the nonlinear impact on demand, supply and distribution holds for AVA. Table 2: Summary of variables’ definitions and sources Variable Definition Source Demand Maize seed demanded by variety (metric ton). MoA, 2024 Supply Maize seed supplied by variety (metric ton). MoA, 2024 Distribution Maize seed marketed by variety (metric ton). MoA, 2024 GDPPC growth GDP per capita growth (annual %) WDI, 2024 Inflation Inflation, consumer prices (annual %) WDI, 2024 AVA of demand Quantity of demand weighted maize AVA. Own computation AVA of supply Quantity of supply weighted maize AVA. Own computation AVA of distribution Quantity of distribution weighted maize AVA. Own computation Age Seed specific varietal age. MoA, 2024 Extension transfer Public expenditures financing provision of extension services MAFAP, FAO, 2024 Cereal yield Cereal yield (kg per hectare) WDI, 2024 Input subsidies Monetary transfers to agricultural producers that are based on on-farm use of inputs: variable inputs, capital, and on-farm services MAFAP, FAO, 2024 Fertilizer consumption Fertilizer consumption (kilograms per hectare of arable land) WDI, 2024 5 2.2.3. Control variables We use two groups of control variables: (i)seed level variables such as lags of demand, supply and distribution of the varieties, and (ii) country level variables. We include GDPPC growth because its increase enhances purchasing power of consumers triggering higher supply and distribution to satisfy the increased demand. We include inflation as a proxy for macroeconomic stability and argue that increased macroeconomic stability promotes demand, production and distribution of maize seeds. We account for public expenditure on extension technology transfer, and input subsidies given that they impact adoption of new varieties thereby affecting demand for, supply and distribution of the varieties. We further account for cereal yield and fertilizer consumption as potential determinants of demand, supply and distribution of seed varieties. 2.3. The empirical model To analyze the dynamics of maize seed varieties focusing on the effects of seed specific varietal age and AVA on demand, supply and distribution of varieties, we develop the empirical model as follows: Seedikt = α0 + θAgeit + 𝑋it ′  + Zt ′ + αi + εikt where i = 1, … , N; k = 1, . , 3;t = 1, … , T (1) where Seedikt represents the seed demand, supply, or distribution of maize variety 𝑖 in year 𝑡. Ageit and X′it are respectively the varietal age of variety 𝑖 in reference to year 𝑡 and vectors of the seed level control variables that are associated with variety 𝑖 at time 𝑡. Zt ′ is a vector of country level control variables. εikt is error term related to demand, supply or distribution of variety 𝑖 at time 𝑡; αi is individual fixed effect. In the case of analyzing the effect of average varietal age AVA, which varies by demand, supply or distribution, the empirical model will be: Seedikt = α0 + θAVAkt + 𝑋it ′  + Zt ′ + αi + εikt (2) 3. Results and discussion 3.1. Descriptive analyses Figure 1 and Table A1 presents percentage share of each variety (in terms of mean value over the period of 2009 to 2023) in the total demand, supply, and distribution of maize seeds; the figure also presents (a) concentration ratio (the percentage of total seed demand, supply and distribution is covered by the top five varieties), (b) disaggregated share of each variety by seed source and type, (c) relative shares of public and private sourced seeds, and (d) relative share of hybrid and OPV seed types. Figure A1 in the Appendix presents mean values (from 2009 to 2023) of demand, supply and distribution by variety. Figure 1 clearly shows that a few maize varieties such as BH661, BH660, BH540, Limu (P3812W), and Shone (30G19) dominate the maize seed system of the country. The concentration ratio corroborates the result: the top five varieties cover more than 6 74% of demand, supply, and distribution of seeds. The result corroborates the studies that claim that the country’s maize seed system is dominated by a few seeds, despite developments of several new varieties (Worku et al., 2012; Marenya et al., 2022). Table A2 in Appendix presents summary statistics of demand for, supply and distribution of maize varieties from 2009 to 2023. 0 5 10 15 20 25 BH 66 1 BH 66 0 BH 54 5Q BH 54 3 BH 54 2 BH 54 0 BH 14 0 BH 54 7 BH 54 6 M el ka sa 1 Q M el ka sa 2 M el ka sa 4 M el ka sa 6 Q M H 14 0 M H 13 8Q Ja bi i (P H B 32 53 ) Sh on e (3 0G 19 ) Sh al a (P 28 59 W ) Li m u (P 38 12 W ) PH B 3 0 V 53 PH B 3 0 H 8 3 Jib at (A M H 8 51 ) H aw as a 1 G ut o G ib e 1 G ib er 2 Percentage share of the varities, mean value from 2009 to 2023 Demand Supply Distribution 0 5 10 15 20 25 30 35 B H 6 6 1 B H 6 6 0 B H 5 4 5 Q B H 5 4 3 B H 5 4 2 B H 5 4 0 B H 1 4 0 B H 5 4 7 B H 5 4 6 M el k as a 1 Q M el k as a 2 M el k as a 4 M el k as a 6 Q M H 1 4 0 M H 1 3 8 Q Ji b at ( A M H 8 5 1 ) G ib e 1 G ib er 2 Percentage share of the varities of public sources, mean value from 2009 to 2023 Demand Supply Distribution 0 10 20 30 40 50 Ja b ii (P H B 3 2 5 3 ) S h o n e (3 0 G 1 9 ) S h al a ( P 2 8 5 9 W ) L im u (P 3 8 1 2 W ) P H B 3 0 V 5 3 P H B 3 0 H 8 3 H aw as a 1 G u to Percentage share of the varities of private sources, mean value from 2009 to 2023 Demand Supply Distribution 7 Figure 1. Percentage share of the varieties in total demand, supply and distribution of maize seed system, from 2009 to 2023. Figure 2 presents the trends of demand, supply, and distribution of maize seed varieties for the overall maize varieties. The result shows an overall increasing trend in total demand, supply and distribution while a slight decline is observed of supply and distribution starting from 2021. Figure A2.1 and Figure A2.2 in the Appendix preset the disaggregated results by seed sources and seed types. 0 5 10 15 20 25 B H 6 6 1 B H 6 6 0 B H 5 4 5 Q B H 5 4 3 B H 5 4 2 B H 5 4 0 B H 1 4 0 B H 5 4 7 B H 5 4 6 M el k as a 4 M H 1 4 0 M H 1 3 8 Q Ja b ii (P H B 3 2 5 3 ) S h o n e (3 0 G 1 9 ) S h al a ( P 2 8 5 9 W ) L im u (P 3 8 1 2 W ) P H B 3 0 V 5 3 P H B 3 0 H 8 3 Ji b at ( A M H 8 5 1 ) G u to Percentage share of the varities of hybrid type, mean value from 2009 to 2023 Demand Supply Distribution 0 10 20 30 40 50 60 70 80 90 M el ka sa 1 Q M el ka sa 2 M el ka sa 6 Q H aw as a 1 G ib e 1 G ib er 2 Percentage share of the varities of OPV type, mean value from 2009 to 2023 Demand 0 100 200 300 400 500 600 700 20 09 20 10 20 11 20 12 20 13 20 14 20 15 20 16 20 17 20 18 20 19 20 20 20 21 20 22 20 23 Trends of total demand, supply and distribution of overall varieties Total demand Total supply Total distribution 8 Figure 2: Trends of overall demand, supply, and distribution of maize seed varieties We also present the trend in terms of percentage shares of the public and private sources of seeds in the overall maize seed system. The overall results presented in Figure A2.3 and Figure A2.4 show that the share of demand, supply and distribution of public sourced seeds have been decreasing while that of private sourced ones have shown an increasing trend until 2021. After 2021, while the share of demand for private sourced seeds has surpassed that of the public sourced ones, share of supply and distribution of private sources seeds show a declining tendency. The results related to dynamics in relative share of public vs private sourced seeds in the maize seed system of the country is in line with literature (Rutsaert et al., 2021). Dynamics of relative share by seed type as presented by Figure A2.5 presents shows persistent dominance of the hybrid seed types in the maize seed system of the country accounting for far more than 90% of market share. 3.2. Average varietal age analysis To mitigate the threats posed by climate change and related stresses, which are the major causes of reduced maize production and food insecurity (Atlin et al., 2017; Lunduka et al., 019; Rutsaert & Donovan, 2020), governments and donors have sought to expand the availability of new maize hybrids for smallholders in Eastern and Southern Africa where the crop plays an outsized role in human diets (Worku et al., 2020; Rutsaert et al., 2021). Accordingly, the lion’s share of public investments in maize breeding have focused on the development of hybrids with increased drought and disease tolerance in East Africa (Rusaert et al., 2024). In improving the productivity of maize- based cropping systems, Chivasa et al. (2022) states the importance of rapid-cycle, climate- adaptive breeding programs and seed systems that deliver new, elite varieties to farmers to replace obsolete ones. Varietal age is an indicator of adoption efficiency and diffusion process in farmers' field; accordingly, varietal age increment is associated with reducing returns to plant breeding (Singh et al., 2020). In this study, we compute quantity weighted AVA as follows. Quantity adjusted AVA of demand, supply or distribution at time 𝑡 is given as: AVAt =pitYSRit (3) Where 𝑝𝑖𝑡 and 𝑌𝑆𝑅𝑖𝑡 are market share, and varietal age of variety 𝑖 at time 𝑡, respectively. Accordingly,, we computed quantity weighted AVA for demand, supply and distribution of the varieties from 2009 to 2023. Figure A2.6 in the Appendix presents the results related to overall varieties. The results show that AVA of demand, supply and distribution of overall maize seed varieties exhibit decreasing trend. Although the overall trend shows a decreasing tendency, the decrement rate appears to be slow highlighting that older varieties are used by farmers for long time before they are replaced by new ones. 9 The result is in line with the literature that shows that varietal turnover is slow in many developing countries in general and in SSA in particular. According to Rutsaert and Donovan (2020), although landraces have been replaced by modern varieties for several crops in many developing countries, the slow speed of varietal turnover has become a topic of concern. Bukero (2023) notes that adoption of second- and third-generation cultivars offering improvements in yield, output quality, and stress resistance seems to be occurring at a much slower pace. The problem is particularly worse in SSA because, despite the consensus related to the importance and benefits of accelerated varietal turnover to climate change adaptation and food security, the rate of maize varietal replacement is slow (Chivasa et al., 2022). 3.2.1. Average varietal age analysis by public vs private sources The aggregate AVA analysis may conceal potential differences between public and private sourced seeds. To unpack the potential difference, we present the analysis by disaggregating the varieties into public and private sources. Figure 3 shows that AVA of public sourced seeds exhibits decreasing trend despite slow decrement pace, but that of private sourced one does not show a decreasing trend1. Figure 3. Average varietal age of demand, supply and distribution by sources of variety 3.2.2. Average varietal age analysis by hybrid vs OPV types Analyzing AVA by unbundling by seed types can shed light on important information that aggregation may hide. Hence, accounting for the cultivars classes of hybrids and OPVs may matter in AVA analysis. Hybrid is freshly purchased hybrid seed while OPV is a seed that has not been recycled for more than three seasons (Abate et al., 2015; Bukero, 2023). Figure 4 shows that while AVA of demand, supply and distribution of the hybrid varieties follow similar trends with that of 1 Distribution is sometimes higher than supply as carryover from the previous year gets into the market. 0 5 10 15 20 25 20 09 20 10 20 11 20 12 20 13 20 14 20 15 20 16 20 17 20 18 20 19 20 20 20 21 20 22 20 23 V ar ie ta l a ge (y ea rs ) Average varietal age of public sources Demand Supply Distribution 0 2 4 6 8 10 12 14 16 20 09 20 10 20 11 20 12 20 13 20 14 20 15 20 16 20 17 20 18 20 19 20 20 20 21 20 22 20 23 V ar ie ta l a ge (y ea rs ) Average varietal age of private sources Demand Supply Distribution 10 the overall varieties; on the contrary AVA of demand, supply, and distribution of the OPV type show an increasing trend. Figure 4. Average varietal age of demand, supply and distribution by variety types Although the AVA of hybrids is decreasing, the slow decrement pace indicates that farmers are not using recently released seed but are rather using relatively old varieties; the situation is worse in the case of OPV given that the AVA exhibits an increasing trend. The findings of lags in terms of adopting new varieties are in line with the literature (Rutsaert & Donovan, 2020). Overall, despite differences in our findings when analyzed by seed source and type, the results indicate that farmers in Ethiopia are not fully reaping the expected benefit from improved maize breeding because they use relatively older varieties. Studies show that Ethiopia is one of the countries where insufficient availability of seed of recently released varieties affects varietal replacement (Fisher et al., 2015; Chivasa et al., 2022). Singh et al. (2020) summarizes constraints related to varietal replacements. They argue that institutional, technical, environmental, and socio-economic factors are major constraints and forward strategies to improve varietal replacement rates through policy change. Accordingly, to accelerate varietal replacement rates, a paradigm shift in policy change is required at institutional level to address: (a) crop breeding for local adaptation including specific cropping pattern, (b) pre-breeding that mainly comprises of steps such as bio-prospecting traits or genes from exotic materials and their transfer into intermediate materials, (c) participatory plant breeding, (d) maintenance breeding, and (e) requirement of seed systems for rapid varietal replacement. 3.3. Dynamic analyses of supply and distribution of maize seed varieties 3.3.1. Baseline results 0 2 4 6 8 10 12 14 16 18 20 20 09 20 10 20 11 20 12 20 13 20 14 20 15 20 16 20 17 20 18 20 19 20 20 20 21 20 22 20 23 V ar ie ta l a ge (y ea r) Average varietal age of hybrid varieties Demand Supply Distribution 0 2 4 6 8 10 12 14 16 18 20 20 09 20 10 20 11 20 12 20 13 20 14 20 15 20 16 20 17 20 18 20 19 20 20 20 21 20 22 20 23 A xi s Ti tl e Average varietal age of OPV varieties Demand Supply Distribution 11 Due to the dynamic nature of the dependent variables (i.e., demand, supply, and distribution are likely impacted by their respective lags), endogeneity problem potentially arises. Therefore, we use system GMM estimation that was initially proposed by Arellano and Bover (1995) and later fully developed by Blundell and Bond (1998). Some studies recommend using the GMM approach, which addresses some endogeneities such as those arising from time invariant sources through removing the fixed effect by differencing and using the lagged values as an instrument (Arellano & Bond, 1991). GMM, nevertheless, suffers from weaknesses including likelihood of dropping valuable information, which could potentially lead to measurement error bias, and possibly suffers from weak instruments (Blundell & Bond, 1998). System GMM, however, by using additional moment conditions, addresses the problem of weak instruments associated with GMM. To estimate using system GMM, we specifically write Equation (1) in a system of equations as follows: ΔSeedikt = ΔθAgeit + Δ𝑋it ′  +Δ Zt ′ + Δεikt (3) Seedikt = α0 + θAgeit + 𝑋it ′  + Zt ′ + αi + εikt (4) Accordingly, system GMM estimates both the differenced equation (Equation 3) and the level equation (Equation 4) while GMM uses only the differenced equation. As such, system GMM addresses the problem of weak instruments and thereby yields more efficiency than GMM by using an additional moment condition. The consistency of system GMM estimation depends on the assumptions that the instruments used are valid, and there is no second-order serial correlation of the error term εikt. In modelling supply, we include lags of demand and distribution as regressors rather than using their contemporaneous values due to potential endogeneity problem arising from the likelihood of reverse causality because demand and distribution can impact supply. To address this potential problem, we use predetermined values of demand and distribution rather than using their contemporaneous value. Similarly, we use lags of demand and supply in modelling distribution, and lags of supply and distribution in modelling demand. Table 3 presents system GMM regression results. Columns 1 to 3 present the results related to the effects of varietal age on distribution, supply and demand of overall varieties. The columns 4 to 6 presents respective results related to the effects of AVA. The result shows that varietal age nonlinearly impacts supply and distribution of the overall maize seed varieties in Ethiopia: varietal age increases supply and distribution while its square reduces them. Supply and distribution of overall maize varieties initially increase as varietal age increases but later decrease after threshold of varietal age is reached. Figure 5 presents average marginal effects of varietal age on distribution, supply and demand of maize seeds. The figure shows the nonlinear effect of varietal age on distribution, supply and demand of overall maize varieties. Table 3. Effects of varietal age and AVA on demand, supply and distribution 12 (1) (2) (3) (4) (5) (6) Variables Distribution Supply Demand Distribution Supply Demand L. distribution 0.425*** 0.625*** 0.524*** 1.003*** 0.973*** 0.131** (0.018) (0.039) (0.047) (0.022) (0.021) (0.062) L. supply -0.158*** -0.324*** 0.190*** -0.382*** -0.623*** 0.320*** (0.010) (0.020) (0.018) (0.029) (0.015) (0.064) L. demand 0.431*** 0.530*** 0.565*** 0.238*** 0.547*** 0.729*** (0.009) (0.018) (0.019) (0.015) (0.013) (0.026) Age 1.109*** 0.839* 0.499*** (0.204) (0.442) (0.171) Age2 -0.069*** -0.040*** -0.0439*** (0.006) (0.010) (0.005) AVAk 125.5*** 59.78*** 31.47** (11.10) (10.61) (15.05) AVAk 2 -4.400*** -2.015*** -1.003** (0.390) (0.343) (0.506) Inflation -0.082*** -0.106** -0.095*** -0.213*** -0.064*** -0.627*** (0.013) (0.050) (0.025) (0.033) (0.025) (0.045) GDPPC growth 0.885*** 3.391*** 1.804*** 0.129 1.670*** 0.085 (0.131) (0.534) (0.222) (0.317) (0.296) (0.217) Extension transfer 7.444*** 10.59*** 10.80*** 8.381* 6.572*** 0.304 (0.686) (3.463) (2.278) (4.773) (1.292) (2.600) Input subsidies 4.720* 16.63 66.62*** 98.43*** 73.69*** 7.466 (2.769) (10.47) (9.686) (16.17) (5.715) (10.16) Cereal yield 0.095 0.930** 3.711*** 2.795*** 0.882*** 2.072*** (0.094) (0.430) (0.113) (0.332) (0.227) (0.149) Fertilizer consumption 20.48*** 22.99*** 94.64*** 85.33*** 41.03*** 115.2*** (2.198) (8.904) (2.280) (4.376) (12.55) (8.474) Constant -9.810*** 21.50 103.3*** -862.3*** -445.6*** -239.8** (3.575) (20.03) (5.443) (73.88) (87.50) (119.6) Observations 253 255 261 253 255 261 Number of varieties 26 26 26 26 26 26 AR1 (PV) 0.0121 0.007 0.061 0.001 0.003 0.076 AR2 (PV) 0.329 0.423 0.912 0.652 0.164 0.993 Hansen test (PV) 0.659 0.125 0.374 0.602 0.611 0.755 Standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1 13 To check the consistency of the system GMM results, we employed: (1) the over-identification test developed by Hansen (1982) to test for the validity of instruments, and (2) the Arellano and Bond (1991) to check the assumption of no second-order serial correlation is satisfied. In summary, the system GMM estimation results are valid given that we fail to reject both the null hypotheses of the over-identification test (instruments are valid) and there is no second-order serial correlation. In other words, post estimation tests of system GMM estimation as provided by the Arellano–Bond (AR2) and Hansen test results show validity of the model. 14 Figure 5. Average marginal effects of varietal age on distribution, supply and demand 3.3.2. Analyses by sources and types of varieties Given that the aggregate result presented in Section 3.3.1 may conceal potential variations of varietal age impacts by seed sources and types, in this section, we present a disaggregated analysis by public and private sources and hybrids and OPV types. Table 4 presents the results2. The results show that varietal age has qualitatively similar impacts on public sourced varieties as that of the aggregate results: varietal age below threshold increases supply and distribution while age above 2 The results associated with control variables are not reported due to word limits but are available on request. 15 threshold reduces them. The results related to the private sourced varieties show not significant impact of varietal age below threshold. Table 4. Effects of varietal age and average varietal age by sources and types of varieties 16 (1) (2) (3) (4) (5) (6) Variables Distribution Supply Demand Distribution Supply Demand Public vs private Public Age 7.312** 1.149** 1.994*** (3.134) (0.505) (0.552) Age2 -0.140** -0.043*** -0.072*** (0.064) (0.0130) (0.016) AVAk 75.47** 84.90*** 155.7** (33.01) (18.60) (68.67) AVAk2 -2.778** -2.788*** -5.009** (1.206) (0.596) (2.192) Control variables Yes Yes Yes Yes Yes Yes Observations 179 180 184 179 180 184 Number of varieties 18 18 18 18 18 18 AR1 (PV) 0.018 0.033 0.084 0.073 0.013 0.014 AR2 (PV) 0.274 0.333 0.255 0.221 0.420 0.371 Hansen test (PV) 0.784 0.985 0.981 0.752 0.903 0.693 Private Age 9.107 14.21 63.91 (6.073) (65.61) (42.09) Age2 -0.775* -0.896 -4.540* (0.462) (3.331) (2.587) AVAk 14.36** 17.30 148.0** (6.073) (65.61) (42.09) AVAk2 -0.775* -0.896 -4.540* (0.462) (3.331) (2.587) Control variables Yes Yes Yes Yes Yes Yes Observations 74 75 77 74 75 77 Number of varieties 8 8 8 8 8 8 AR1 (PV) 0.046 0.039 0.001 0.031 0.010 0.013 AR2 (PV) 0.300 0.922 0.564 0.253 0.714 0.329 Hansen test (PV) 0.153 0.122 0.775 0.140 0.764 0.892 Hybrid vs OPV Hybrid Age 2.371** 3.705 4.462*** (1.056) (2.394) (1.408) Age2 -0.114*** -0.184** -0.165*** (0.028) (0.076) (0.037) AVAk 42.64*** 50.13*** 32.42 (11.58) (11.97) (43.83) AVAk2 -1.543*** -1.825*** -0.963 (0.415) (0.432) (1.427) Control variables Yes Yes Yes Yes Yes Yes Observations 196 198 204 196 198 204 Number of varieties 20 20 20 20 20 20 AR1 (PV) 0.013 0.027 0.016 0.054 0.016 AR2 (PV) 0.314 0.513 0.110 0.188 0.985 Hansen test (PV) 0.995 0.909 0.982 0.205 0.987 OPV 17 Age 3.395 0.0946 0.032 (2.995) (0.103) (0.958) Age2 1.634* 0.059** 0.259** (0.849) (0.023) (0.129) AVAk 0.018 0.113 0.047 (1.041) (0.100) (0.051) AVAk2 0.006* 0.005** 0.164* (0.004) (0.002) (0.088) Control variables Yes Yes Yes Yes Yes Yes Observations 57 57 57 57 57 57 Number of varieties 6 6 6 6 6 6 AR1 (PV) 0.015 0.089 0.001 0.031 0.001 0.023 AR2 (PV) 0.402 0.514 0.143 0.430 0.432 0.843 Hansen test (PV) 0.395 0.336 0.582 0.403 0.364 0.591 Standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1 Analysis by seed type shows that while the results related to the impact of varietal age on hybrid seeds are by and large similar to that of the overall varieties, in contrast, varietal age persistently increases supply and distribution of OPV type. The results confirm the analysis presented in Section 3.2 that shows that the AVA of supply and distribution of OPV varieties persistently increases over time. 3.3.3. Market share analysis Besides investigating the effects of varietal age and AVA on the distribution, supply and demand of the varieties, we further examine their effects on the market share of the varieties in the distribution, supply and demand of the seed dynamics. To this end, we use the ratios obtained dividing distribution, supply and demand of each variety by the respective overall distribution, supply and demand of the varieties in the corresponding years. Table 5 presents the results and shows that varietal age initially increases market shares of distribution, supply and demand of the varieties in the maize seed dynamics; however, its effect on market shares of the varieties turns out to be negative as it surpasses its thresholds level. Similar results hold for AVA. Table 5. Dynamics of market share analysis (1) (2) (3) (4) (5) (6) Variables Distribution share Supply share Demand share Distribution share Supply share Demand share L. Distribution share -0.130*** -0.174*** -0.192*** 0.329*** 0.240*** -0.152*** (0.043) (0.047) (0.069) (0.015) (0.042) (0.026) L. Supply share 0.677*** 0.649*** 0.420*** 0.197*** 0.239*** 0.394*** (0.050) (0.040) (0.067) (0.020) (0.053) (0.028) L. Demand share 0.425*** 0.464*** 0.714*** 0.417*** 0.424*** 0.662*** (0.039) (0.020) (0.010) (0.009) (0.021) (0.003) Age 0.001*** 0.001* -0.000 (0.0004) (0.000) (0.000) Age2 -0.0002*** -0.0001*** -0.000*** 18 (0.000) (0.000) (0.000) AVAk 0.075*** 0.086*** 0.086*** (0.010) (0.007) (0.024) AVAk2 -0.003*** -0.003*** -0.003*** (0.000) (0.000) (0.001) Inflation -0.0003*** -0.0002*** -0.000 -0.000*** -0.000*** -0.000*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) GDPPC growth 0.000 0.001*** 0.000** 0.001*** 0.002*** 0.001** (0.001) (0.000) (0.000) (0.000) (0.000) (0.000) Extension transfer 0.004 0.003 0.002 0.039*** 0.049*** 0.017** (0.004) (0.003) (0.004) (0.005) (0.004) (0.007) Input subsidies 0.041*** 0.046*** 0.005 0.096*** 0.109*** 0.078** (0.008) (0.007) (0.010) (0.017) (0.013) (0.0312) Cereal yield 0.001*** 0.001*** 0.000*** 0.003*** 0.003*** 0.001*** (0.000) (0.000) (0.000) (0.000) (0.001) (0.000) Fertilizer consumption 0.003 0.004 0.025*** 0.042*** 0.102*** 0.050*** (0.006) (0.004) (0.003) (0.005) (0.010) (0.017) Constant -0.030*** -0.013** -0.020*** -0.417*** -0.440*** -0.730*** (0.009) (0.006) (0.008) (0.058) (0.047) (0.209) Observations 253 255 261 253 255 261 Number of varieties 26 26 26 26 26 26 AR1 (PV) 0.033 0.011 0.055 0.043 0.009 0.064 AR2 (PV) 0.497 0.210 0.510 0.621 0.305 0.501 Hansen test (PV) 0.857 0.388 0.438 0.896 0.822 0.451 Standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1 19 20 Figure 6. Average marginal effects of varietal age on the market share of the varieties 21 3.4. Causality analysis of maize seed system While the system GMM results discussed in Section 3.3 account for the lags of: (i) demand and distribution when modelling supply, (ii) demand and supply when estimating distribution, and (iii) supply and distribution when modelling demand, the results do not tell us potential long run relationship between the variables. Thus, in this section, we analyze the potential presence and nature of the long run relationship between demand, supply and distribution of the maize seed dynamics of Ethiopia. To investigate the potential long run relationship, we undertake cointegration test between the variables. Table A3 in the Appendix shows that, in all cases, we reject the null of no-cointegration in favor of the presence of co-integration between the variables indicating the presence of long run relationships between them. Although the cointegration results show the presence of long run relationship between the variables, they do not tell the direction of their relationship. For instance, the presence of long-term relationship between supply and distribution doesn’t tell us whether: (a) supply causes distribution, (b) distribution causes supply, or (c) bidirectional causality exists between them. Hence, we undertake causality analysis to determine the direction of the relationship between the variables. Table 6 presents the results of causality tests. In the case of the overall seeds, the result shows that we reject the null hypothesis that distribution does not granger-cause supply in favor that distribution causes supply indicating that distribution of maize seeds causes supply of maize seeds. We also reject the null hypothesis that supply does not granger cause distribution. The two results indicate the presence of bidirectional causality between supply and distribution of maize seeds. Table 6. Causality results Null hypothesis W-bar p-value Optimal number of lags BIC AIC HQIC Overall seeds Distribution does not Granger-cause supply 9.942 0.012 3 3 3 Demand does not Granger-cause supply 1.236 0.972 1 1 1 Supply does not Granger-cause distribution 1.286 0.062 2 2 2 Demand does not Granger-cause distribution 2.768 0.891 2 2 2 Distribution does not Granger-cause demand 0.002 0.003 1 1 1 Supply does not Granger-cause demand 7.359 0.000 2 2 2 Public vs private seeds Public seeds Distribution does not Granger-cause supply 1.286 0.051 2 2 2 Demand does not Granger-cause supply 1.142 0.122 1 1 1 Supply does not Granger-cause distribution 9.942 0.036 3 3 3 Demand does not Granger-cause distribution 1.302 0.295 1 1 1 Distribution does not Granger-cause demand 2.768 0.403 2 2 2 Supply does not Granger-cause demand 1.236 0.977 1 1 3 Private seeds 22 Distribution does not Granger-cause supply 1.286 0.301 2 2 2 Demand does not Granger-cause supply 1.302 0.108 1 1 1 Supply does not Granger-cause distribution 9.942 0.162 3 3 3 Demand does not Granger-cause distribution 1.202 0.148 1 1 1 Distribution does not Granger-cause demand 2.768 0.940 2 2 2 Supply does not Granger-cause demand 1.236 0.985 1 1 3 Hybrid vs OPV Hybrid seeds Distribution does not Granger-cause supply 1.286 0.008 2 2 2 Demand does not Granger-cause supply 1.359 0.103 2 2 2 Supply does not Granger-cause distribution 9.942 0.027 3 3 3 Demand does not Granger-cause distribution 1.564 0.120 Distribution does not Granger-cause demand 2.768 0.086 2 2 1 Supply does not Granger-cause demand 1.236 0.093 1 1 3 OPV seeds Distribution does not Granger-cause supply 1.286 0.370 2 2 2 Demand does not Granger-cause supply 1.305 0.148 2 2 2 Supply does not Granger-cause distribution 9.942 0.226 3 3 3 Demand does not Granger-cause distribution 1.447 0.102 2 2 2 Distribution does not Granger-cause demand 2.768 0.034 2 2 2 Supply does not Granger-cause demand 1.236 0.987 1 1 3 We fail to reject the null that demand does not granger-cause supply indicating that demand for maize seeds does not cause supply of same. However, we reject the null hypothesis that supply does not granger-cause demand indicating that supply causes demand for maize seeds. The two results imply that the relationship between supply and demand is unidirectional: supply cause demand but not the other way around. Similarly, we fail to reject the null hypothesis that demand does not granger-cause distribution while we reject the null hypothesis that distribution does not granger-cause demand. The results show unidirectional causality between demand and distribution: distribution causes demand but demand does not cause distribution. Disaggregated causality analysis shows that the results related to the public sourced and hybrid are by and large similar with that of overall seeds while causality between the variables is not much observed in the private sourced and OPV seeds. 3.5. Robustness checks 3.5.1. Instrumental variables regression One of the potential endogeneity problems is possible reverse causality given that demand, supply, and distribution of seeds apart being impacted by varietal age they can also impact varietal age. Under such a scenario, endogeneity from reverse causality arises leading to inconsistent estimation results. To address this, we use instrumental variable (IV) estimation approach. Following the literature (Edwards & Waverman, 2006; Jones & Jorgensen, 2012; Karagyozova, 2023), we use 23 ranking of the endogenous variable as IV. We argue that the rank of a seed in terms of varietal age impacts varietal age but does not directly impact the demand, supply or demand of the seed. In other words, the higher the variety’s rank in terms of varietal age the higher its varietal age compared to the other varieties. However, the variety in terms of varietal age does not directly impact demand, supply or distribution of the variety but only through impacting varietal age. Table 7 presents the results of the IV regression and shows that the finding that varietal age below (above) threshold increases (decreases) the demand, supply and distribution of the seed varieties holds when instrumenting the varietal age. The postestimation tests of the IV regression show the model is well identified: we reject the null hypothesis that the model is under-identified, and we also reject the null hypothesis of a weak instrument because the F statistic is greater than critical values. Table 7. IV regression results (1) (2) (3) Variables Distribution Supply Demand L. distribution 0.500*** 0.165** 0.143 (0.100) (0.073) (0.164) L. supply 0.109 0.486*** 0.259* (0.075) (0.077) (0.136) L. demand 0.231*** 0.436*** 0.419*** (0.045) (0.048) (0.047) Age 4.960** 2.614*** 8.098* (2.480) (0.698) (4.681) Age2 -0.111** -0.077*** -0.193** (0.043) (0.024) (0.081) Inflation 0.093 -0.020 -0.241* (0.074) (0.085) (0.137) GDPPC growth 1.637** 1.475* 3.138** (0.698) (0.795) (1.316) Extension transfer 3.368 3.758 2.235 (3.910) (4.438) (7.322) Input subsidies 52.51* 12.33 92.56* (29.56) (33.46) (55.56) Cereal yield 2.199 1.680 3.975 (2.595) (2.934) (4.872) Fertilizer consumption 37.04* 3.698 62.16 (22.29) (25.24) (41.68) Constant -5.847 -0.497 -10.24 (4.115) (4.669) (7.748) Observations 224 227 232 R-squared 0.246 0.322 0.392 Number of varieties 26 26 26 Under-identification test LM statistic 170.905 172.832 177.759 p-value 0.000 0.000 0.000 24 Weak identification test Cragg-Donald Wald F statistic 679.185 685.994 720.978 Stock-Yogo test critical value 16.38 16.38 16.38 Standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1 3.5.2. High vs low varietal age analysis To examine the robustness of our finding that varietal age nonlinearly impacts demand, supply and distribution, we use a “High age” dummy variable that takes a value 1 if varietal age is greater than or equal to the mean value and 0 otherwise. Table 8 presents the results; high varietal age reduces demand, supply, and distribution of the varieties confirming robustness of the non-linear results. Table 8. High vs low varietal age as a robustness check for nonlinear effects (1) (2) (3) Variables Distribution Supply Demand L. distribution 0.341*** 0.281*** 0.509*** (0.030) (0.027) (0.021) L. supply -0.107*** -0.300*** 0.062*** (0.013) (0.019) (0.014) L. demand 0.474*** 0.842*** 0.659*** (0.015) (0.031) (0.009) High age -14.23*** -19.39*** -7.457*** (0.681) (1.224) (0.786) Inflation -0.090*** -0.050 -0.219*** (0.025) (0.037) (0.025) GDPPC growth 1.042*** 0.889*** 1.664*** (0.113) (0.327) (0.093) Extension transfer 7.164*** 6.604** 12.87*** (1.350) (2.956) (1.639) Input subsidies 10.07** 15.84 4.987 (4.819) (13.67) (5.189) Cereal yield 0.328** 0.624** 0.456*** (0.164) (0.303) (0.125) Fertilizer consumption 21.33*** 15.78*** 30.13*** (1.836) (2.313) (2.647) Constant -8.482 -12.71 2.173 (6.228) (10.74) (6.823) Observations 253 255 261 Number of varieties 26 26 26 AR1 (PV) 0.013 0.064 0.020 AR2 (PV) 0.322 0.921 0.112 Hansen test (PV) 0.870 0.874 0.766 Standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1 25 3.5.3. Semiparametric approach of non-linearity analysis The results in Section 3.3 presupposes that the functional nexus related to the effect of varietal age on distribution, supply and demand is of quadratic nature implying the presence of one structural break. Nevertheless, the nature of the effect of varietal age on the dependent variables may take other than quadratic functional form and thus more than one structural break may exist. Thus, to check the robustness of our nonlinear result, without imposing prior assumption about the functional relationship, we follow Libois and Verardi (2013) and employ panel semiparametric model as follows: Seedikt = α0 + 𝑓(Ageit) + 𝑋it ′  + Zt ′ + αi + εikt i = 1, … , N;k = 1, . , 3;t = 1, … , T (5) The model consists of parametric and non-parametric components—the estimation deals with an estimator of parametric components 𝛽̂ and 𝛾, and standard non-parametric rate of estimated unknown function 𝑓(Ageit). Hence, the model consists of two groups of regressors—varietal age and control variables that are respectively assumed to be nonparametrically and parametrically associated with the dependent variables. Figure 6 presents the nonparametric results3 that are by and large consistent with the findings presented in Section 3.3.1. 3 The parametric results associated with control variables are not reported due to word limits but are available on request. 26 Figure 7. Semiparametric results related to the effect of varietal age on distribution, supply and demand of seed varieties 27 4. Conclusion Agricultural production and productivity are essential to meet the sustainable development goals (SDGs) such as ending hunger, eradicating poverty, and reducing income inequality across and among countries. In an increasingly growing climate change and ensuing risks and threats, development of stress tolerant seeds is vital to enable farmers cope with the challenges. Equally important is also the adoption of the newly developed varieties from the farmers’ end. However, in many developing countries in general, and in SSA in particular, farmers lag in terms of adoption of improved maize seed varieties. Against this background, using 26 varieties from 2009 to 2023, we analyze supply and distribution dynamics of maize seed varieties in Ethiopia given the importance of the crop in the country. The results show that while the overall varieties’ AVA is decreasing over time, the decrement pace appears to be slow, indicating that farmers use relatively old varieties of maize seed. Analyzing AVA by disaggregating the varieties into public and private source, and hybrid and OPV type revealed a deeper insight. While the result related to public source seeds is found to be qualitatively similar to that of the overall varieties, the AVA of seeds from the private source does not show a decreasing trend. Disaggregated analysis by seed type was more revealing: while the hybrid related findings are qualitatively similar with that of overall varieties, the AVA of OPV type increases over time. The dynamic panel analyses show that varietal age nonlinearly impacts supply and distribution of overall varieties. Accordingly, varietal age initially increases supply and distribution of maize seeds; however, varietal age above threshold reduces supply and distribution. Disaggregated analysis by source of seed by and large shows qualitatively similar nonlinear impacts of varietal age. However, disaggregated analysis by seed type reveals a different story: while the results related to the hybrid are qualitatively similar to that of overall varieties, varietal age persistently and positively impacts supply and distribution of OPV. Moreover, the nonlinear effects of varietal age hold in the cases of demand, supply and distribution of varieties’ market shares. The impact of AVA on demand, supply and distribution of the varieties (and on their market shares) is similar with that of varietal age. The results are robust to several sensitivity analyses including IV regression and semiparametric estimation. We further demonstrate long-term relationships in maize seed dynamics. The nexus between supply and distribution exhibits bidirectional causality while the relationship between supply and demand is found to be unidirectional: supply causes demand but demand does not cause supply. Similarly, while distribution of maize seeds causes farmers’ demand for the seeds, demand for same does not cause distribution. In general, the causality analyses highlight that the supply side of maize seed market shapes the long run relationship of demand for and supply of the maize seed system in Ethiopia. 28 The results imply that maize seed production and marketing focused on more recently released varieties in general and that of OPV variety in particular can be used as an instrument to reduce AVA and thereby increase varietal turnover. Put differently, besides promoting development of new maize varieties, enhancing demand for recently released maize varieties through varietal promotion accounting for the supply chain of the market is crucial. The result also suggests that future studies that pursue a disaggregated analysis of maize seed system in Ethiopia by seed source, seed type and other attributes potentially reveal more insight than pursuing an aggregated approach. One of the limitations of this study is, due to data unavailability, not accounting for more variables that potentially explain the dynamics of maize seed system in Ethiopia. Hence, as data becomes available, analyzing the maize seed system in Ethiopia along the line of targeted agroecology is one of the potential future research areas. 29 References Abate Tsedeke, Shiferaw Bekele, Menkir, A., Wegary Dagne, Kebede, Y., Tesfy, K., Menale Kassie, Bogale, G., Tadesse, B., Keno, T. (2015). Factors that transformed maize productivity in Ethiopia. Food Security, 7(5): 965-981. Arellano, M., & Bond, S. (1991). Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. The Review of Economic Studies 58, 277-297.https://doi.org/10.2307/2297968 Arellano, M., & Bover, O. (1995). Another look at the instrumental variable estimation of error- components models. Journal of econometrics, 68(1), 29-51. https://doi.org/10.1016/0304- 4076(94)01642-D Atlin, G. N., Cairns, J. E., & Das, B. (2017). Rapid breeding and varietal replacement are critical to adaptation of cropping systems in the developing world to climate change. Global food security, 12, 31-37. https://doi.org/10.1016/j.gfs.2017.01.008 Blundell, R., & Bond, S. (1998). Initial conditions and moment restrictions in dynamic panel data models. Journal of econometrics, 87(1), 115-143. https://doi.org/10.1016/S0304- 4076(98)00009-8 Bukero, G. (2023). Determinants of Maize (Zea mays L.) Varietal Turnover in Ethiopia. Bukero, G., Yirga, C., Haji, J., & Jaleta, M. (2019). Maize Varietal Adoption Rate in Ethiopia, the Farmer Self Identification and DNA Finger Printing Approaches. Chen, F., Qiu, H., Zhao, Y., Wei, X., & Wan, X. (2024). Impact of new maize variety adoption on yield and fertilizer input in China: Implications for sustainable food and agriculture. Agricultural Systems, 218, 104004. https://doi.org/10.1016/j.agsy.2024.104004 Chivasa, W., Worku, M., Teklewold, A., Setimela, P., Gethi, J., Magorokosho, C., ... & Prasanna, B. M. (2022). Maize varietal replacement in Eastern and Southern Africa: Bottlenecks, drivers and strategies for improvement. Global Food Security, 32, 100589. https://doi.org/10.1016/j.gfs.2021.100589 Choi, I. (2001). Unit root tests for panel data. Journal of International Money and Finance, 20(2), 249–272. https://doi. org/10.1016/S0261-5606(00)00048-6 De Groote, H., & Omondi, L. B. (2023). Varietal turn-over and their effect on yield and food security–Evidence from 20 years of household surveys in Kenya. Global Food Security, 36, 100676. https://doi.org/10.1016/j.gfs.2023.100676 Donovan, J., Rutsaert, P., Spielman, D., Shikuku, K. M., & Demont, M. (2021). Seed value chain development in the Global South: Key issues and new directions for public breeding programs. Outlook on Agriculture, 50(4), 366-377. https://doi.org/10.1177/00307270211059551 Edwards, Geoff, and Leonard Waverman. 2006. The effects of public ownership and regulatory independence on regulatory outcomes. Journal of Regulatory Economics 29: 23–67. Fisher, M., Abate, T., Lunduka, R. W., Asnake, W., Alemayehu, Y., & Madulu, R. B. (2015). Drought tolerant maize for farmer adaptation to drought in sub-Saharan Africa: Determinants of adoption in eastern and southern Africa. Climatic Change, 133, 283-299. https://doi.org/10.2307/2297968 https://doi.org/10.1016/0304-4076(94)01642-D https://doi.org/10.1016/0304-4076(94)01642-D https://doi.org/10.1016/j.gfs.2017.01.008 https://doi.org/10.1016/S0304-4076(98)00009-8 https://doi.org/10.1016/S0304-4076(98)00009-8 https://doi.org/10.1016/j.agsy.2024.104004 https://doi.org/10.1016/j.gfs.2021.100589 https://doi.org/10.1016/j.gfs.2023.100676 https://doi.org/10.1177/00307270211059551 30 Gatto, M., de Haan, S., Laborte, A., Bonierbale, M., Labarta, R., & Hareau, G. (2021). Trends in varietal diversity of main staple crops in Asia and Africa and implications for sustainable food systems. Frontiers in Sustainable Food Systems, 5, 626714. https://doi.org/10.3389/fsufs.2021.626714 Habte, E., Yirga, C., Jaleta, M., Tesfaye, K., Bekele, A., & Erenstein, O. (2023). Wheat seed demand assessment assisted by genotyping in Ethiopia. Experimental Agriculture, 59, e8. Hansen, L.P. (1982). Large sample properties of generalized method of moments estimators. Econometrica, 50 (4), 1029-1054. Im, K. S., Pesaran, M. H., & Shin, Y. (2003). Testing for unit roots in heterogeneous panels. Journal of Econometrics, 115(1), 53–74. https://doi.org/10.1016/S0304-4076(03)00092-7 Jaleta, M., Kassie, M., Marenya, P., Yirga, C., & Erenstein, O. (2018). Impact of improved maize adoption on household food security of maize producing smallholder farmers in Ethiopia. Food Security, 10, 81-93. https://doi.org/10.1007/s12571-017-0759-y Johansen, S. (1988). Statistical analysis of cointegration vectors. Journal of economic dynamics and control, 12(2-3), 231-254. https://doi.org/10.1016/0165-1889(88)90041-3 Jones, M. D., & Jorgensen, P. D. (2012). Mind the gap? Political advertisements and congressional election results. Journal of Political Marketing, 11(3), 165-188. https://doi.org/10.1080/15377857.2012.703909 Kalsa, K. K., Kolech, S., Worku, M., Abate, T., & Teklewold, A. (2024). The Formal Seed System of Maize in Ethiopia: Implications for Reducing the Yield Gap. International Journal of Agronomy, 2024(1), 2135433. https://doi.org/10.1155/2024/2135433 Karagyozova, T. (2023). Microinsurance and Economic Growth in Africa. Risks, 11(10), 175. https://doi.org/10.3390/risks11100175 Levin, A., Lin, C. F., & Chu, C. S. J. (2002). Unit root tests in panel data: Asymptotic and finite- sample properties. Journal of Econometrics, 108(1), 1–24. https://doi.org/10.1016/S0304- 4076(01)00098-7 Libois, F., & Verardi, V. (2013). Semiparametric fixed-effects estimator. STATA Journal, 13(2), 329–336. https://doi.org/10.1177/1536867X1301300207 Lunduka, R. W., Mateva, K. I., Magorokosho, C., & Manjeru, P. (2019). Impact of adoption of drought-tolerant maize varieties on total maize production in south Eastern Zimbabwe. Climate and development, 11(1), 35-46. https://doi.org/10.1080/17565529.2017.1372269 Marenya, P., Wanyama, R., Alemu, S., Westengen, O., & Jaleta, M. (2022). Maize variety preferences among smallholder farmers in Ethiopia: Implications for demand-led breeding and seed sector development. Plos one, 17(9), e0274262. Nuthalapati, C. S., Kumar, A., Birthal, P. S., & Sonkar, V. K. (2024). Demand-side and supply-side factors for accelerating varietal turnover in smallholder soybean farms. Journal of Cleaner Production, 447, 141372. https://doi.org/10.3389/fsufs.2021.626714 https://doi.org/10.1016/S0304-4076(03)00092-7 https://doi.org/10.1016/0165-1889(88)90041-3 https://doi.org/10.1080/15377857.2012.703909 https://doi.org/10.1155/2024/2135433 https://doi.org/10.3390/risks11100175 https://doi.org/10.1016/S0304-4076(01)00098-7 https://doi.org/10.1016/S0304-4076(01)00098-7 https://doi.org/10.1080/17565529.2017.1372269 31 Rutsaert, P., & Donovan, J. (2020). Sticking with the old seed: input value chains and the challenges to deliver genetic gains to smallholder maize farmers. Outlook on Agriculture, 49(1), 39-49. https://doi.org/10.1177/0030727019900520 Rutsaert, P., Donovan, J., & Kimenju, S. (2021). Demand-side challenges to increase sales of new maize hybrids in Kenya. Technology in Society, 66, 101630. https://doi.org/10.1016/j.techsoc.2021.101630 Rutsaert, P., Donovan, J., Murphy, M., & Hoffmann, V. (2024). Farmer decision making for hybrid maize seed purchases: Effects of brand loyalty, price discounts and product information. Agricultural Systems, 218, 104002. https://doi.org/10.1016/j.agsy.2024.104002 Singh, R. P., Chintagunta, A. D., Agarwal, D. K., Kureel, R. S., & Kumar, S. J. (2020). Varietal replacement rate: Prospects and challenges for global food security. Global Food Security, 25, 100324. https://doi.org/10.1016/j.gfs.2019.100324 Worku, M., Twumasi Afriyie, S., Wolde, L., Tadesse, B., Demisie, G., Bogale, G., ... & Prasanna, B. M. (2012). Meeting the challenges of global climate change and food security through innovative maize research. Proceedings of the National Maize Workshop of Ethiopia, 3; Addis Ababa, Ethiopia; 18-20 April, 2011. CIMMYT. Worku, M., De Groote, H., Munyua, B., Makumbi, D., Owino, F., Crossa, J., ... & Prasanna, B. M. (2020). On-farm performance and farmers’ participatory assessment of new stress-tolerant maize hybrids in Eastern Africa. Field Crops Research, 246, 107693. https://doi.org/10.1016/j.fcr.2019.107693 https://doi.org/10.1177/0030727019900520 https://doi.org/10.1016/j.techsoc.2021.101630 https://doi.org/10.1016/j.agsy.2024.104002 https://doi.org/10.1016/j.gfs.2019.100324 https://doi.org/10.1016/j.fcr.2019.107693 32 Appendix Table A1. Summary of mean and share of maize varieties under consideration Seed Demand Supply Distribution Demand (%) Supply (%) Distribution (%) BH661 872.691 872.576 789.926 12.867 21.628 22.473 BH660 1114.716 702.718 582.776 16.435 17.418 16.580 BH545Q 16.160 17.551 3.834 0.238 0.435 0.109 BH543 206.786 109.000 89.884 3.049 2.702 2.557 BH542 21.715 55.543 45.387 0.320 1.377 1.291 BH540 1157.487 568.508 448.734 17.066 14.091 12.766 BH140 454.256 240.687 228.873 6.697 5.966 6.511 BH547 11.928 17.975 17.159 0.176 0.446 0.488 BH546 106.098 166.852 148.150 1.564 4.136 4.215 Melkasa 1Q 15.088 2.097 1.768 0.222 0.052 0.050 Melkasa 2 111.754 89.503 74.465 1.648 2.218 2.118 Melkasa 4 43.625 27.469 19.768 0.643 0.681 0.562 Melkasa 6Q 22.414 13.341 7.589 0.330 0.331 0.216 MH140 25.571 23.185 22.076 0.377 0.575 0.628 MH138Q 7.776 10.444 6.756 0.115 0.259 0.192 Jabii (PHB3253) 260.015 80.246 70.190 3.834 1.989 1.997 Shone (30G19) 894.400 458.448 430.820 13.187 11.363 12.256 Shala (P2859W) 121.444 45.247 37.129 1.791 1.122 1.056 Limu (P3812W) 1100.198 416.668 399.297 16.221 10.328 11.360 PHB 30 V 53 159.279 86.537 64.618 2.348 2.145 1.838 PHB 30 H 83 38.418 14.779 13.722 0.566 0.366 0.390 Jibat (AMH 851) 6.540 3.842 3.531 0.096 0.095 0.100 Hawasa 1 3.904 1.104 0.542 0.058 0.027 0.015 Guto 0.800 1.812 1.821 0.012 0.045 0.052 Gibe 1 4.871 4.915 3.971 0.072 0.122 0.113 Giber 2 4.605 3.365 2.255 0.068 0.083 0.064 33 Table A2. Summary statistics of the main variables of the maize seed system in metric tons Overall varieties Variable Obs Mean Std. Dev. Min Max Demand 352 19.269 37.857 0 252.635 Supply 335 12.043 24.692 0 185.502 Distribution 333 10.556 21.976 0 173.135 Public Demand 251 16.749 33.815 0 252.031 Supply 235 12.466 27.157 0 185.502 Distribution 233 10.716 23.897 0 173.135 Private Demand 101 25.529 46.003 0 252.635 Supply 100 11.048 17.674 0 75.111 Distribution 100 10.181 16.768 0 74.940 Carryover 99 0.819 3.750 -19.930 21.437 Shortage 97 14.776 34.421 -20.738 199.924 Hybrid varieties Demand 273 24.249 41.650 0 252.635 Supply 258 15.194 27.325 0 185.502 Distribution 257 13.325 24.310 0 173.135 OPV varieties Demand 79 2.059 3.645 0 20.675 Supply 77 1.485 2.811 0 12.050 Disttribution 76 1.192 2.369 0 10.019 34 Table A3. Cointegration test results Null hypothesis Tests (P value) Modified Phillips–Perron Phillips–Perron No cointegration, between demand and supply 0.011 0.023 No cointegration, between demand and distribution 0.003 0.015 No cointegration, between supply and distribution 0.000 0.000 Figure A1. Mean values of the varieties, from 2009 to 2023 Figure A1.1. Mean values (in metric tons) of maize seed varieties by year, from 2009 to 2023 Figure A2. Trends of demand, supply and distribution 0 200 400 600 800 1000 1200 BH 66 1 BH 66 0 BH 54 5Q BH 54 3 BH 54 2 BH 54 0 BH 14 0 BH 54 7 BH 54 6 M el ka sa 1 Q M el ka sa 2 M el ka sa 4 M el ka sa 6 Q M H 14 0 M H 13 8Q Ja bi i (P H B 32 53 ) Sh on e (3 0G 19 ) Sh al a (P 28 59 W ) Li m u (P 38 12 W ) PH B 3 0 V 53 PH B 3 0 H 8 3 Jib at (A M H 8 51 ) H aw as a 1 G ut o G ib e 1 G ib er 2 Mean values of the varities by year, from 2009 to 2023 Demand Supply Distribution 35 Figure A2.1. Trends of demand, supply, and distribution of maize by sources of seeds Figure A2.2. Trends of demand, supply, and distribution in metric tons of maize by variety types 0 50 100 150 200 250 300 350 400 450 20 09 20 10 20 11 20 12 20 13 20 14 20 15 20 16 20 17 20 18 20 19 20 20 20 21 20 22 20 23 Trends of demand, supply and distribution by public and private sources Demand public Supply public Distribution public Demand private Supply private Distribution private 0 100 200 300 400 500 600 700 20 09 20 10 20 11 20 12 20 13 20 14 20 15 20 16 20 17 20 18 20 19 20 20 20 21 20 22 20 23 Trends of demand, supply and distribution by hybrid and OPV variety types Demand hybrid Supply hybrid Distribution hybrid Demand OPV Supply OPV Distribution OPV 36 Figure A2.3. Trend of public and private shares in demand of maize seed varieties Figure A2.4. Trend of public and private shares in supply and distribution of maize seed varieties 0 10 20 30 40 50 60 70 80 90 100 2 0 0 9 2 0 1 0 2 0 1 1 2 0 1 2 2 0 1 3 2 0 1 4 2 0 1 5 2 0 1 6 2 0 1 7 2 0 1 8 2 0 1 9 2 0 2 0 2 0 2 1 2 0 2 2 2 0 2 3 Maize seed demand by share of public and private source Demand public (%) Demand private (%) 0 10 20 30 40 50 60 70 80 90 100 2 0 0 9 2 0 1 0 2 0 1 1 2 0 1 2 2 0 1 3 2 0 1 4 2 0 1 5 2 0 1 6 2 0 1 7 2 0 1 8 2 0 1 9 2 0 2 0 2 0 2 1 2 0 2 2 2 0 2 3 Maize seed supply and distribution by share of public and private source Supply public (%) Supply private (%) Distribution public (%) Distribution private (%) 37 Figure A2.5. Trend of hybrid and OPV shares in demand, supply, and distribution of maize seed varieties 0 10 20 30 40 50 60 70 80 90 100 2 0 0 9 2 0 1 0 2 0 1 1 2 0 1 2 2 0 1 3 2 0 1 4 2 0 1 5 2 0 1 6 2 0 1 7 2 0 1 8 2 0 1 9 2 0 2 0 2 0 2 1 2 0 2 2 2 0 2 3 Trends of demand, supply and distribution by share of hybrid and OPV type Demand hybrid (%) Demand OPV (%) Supply hybrid (%) Supply OPV (%) Distribution hybrid (%) Distribution OPV (%) 38 Figure A2.6. Average varietal age of demand, supply and distribution of overall maize varieties 0 2 4 6 8 10 12 14 16 18 20 09 20 10 20 11 20 12 20 13 20 14 20 15 20 16 20 17 20 18 20 19 20 20 20 21 20 22 20 23 Average varietal age of overall varities Demand Supply Distribution