Assessing homogeneous and heterogeneous economic impacts of fall armyworm management practices on farm performance in Ghana

One of the emerging challenges impinging on sustainable food production in sub-Saharan Africa is the invasion of the fall armyworm (FAW) pest. Data collected from farm households in different agro-ecological zones in Ghana and a Multivalued Treatment Effect (MVTE) model were used to argue that FAW management practices are key to stemming the debilitating effects of FAW infestations on farm performance. Previous studies have analysed homogeneous treatment effects to explain FAW management practices. The present study employs heterogeneous treat-ment effects to account for the differential effects of FAW management practices such as early planting, application of prescribed pesticides, and a combination of both practices while accounting for socioeconomic characteristics. Findings from the study reveal that distance to extension services exerts negative effects on adoption of early planting only, and adoption of both FAW management practices. Moreover, past FAW infestations tend to exert a positive effect on adoption of pesticide application only and adoption of a combination of the FAW management practices. Adopting the FAW management practices examined in this study assisted the maize producers in raising their farm performance. Socioeconomic characteristics also tend to influence the homogeneous treatment effects of adoption of the FAW management practices on farm performance. This finding indicates that heterogeneity within farm households is crucial for policy regarding adoption of FAW management practices.


Introduction
The spread of the invasive crop pest FAW (Spodoptera frugiperda) is one of the emerging environmental and ecological challenges facing many countries in sub-Saharan Africa (Tambo et al. 2020) whose economies heavily depend on agriculture (De Groote et al. 2020;Tambo et al. 2020).The fall armyworm (FAW) is an insect pest that feeds on over 80 different crop plants, including cereals such as maize, rice, and sorghum, as well as legumes, vegetable crops, and cotton (Goergen et al. 2016;FAO 2017;Day et al. 2017;Hailu et al. 2018).
Maize is a staple food crop in sub-Saharan Africa (SSA) and a source of on-farm income generation for smallholder farmers.In Ghana, for instance, income from maize production alone constitutes about 30-40% of the total household earnings of smallholder farmers (GSS 2017;2020) and it constitutes about 50-60% of the Page 2 of 18 Kondo et al. CABI Agriculture and Bioscience (2024) 5:70 country's cereal production (Obour and Arthur 2022).
Growing evidence on FAW outbreaks in sub-Saharan Africa (SSA) suggests that smallholder farmers producing maize experienced the greatest impact, recording the highest economic losses in terms of yield and income, especially in 2017 and 2018 (Day et al. 2017;Rwomushana et al. 2018;Tambo et al. 2019Tambo et al. , 2020;;Baudron et al. 2019).In contextualising the effect on Ghana, the success of the flagship programme of Planting for Food and Jobs (PFJ) by the Government of Ghana (GOG), for instance, would not be tenable if the current trend of FAW devastation continues.
To be able to keep FAW pests below the economic injury level in maize production, some studies have argued that policy should focus on long-term management of FAW infestations instead of eradication (Kumar et al. 2022;Lowry et al. 2022).In line with this, the Food and Agriculture Organisation (FAO) has advocated for the implementation of pragmatic policies that promote the management of FAW to curb the destruction of food crops (FAO 2018).The FAW management practices recommended by the FAO include early planting (or avoiding late planting) and staggered planting (plots of different ages), intercropping of maize with tuber crops such as yam or cassava, and maize-legume intercropping.Other recommended practices include hand picking and destroying of egg masses, and the broad adoption of integrated pest management practices (FAO 2017;Hailu et al. 2018;Tefera et al. 2019;Kenis et al. 2019;Tambo et al. 2019;Babendreier et al. 2020).The prescribed use of synthetic pesticides in farms under the guidance of Agricultural Extension Agents (AEAs) during periods of heavy FAW infestations is also part of the recommended management protocols (FAO 2017;Day et al. 2017).Within the broader framework of agricultural technology adoption, adopting technologies or management practices as a package rather than as a stand-alone technology or practice has been highlighted (Manda et al. 2016;Suri and Udry 2022).This study focussed on early planting and pesticide application, or a combination of these two management practices because they were the two most predominant cross-cutting FAW management practices adopted by the farmers in the region.
Most of the existing studies on FAW have focused on the perceptions, knowledge, and current FAW management practices (Asare-Nuamah 2022; Koffi et al. 2020;Bariw et al. 2020).Other studies have analysed the determinants of FAW management practices (Tambo et al. 2020), and the potential constraints to the use of FAW management practices (Chimweta et al. 2020;Ngangambe et al. 2020).The study by Tambo et al. (2020) employed cross-country data from Ghana, Rwanda, Uganda, Zambia, and Zimbabwe to analyse the determinants of FAW management practices.Using a household and village fixed-effects modelling approach, Kassie et al. (2020) analysed how FAW pest affects maize yield, quantities of maize sales, insecticide use, and maize consumption in Ethiopia.Similarly, Tambo et al. (2023) examined the dynamics of the effect of FAW on maize productivity using fixed effects and correlated random effects models in a panel data setting; Banson et al. (2020) employed the systems thinking approach to examine the threat posed by FAW attacks on maize production and its implications on food security in Ghana; and Michler et al. (2019) used a three-wave panel data to estimate a correlated random coefficient model and calculated the returns to improved chickpea in terms of yields, costs, and profits in Ethiopia.The knowledge gap in the existing literature regarding FAW infestations is apparent because most of the studies have not analysed the direct impacts of FAW management practices on farm performance, instead only emphasising homogeneous treatment effects.Because of this, it is important to fill the gap by assessing the economic impacts of fall armyworm management practices on farm performance from both homogeneous and heterogeneous perspectives.Sauer and Moreddu (2020) pointed out that farm performance may include attributes of the farm that relate to economic performance, environmental sustainability performance, as well as social or cultural performance.
FAW management practices come in the form of packages or alternatives, highlighting the relevance of applying appropriate empirical approaches to avoid spurious findings that have the tendency to distort policy recommendations.In the present study, we employ the Multivalued Treatment Effects model to explain the effects of the different alternatives of FAW management practices on farm performance.The model employed in this study allowed us to control for observable characteristics.The novelty of the present study in expanding the literature on FAW infestation, and its implications on farm performance is that (1) We use representative data from maize-producing households located in different agroecological zones in Ghana to analyse the direct impacts of FAW management practices on farm performance, and (2) The study enriches the literature on the impacts of FAW management practices by analysing both homogeneous and heterogeneous effects of FAW management practices on farm performance.Specifically, two outcome variables, namely maize yield and household maize income, were assessed to ensure succinct policy recommendations on sustainable food production in sub-Saharan Africa.
The findings from the study suggest significant variations in maize yield and household maize income between adopters and non-adopters of early planting only, pesticide application only, and a combination of the two practices.We also find that heterogeneous treatment effects across socioeconomic characteristics are crucial for policy regarding adoption of FAW management practices.Succinct policy recommendations on FAW management practices to ensure the sustainability of the Government of Ghana's Planting for Food and Jobs (PFJ) programme have been provided in the study.
The paper is organised as follows.Sect."Introduction" has introduced the study.Sect."The fall armyworm epidemic and farm performance in Ghana" provides a brief overview of the FAW epidemic and farm performance in Ghana.Sect."Materials and methods" discusses the materials and methods.Sect."Results and discussion" presents the results and discussion.Sect."Conclusion and policy recommendations" provides the conclusion and policy recommendations.

The fall armyworm epidemic and farm performance in Ghana
The fall armyworm originated from the tropical and subtropical regions of the Americas, with the adult moth having the ability to move over 100 km per night (Goergen et al. 2016;FAO 2017).The FAW pest lays its eggs on plants, the larvae then hatch and begin feeding.According to FAO (2017), farmers in the Americas have made an effort to manage the pest, but with significant associated costs.In Africa, the outbreak of FAW was first reported in 2016 in São Tomé and Príncipe Islands and Nigeria (Goergen et al. 2016).Studies indicate that the introduction of the FAW pest into Africa and Asia may have come from direct air flights (Baudron et al. 2019;Tambo et al. 2019).It is now known that about 46 African and some Asian countries have recorded cases of FAW infestations in agricultural production (Goergen et al. 2016).Findings suggest that FAW infestation levels in the Semi-deciduous, Guinea Savannah and Transitional agro-ecological zones were higher than in the Sudan Savannah agro-ecological zone (Koffi et al. 2020;Babendreier et al. 2020).
The FAW pest, therefore, poses a significant threat to farm performance in Ghana in terms of possible yield losses and income shocks to thousands of smallholder maize farmers (MoFA 2017; Banson et al. 2020;Bariw et al. 2020).Pasanna et al. (2018) argued that the FAW pest has the capability of causing extensive harm to other staple food crops as well.The invasion is a food security threat and must be addressed from a management perspective by African governments since over 300 million people in sub-Saharan Africa derive their livelihood from the crop (Macauley 2015;Day et al. 2017;Wossen et al. 2017;Tambo et al. 2019;2020).From the 2016 to 2018 farming seasons about 14,247 hectares (ha) of cultivated farmlands were destroyed by the FAW outbreak leading to reductions in both potential and actual maize yields in Ghana (Bariw et al. 2020;Koffi et al. 2020).
Maize yields increased annually from 2013 to 2015 but slowed down from 2016 to 2018 in Ghana as FAW infestation was severe during these periods (Koffi et al. 2020).Reduction in maize yield due to FAW attack reduces the quantity of maize available for sale and consumption, which has economic implications.Household maize income constitutes about 30-40% of total household income for most smallholder farmers (Rapsomanikis 2015).During the 2018 minor and the 2019 major cropping seasons, the period of the study, FAW infestation accounted for 30% and 43% of maize grain yield loss, respectively, in the surveyed regions and districts (Yeboah et al. 2023).At the peak of the FAW outbreak in Ghana during the 2016/2017 cropping season, an estimated 125,000 ha of maize farms were affected, leading to economic losses of about US$64 million (Day et al. 2017).Hence effective national capacity is required to detect and monitor plant pests to create the enabling environment for the private-sector-led modernisation of Ghana's agriculture by ensuring access to agro-chemicals and promoting plant disease prevention and control measures (Williams et al. 2021).

Study area, sampling procedure, and data collection
The present study uses representative data collected from maize farming households in nine administrative regions of Ghana across five agro-ecological zones in 2019 (see Fig. 1).The nine administrative regions selected for the study account for more than 80% of the regions producing maize grains in Ghana (Obour and Arthur 2022).
A multi-stage sampling procedure was employed in the data collection.First, the nine administrative regions were purposively selected because of the predominant maize production, and occurrence of FAW infestations in these regions.Second, representative districts were randomly sampled from each agro-ecological zone (see Table 1).Third, four communities from each of the selected districts in the regions and agro-ecological zones were randomly sampled.Fourth, a total of 455 maizefarming household heads were randomly sampled from the selected regions covering the five agro-ecological zones in the country (see the distribution of the selected maize-producing households in Table 1).The different number of districts and farm households sampled for the study was done to reflect the disparities in the volume of maize production and the share of the population engaged in the production of the cereal.
We collected the primary data with semi-structured questionnaires.The data collected included household, farm, institutional, and agro-ecological characteristics, and data on outcome indicators.Outcome indicators comprise maize yield (maize output per hectare) and household maize income, measured as the value of maize sales representing the amount of maize the household sold during the 2019 marketing season (Ghana Cedis per hectare), and the treatment variable, comprising FAW management practices adopted by the households.The predominant cross-cutting FAW management practices adopted by the farmers include early planting only, pesticide application only, and a combination of the two FAW management practices.The proportions of maize farmers adopting other management practices such as maize-legume intercropping, mixed cropping, removal, and burning of affected plants were relatively minimal, and hence could not be considered in the empirical estimation of this study.

Conceptual framework and hypotheses
The conceptual framework providing the interlinkages between the adoption of FAW management practices, the independent variables that drive the adoption of the respective FAW management practice, and their impacts on the outcome variables under consideration is shown in Fig. 2.
Adoption of FAW management practices may be affected by household, farm, and institutional characteristics, and agro-ecological dummies.The household characteristics include age, gender, education, and household size.The farm characteristics include farm size, owner-cultivation, major-minor production seasons, other pest and disease attacks, and past FAW infestation.The institutional characteristics include membership in a farmer-based organisation, extension The adoption of FAW management practices is likely to be affected by a wide range of explanatory variables or covariates in both the outcome and treatment equations as indicated in the conceptual framework.We consider the age of the farmer, among other variables.Older farmers are expected to be more experienced in methods of land preparation and production technologies (Ng'ombe et al. 2017;Bariw et al. 2020).The household size variable is used as a proxy for labour availability (Ng'ombe et al. 2017;Khonje et al. 2018).We postulate that households with healthier and more productive aged group members could increase adoption of FAW management practices.However, households with many young children and members needing care could result in a labour-reducing effect; hence, the expected adoption effect would be indeterminate.
Evidence suggests that attacks by other pests and diseases and past experience with FAW infestations may potentially stifle maize production and contribute to a reduction in maize yield (Tambo et al. 2019).Therefore, attacks by other pests and diseases, and past experience with FAW infestations are likely to increase the probability of adopting FAW management practices.Shorter distance to the nearest extension office would facilitate easy interactions between farmers and extension agents.Therefore, we expect a negative correlation between the distance to the nearest extension office and the adoption of FAW management practices.We expect extension contacts between farmers and agricultural extension agents to have a positive effect on the adoption of FAW management practices.Procuring farm inputs for FAW control requires substantial investment (Kassie et al. 2020).Therefore, credit-constrained households who are willing to access formal or informal credit but are unable to do so due to high transaction costs are less likely to adopt FAW management practices.The infestations of FAW pests tend to be associated with agro-ecological characteristics of farming households (Tambo et al. 2023;Kassie et al. 2020).We use agro-ecological zone dummy variables to explain the agro-ecological effects of FAW invasion on maize production.The Guinea Savannah, Sudan Savannah, Transitional, Semi-deciduous, and Coastal Savannah agro-ecological zones in Ghana were examined.The Guinea, Sudan, and Coastal Savannah agro-ecological zones, for instance, are unique in that they lack the diverse landscapes that provide shelter and perches for preying birds, parasites, predators, and other natural enemies that mitigate against the damage of FAW invasion (Maas et al. 2013;2016;Baudron et al. 2019;Harrison et al. 2019).Maize farmers in these zones recorded higher levels of larval infestations during the 2016 to 2018 maize production seasons (Koffi et al. 2020).We expect the probability of adopting FAW management practices by maize farmers in these zones to be higher.The Semi-deciduous agro-ecological zone, on the other hand, is characterised by big trees and thick forests, which enhance the abundance of birds and bats as natural enemies of FAW pests (Maas et al. 2013(Maas et al. , 2016;;Baudron et al. 2019).Therefore, we expect the Semi-deciduous agro-ecological zone dummy to exert a negative effect on the adoption of FAW management practices.Moreover, the practice of minimum or zero-tillage maize farming in this zone tends to lower FAW damage relative to the Guinea and Savannah agro-ecological zones where maize farmers use plough-based tillage systems (Harrison et al. 2019;Baudron et al. 2019).The Transitional agro-ecological zone possesses both attributes of Guinea Savannah, Sudan Savannah, and Semi-deciduous agro-ecological zones.Apriori, the effect of the Transitional zone on the adoption of FAW management practice is indeterminate.

Empirical strategy
In this study, we analyse the economic effects of adoption of FAW management practices on farm performance (i.e., maize yield and household maize income) with the Multivalued Treatment Effects (MVTE) model.The treatment effects in the MVTE model were computed using the Inverse Probability Weighting Regression Adjustment (IPWRA) approach.The MVTE could control for observables based on the unconfoundedness assumption (Liden et al. 2016) Assuming that a rational maize farmer i maximises expected utility by adopting a FAW management practice Ŵ i from j alternatives, then the multivalued treatment variable, 2016), the potential outcome, G i from adopting the FAW management practices is expressed as: The population treatment effect is computed as the difference between the means of the two potential outcomes: where G ij − G ik denotes the individual treatment effect of adopting the FAW management practices.
The average treatment effect ATE jk of adopting FAW management practice j relative to FAW management practice k over the whole population is computed as: where Z i denotes a vector of observed pre-treatment var- iables (e.g., household, farm, and institutional characteristics, and agro-ecological zone effects), δ i = α 0j − α 0k , with α 0j and α 0k are constants, γ i = (α 1j − α 1k ) where α 1j and α 1k are the estimated parameters, and N denotes the total number of observations.
The average treatment effect on the treated (ATET) is computed (1) Although previous studies have analysed homogeneous treatment effects, this study further computes heterogeneous treatment effects to account for heterogeneity within adopters of the FAW management practices in terms of their socioeconomic characteristics.Wonsen et al. ( 2017) rightly pointed out that focusing only on homogeneous effects may distort differential policyrelated effects of the adoption of a technology across rural households.We computed heterogeneous treatment effects of adoption of FAW management practices on farm performance using the approach by Verhofstadt and Maertens (2014) and Wonsen et al. (2017).With this estimation strategy, we first regressed the socioeconomic characteristics on predicted homogeneous treatment effects using Ordinary Least Square (OLS).Second, the average heterogeneous treatment effects on farm performance (maize yield and household maize income) were computed.In this study, we ensure that the treatment effect estimates from the MVTE estimates were robust by comparing them with those from the IPWRA and PSM models.

Descriptive results
Table 2 presents the descriptive statistics of the variables used in the regression models.The proportion of farmers adopting early planting only was 13%, pesticide application only was 33%, and both FAW management practices was 36% (see Table 2).We find variations in the outcomes between adopters and non-adopters of the FAW management practices.Notably, farmers adopting early planting only were less credit-constrained than non-adopters.Farmers located in the Sudan, Guinea, and Transitional agro-ecological zones adopting early planting only and pesticide application were statistically different from non-adopters.

Factors influencing FAW management practices
The marginal effects from the multinomial logit model, which are relevant for policy (Khonje et al. 2018), are presented in Table 3.The estimated coefficients are found in Table A1 in the Appendix.The results show that a kilometre increase in the distance to the nearest extension office tends to decrease the probability of adopting only early planting by 1.3%, pesticide application only by 1.9%, and adopting both FAW management practices by 1.7%, respectively; and maize farmers who have experienced FAW infestations before tend to increase their probability of adopting pesticide (4) application only by 9.6%.The probability of adopting only pesticide application increases by about 10% for maize farmers who experience other pests and disease attacks compared to those who did not experience such attacks.For instance, farmers whose maize farms are often attacked by maize stem borer, maize streak virus, and downy mildew disease, among others, are likely to put in more effort in adopting management practices to control them.The probability of adopting both early planting and pesticide application as FAW management practices tends to decrease by 12% for farmers who experience other pests and diseases on their maize farms relative to those who do not.This result is likely because farmers experiencing multiple pests and diseases tend to face significant resource constraints, including financial, labour, and time limitations.Managing various pest infestation threats simultaneously complicates their pest management strategies, leading to a prioritisation of the most immediate or severe issues, and subsequent lower adoption rates of the combination of the FAW management practices.The probability of adopting early planting only by the maize producers in the Sudan Savannah, Semi-deciduous Forest, and the Coastal-Savannah agro-ecological zones tends to increase by 14%, 11.2%, and 18%, respectively, relative to those in the Transitional agro-ecological zone.These results suggest that smallholder maize farmers rely heavily on early planting only as FAW management practice in maize production in these three agro-ecological zones compared to farmers in the Transitional agro-ecological zone.These results are consistent with previous studies on FAW larval infestation levels in agro-ecological zones of Ghana (Harrison et al. 2019;Koffi et al. 2020).Relative to maize producers in the Transitional agro-ecological zone of Ghana, those in the Guinea-Savannah agro-ecological zones tend to increase their probability of adopting both early planting and pesticide application by 14% whilst farmers located in the Semi-deciduous Forest and the Coastal Savannah agroecological zones tend to decrease their probability of adopting both early planting and pesticide application as FAW management practice by 29% and 32%, respectively.This implies that in the Coastal Savannah agro-ecological zone of Ghana, farmers use mostly early planting only to combat FAW pests in maize farms.The empirical results generally suggest that the agro-ecological zone in which the farmer is located plays a significant role in the adoption of FAW management practices in Ghana.

Homogeneous impacts of FAW management practices
The homogeneous average treatment effects (ATE) and average treatment effects on the treated (ATET) from the MVTE models are presented in Table 4. Generally, adopting FAW management practices assisted the farmers in increasing their maize yields and household maize income.Notably, adopting early planting only increased maize yield by 16.8% and household maize income by 9.6%.Adopting pesticide applications only increased maize yield by 19.6% and household maize income by 9.6% whilst adopting both FAW practices increased maize yield by 21.6% and household maize income by 10.9%.Our findings are consistent with a report by GSS (2017) that increased maize yields as a result of adoption of FAW management practices may have translated into increased household maize income in Ghana.Michler et al. (2019), however, cautioned that yield increases may not necessarily drive technology adoption.
Notably, farming households adopting both early planting and pesticide application as FAW management practices tend to have relatively higher expected maize yields and household maize income than those adopting either early planting only or pesticide application only (see Table 4).We observe relatively higher percentage changes in treatment effects for a combination of FAW management practices than that of the individual FAW management practices.These empirical results imply that adopting FAW management practices in multiple packages provides higher economic gains than adopting single or stand-alone management practices.Our empirical results agree with Tambo et al. (2019) and Koffi et al. 2020 who found that adopting both early planting and pesticide application as FAW management practice is more likely to suppress the multiplication of the FAW larvae in maize farms.This way, maize producers would be able to harvest their crops as early as possible to escape the onslaught of FAW attack during the vegetative stage of maize growth when the damage of FAW pest attack is most severe.
The empirical estimates from the Inverse Probability Weighted Regression Adjustment (IPWRA) and the Propensity Score Matching (PSM) models used to judge whether the treatment effect estimates from the MVTE model are robust are presented in Table A2 in the Appendix.The treated and control groups of the three FAW management practices examined in this study balanced quite well in the PSM model (see Fig. 3).We find that the percentage change estimates from the IPWRA model were relatively higher (both ATE and ATET) compared to our MVTE model, but the percentage change estimates for the PSM model were rather lower.These empirical results suggest that the treatment effect estimates from our MVTE tend to be more robust when compared to those from the IPWRA and the PSM models.

Heterogeneous impacts of FAW management practices
The heterogeneous average treatment effects (ATE) of adoption of FAW management practices on maize yield and household maize income are shown in Table 5.The results are presented in two parts.First, we show the estimated Ordinary Least Squares (OLS) results after we regressed the socioeconomic characteristics on the homogeneous treatment effects; and second, we present the estimated heterogeneous treatment effects after controlling for the socioeconomic characteristics within the farm households.The OLS results reveal significant heterogeneous effects for adoption of early planting only after controlling for socioeconomic characteristics except being a male, number of years of schooling, FBO, and Guinea-Savannah agro-ecological zone (see column 1 of Table 5).Moreover, we find significant heterogeneous effects for adoption of pesticide application only after controlling for socioeconomic characteristics except for age, being a male, and farm size (see column 2 of Table 5); and for adoption of both FAW management practices after controlling for socioeconomic characteristics except gender (male), own land, and Guinea Savannah agroecological zone.Similarly, age, being a male, number of years of schooling, other pests and disease attacks, membership of farmer-based organization, Guinea-Savannah, Semi-deciduous, and Coastal-Savannah agro-ecological zones, all show highly statistically significant heterogeneous effects of early planting only, pesticide application only and both FAW management practices on household maize income (see columns 4, 5, and 6 of Table 5).The findings from the study underscore the relevance of not neglecting the heterogeneous effects of adoption of FAW management practices on maize yield and household maize income across socioeconomic characteristics (Verhofstadt and Maertens 2014; Wossen et al. 2017).
The heterogeneous ATEs indicate that adopting the FAW management practices makes the maize farmers better off than non-adoption.The estimates suggest differential effects of the adoption of FAW management practices across the socioeconomic characteristics examined in this study (see Table 5).From the treatment effect results it is clear that the heterogeneous ATEs are slightly lower than the homogeneous ATEs (see Tables 4 and 5).The significant heterogeneous effects suggest a statistically significant difference in effects within the farm households based on their socioeconomic characteristics (see Table 5); and suggest that by controlling for socioeconomic characteristics, adoption of FAW management practices tends to have statistically significant heterogeneous impacts on maize yields and household maize income.
Heterogeneous treatment effects of adoption of FAW management practices on farm performance are therefore crucial in the design of relevant policy instruments on FAW management practices in Ghana in particular, and sub-Saharan Africa in general.The work by Suri and Udry (2022) argued that adoption of technology by rural households in Africa is sensitive to local circumstances such as differences in agro-ecological locations of the farm households.Notably in Ghana, expected gains from the Government's flagship programme of Planting for Food and Jobs (PFJ) can only be sustained if stakeholders pay particular attention to the promotion of adoption of FAW management practices due to the outbreak of FAW during the initial stages of implementation of the programme.

Conclusion and policy recommendations
The invasion of the fall armyworm (FAW) pest in sub-Saharan Africa, and its impacts on farm performance are gaining the attention of researchers and policymakers.Maize producers across the affected countries have recorded massive destruction of farms due to FAW pest attacks.Using data collected from maize-producing households in the majority of the regions in Ghana, the present study argues that using FAW management practices holds the key to stemming the debilitating effects of FAW infestations on sustainable food production.
Our study provides key empirical findings and policy recommendations that researchers, relevant stakeholders, and policymakers could use to ensure sustainable food production in the affected FAW infestation regions.
Our findings indicate that household and farm characteristics tend to influence adoption rates of early planting only or pesticide application only.Moreover, agro-ecological dummies correlate with adoption of early planting only and pesticide application only as FAW management practices.The results from the study generally show that maize-producing households adopting FAW management practices significantly increased their farm performance.These findings are particularly important as farming households with increased maize yields and income are economically better off in overcoming household food deficits and income shocks during periods of peak FAW infestations.We find that maize producers who adopted a combination of the FAW management practices were better off and would have obtained lower maize yield and Balance plot for early planting only   household maize income had they not adopted the FAW management practices in maize farms.These findings suggest that relevant stakeholders promote targeted policies that take the specific agro-ecological zones where the farm households being affected by the FAW pest infestations are located into considera-We recommend that adoption of early planting only, pesticide application only, and the combination of these FAW management practices should be promoted among maize producers in regions and countries susceptible to FAW infestations.In particular, in Ghana, we recommend that the policy objectives outlined in the Food and Agricultural Sector Development Policies (FASDEP II and FASDEP III) on the need to guide the management of pest and disease incidences to spur agricultural production and productivity should be fully implemented by the Ministry of Food and Agriculture (MoFA) by highlighting the FAW management practices analysed in this study.Attention should be paid to the objective of putting in place adequate plant pests and disease monitoring and surveillance systems while ensuring that excessive pesticide use by farmers is closely monitored.
The present study would be useful to the public sector ministries, departments and agencies, the private sector, industry, and academic research institutions in the design and implementation of sustainable food production and economically tractable policies.The study could be used by development partners, non-governmental organisations, civil society organisations, and international and inter-governmental organisations in their advocacy drive on sustainable food production.
While we recognize the importance of incorporating current FAW infestation levels in our analysis, we couldn't do so in the present study due to data limitations.We suggest that future studies should explore the effect of current FAW infestation levels on adoption of FAW management practices.Our empirical identification strategy could not control for unobserved heterogeneity.Therefore, future studies should explore this possibility.Furthermore, future studies could utilise cross-country or panel data to explore the theme of the present study; and could examine an insurance policy as a risk-mitigating strategy for farmers against FAW infestations.Such a policy would need to take into consideration the agro-ecological zones in which smallholder farmers are located.

Appendix
See Tables 6 and 7.

Fig. 1
Fig. 1 Map of study area.Source: Ghana Geological Survey

Fig. 3
Fig. 3 Test of covariate balancing on the three treatments categories.Source: Authors' own construct

Table 1
Sampled maize farming households in the study area . The treatment variable used in the MVTE model is categorised into FAW tion, as it is in the instrumental variable or the recursive models.The covariates specified in both the treatment and outcome equations could therefore be the same.

Table 2
Descriptive statistics of the variables used in the regression models Figures are mean and those in parentheses are standard deviations Exchange rate: 1 USD = GH¢5.43 in 2019 Source: Authors' computation a Natural logarithms of maize yield and household maize income were used in the regression models

Table 3
Marginal effects of the covariates on adoption of FAW management practices

Table 4
Estimates of multivalued treatment effects models

Table 6
Coefficient estimates from multinomial logit model

Table 7
Treatment effect estimates from IPWRA and PSM models Source: Authors' own computation Robust standard errors are in parentheses PSM denotes Propensity Score Matching Nearest neighbour was used as the matching algorithm for the PSM estimates ATE denotes average treatment effect, ATT denotes average treatment effect on the treated IPWRA denotes Inverse Probability Weighted Regression Adjustment