Enhancing food security through keyhole gardening in Lesotho

ABSTRACT

Keyhole gardens are gaining prominence worldwide due to their sustainable agricultural benefits. Additional evidence is required to inform keyhole gardening policy guidelines for combating food insecurity and poverty. This study employed an endogenous treatment effects estimator with ordered outcomes to assess keyhole gardening impact on food security. The study included data from 2,014 households in Lesotho. Our analysis shows that keyhole gardens have a 30.60% likelihood of achieving food security and reducing moderate and severe food insecurity by 11.02% and 41.62%, respectively. As a result, government and aid agencies should prioritize food-insecure households in keyhole garden initiatives.

KEYWORDS:

• Home food gardens
• ordered outcomes
• food insecurity experience scale
• Lesotho

1. Introduction

As the world’s population approaches 9 billion by 2050, food production must rapidly expand to satisfy rising demand, particularly in Sub-Saharan Africa, where poverty and food insecurity are pervasive. This increase in food production is required to meet people’s food security and nutrition needs, which is a major challenge for most economies, especially in African countries. This is especially important given that over two billion individuals do not have reliable access to a nutritious, balanced meal, with over 820 million of them going hungry (FAO, UNICEF, WFP and WHO 2019). Global food security programmes are increasingly focusing on household-level solutions. Establishing household food gardens is a potentially effective intervention. According to Galhena et al. (2013), traditionally, home gardening has been defined as cultivating a small piece of land that is either directly adjacent to or within a short distance from a family’s residence. A home garden, according to Eyzaguirre and Linares (2004), is a smaller scale alternative to the conventional agricultural system that is managed by household members and has a variety of plants and animals that resemble the natural eco-system. Existing studies (e.g. Castañeda-Navarrete 2021; Tesfamariam et al. 2018) show that the magnitude and composition of social-economic impacts of home gardening on the livelihoods of rural communities vary greatly. Home gardens have been linked to increased food consumption and a more diverse diet in many countries in Africa and the rest of the world. In South Africa, Tesfamariam et al. (2018) indicated that participation in homestead agroforestry system helps to reduce food insecurity by approximately 41.5%. In rural Guatemala, home gardens enhanced dietary diversity score, a measure of food security (Luna-González and Sørensen 2018). A study conducted by Baliki et al. (2019) in Bangladesh also identified a positive impact of integrated home gardens that combines training in homestead garden and nutrition education. A domestic food garden can offer a cost-effective supply of healthy food, improving food accessibility and availability, as well as generating revenue (Darby, Hinton, and Torre 2020; Rammohan, Pritchard, and Dibley 2019). This study explores the impact that keyhole gardens, which are not yet a prevalent type of home garden, have on food security in the context of Lesotho, which is characterised by its unique socioeconomic and agroecological environment.

Lesotho, a landlocked country completely enclosed by South Africa, encounters substantial difficulties concerning food security, hunger, and poverty. There are several reasons for this: high rates of unemployment, a decline in agricultural production, a shorter life expectancy, and higher rates of child mortality (Crush et al. 2010). WFP (2016) estimated that about 30% of Lesotho’s population are vulnerable to food insecurity. In 2018, about 11% of the Lesothos’ population suffered from severe food insecurity (IPC 2020). The country still faces food insecurity despite attempts to enhance agricultural productivity, resulting in a significant section of the population lacking access to nutritious as well as culturally suitable food (Muroyiwa and Ts’elisang 2021). The current issues highlight the necessity for specific programmes that address both immediate hunger and enhance resilience in local communities. Home food gardens provide a potential answer by allowing households to diversify their meals, decrease reliance on outside food sources, and improve nutritional self-reliance.

Previously, Lesotho’s government collaborated with non-governmental organisations to establish community gardens, but the expected benefits were not realised due to sustainability issues (Mashinini 2001). While operational community gardens exist in Lesotho, they have suffered owing to operational constraints, highlighting the need for a culturally relevant strategy. The keyhole garden, a sort of homestead garden, was then introduced, having been successfully implemented by NGOs in other countries to fight hunger and poverty. The term ‘keyhole garden (henceforth abbreviated as KHG)’ is constructed using inexpensive, locally accessible materials. It is usually a round raised garden supported with stones. Multiple layers of locally produced compost (compost made from manure, organic waste, wood ash, plant waste, yard sweepings, etc.) are used to cover it. The gardens are easier to maintain (great for the elderly, young children or sick), use less water, and don’t require expensive pesticides or fertilisers. The most popular crops grown include spinach, butternuts squash, carrots, pumpkin, potatoes, peas, beans, among others. KHG provides various benefits, including soil fertility for 5–7 years, year-round production, dietary diversity, and abundant production enough to feed an 8-person family (Nesbitt-Ahmed and Pozarny 2018).

Currently, households in Lesotho have embraced KHG through its implementation by NGOs, necessitating a critical analysis of whether the intended benefits, particularly improved food security through increased access and availability, are being realised. This study attempts to fill a gap in the literature by providing evidence of KHG’s successful contribution to the food security of Lesotho households practicing it. Unfortunately, there is limited evidence to support the effectiveness of KHG in improving food security for needy households.

This study is based on data obtained by the Child Grant Programme plus Sustainable Poverty Reduction via Income, Nutrition, and Access to Government Services (CGP-SPRING) project impact evaluation survey. Despite the fact that Pace et al. (2021) have already conducted an impact evaluation of the entire CGP-SPRING project, this work remains important for improving understanding and validity. The impact evaluation conducted by Pace et al. (2021) examines broader outcome variables that are associated with the following domains: consumption and poverty, dietary diversity and food security, agricultural inputs and assets, children’s well-being and financial inclusion. On the other hand, this study focuses on a particular aspect of the CGP-SPRING, which is KHG, and its impact on a particular outcome, which is food security. Concentrating on a particular project element enables more accurate attribution of observed results. Thus, isolating the impact of KHG component of the project allows for a more precise identification of the causal relationship between KHG and the observed changes in food security. Additionally, evaluating the impact of a specific element of a project helps pinpoint highly effective interventions within the overall project. Furthermore, the study’s empirical strategy used an endogenous treatment effect model with ordered outcomes that account for differences in observable and hidden household variables, as opposed to the propensity score matching technique applied by Pace et al. (2021), which only account for household observable characteristics. Thus, this study provides a more nuanced methodological diversity. The impact of KHG is better understood through methodological diversity, which ensures that results are not restricted to one study paradigm. This contextual sensitivity is necessary for tailoring treatments to local needs. The results of this study can also provide insights for making customised programme changes in the future. Thus, if this demonstrate greater impacts, programme implementers in the future can adjust techniques, redistribute resources, or increase efforts in the most effective areas to enhance the project’s overall impact.

2. Methodology

2.1. The study area, CGP-SPRING description and data collection procedure

2.1.1. The study area

With a surface area of 30,355 km2 (11,720 sq. miles) and an estimated population of 2,160,995, Lesotho is the southernmost landlocked country in the world. It is totally surrounded by South Africa. Lesotho is divided into 10 districts: Thaba-Tseka, Maseru, Leriba, Quithing, Mafeteng, Mokhotlong, Qacha’s Nek, Berea, Butha-Buthe, and Mohale’s Hoek. The climate in Lesotho is described as moderate, with irregular rainfall patterns that make it minimally suitable for agriculture. Rainfall ranges from 500 mm per year in the Senqu River valley to 1,200 mm per year in some areas along the eastern and northern boundaries.

2.1.2. About CGP-SPRING

Social protection is a primary focus in Lesotho’s national Strategic Development Plan 2012–2017 and the National Policy on Social Development established in 2014 by the Government of Lesotho. Ten social/assistance programmes are currently in place in Lesotho, with the Old Age Pension and the Child Grants Programme being the two largest (Pace et al. 2021). The CGP is a no-strings-attached cash distribution programme aimed at children from impoverished and at-risk households.

In July 2013, the FAO of Lesotho initiated a pilot project named ‘Linking Food Security to Social Protection Programme (LFSSP)’. The project aimed to enhance the food security of economically deprived and disadvantaged households by distributing vegetable seeds and offering homestead gardening guidance to qualified households under the CGP. To enhance the efficiency of both projects in increasing the food security of the beneficiaries, some households were selected to maximise the impact of both programmes. The FAO of the United Nations determined through an impact evaluation that the LFSSP had a positive impact on homestead gardening and agricultural productivity (Daidone et al. 2017). In 2015, the United Nations Children’s Fund, the Ministry of Social Development, and the Catholic Relief Services collaborated to establish a comprehensive livelihood project called Sustainable Poverty Reduction via Government Service Support (SPRINGS). This was done due to the favourable outcomes that were achieved. The SPRING programme focused on CGP beneficiaries but also allowed interested communities to participate.

The SPRINGS project focuses on addressing the gaps and priorities identified in the National Social Protection Strategy for 2012–2017, with a specific emphasis on reducing vulnerability through social protection. The goal is to improve the effectiveness and scope of social protection while helping at-risk individuals develop sustainable livelihood approaches (Daidone et al. 2017). In order to improve the cash transfer that is provided by the CGP and other social assistance programmes, SPRINGS is offering a community development package that consists of the following components:

1. Community-based savings and internal lending groups, along with financial education, which are commonly referred to as Savings and Internal Lending Communities (SILC).

2. The practice of homestead gardening, which includes keyhole gardening and the distribution of vegetable seeds.

3. Market clubs that provide nutrition education through the use of Community-led Complementary Feeding and Learning Sessions (CCFLS).

4. Integrated service center/Citizen Engagement Events.

Once again, it is essential to emphasise that the scope of this study is limited to evaluating the impacts of KHG, which is a component of the CGP-SPRING initiative, on the food security of households. Within the scope of this investigation, we categorised the individuals who benefited from CGP-SPRING and participated in KHG as adopters, whereas those who did not have the opportunity to engage in KHG were categorised as non-adopters.

2.1.3. Data collection

As mentioned previously, the data that was used for this study was part of the data collected by the CGP-SPRING project impact evaluation that was conducted in Lesotho. Spatial intelligence (SiQ) was used for the data collection that took place between the months of November 2017 and the middle of January 2018. In accordance with the goals and objectives of the project, six districts were chosen from all around the country. These districts are Maseru, Mafeteng, Leribe, Berea, Butha-Buthe, and Mohale’s Hoek. Out of the total 2,014 households that were surveyed by the SiQ, 1550 of them were qualified to receive the CGP, while 464 were not qualified. The geographic distribution of the households that were sampled is summarised in Table 1, which is organised according to district, treatment status, and eligibility parameters.

Table 1. Survey sample by districts, treatment status and eligibility.

	eligible				ineligible
Districts	Comparison	CGP	CGP+SPRING	Total	Comparison	CGP	CGP+SPRING	Total
Maseru	272	22	164	458	40	13	59	112
Butha-Buthe	1	66	123	190	0	34	60	94
Leribe	81	61	154	296	16	18	62	96
Berea	67	230	0	297	10	48	0	58
Mafeteng	130	80	0	210	20	61	0	81
Mohale’s Hoek	99	0	0	99	23	0	0	23
Total	650	459	441	1550	109	174	181	464

Source: CGP-SPRING impact evaluation survey by Parce et al. (2021).

2.2. Conceptual framework

The assessment of the impact of KHG on household food security originates from the CGP-SPRING theory of change that disentangles the different pathways along which the interventions could tackle food insecurity, while promoting broader developmental impacts. Conservation agricultural practices, such as KHG, have been shown to have an impact on food security and improved lives as a result of income that they provide, as demonstrated by the conceptual framework that is illustrated in Figure 1. KHG support intervention provided by SPRINGS, can improve the diversity of food produced, consequently contribute to a better diet. Beneficiary households are expected to improve nutrition and dietary diversity, by producing diverse vegetables and adopting better feeding practices. This will consequently improve the food security status of the beneficiaries.

Figure 1. Conceptual framework for determinants of KHGs’ adoption, food security and causal effect of KHG on food security.

Although the project selected the districts and communities in line with the project goals and objectives, individual households were not coerced into participation in the project. Thus, participation of households was voluntary. In this case, individuals’ to participate may be influenced by demographic, institutional, and household assets, as illustrated in Figure 1. The KHG adoption defines any household in our data sample that cultivated any vegetable or range of vegetables using the keyhole gardening system and receives seeds from the project.

It was hypothesised that the socioeconomic, institutional, and asset endowments of a household, in addition to the adoption of KHG, would have an effect on the food security status of the household. The adopters (beneficiaries) of KHG are expected to have higher food security status than non-adopters (non-beneficiaries) after correcting for self-selection biases in observed and unobserved characteristics (Faber, Witten, and Drimie 2011; Galhena, Freed, and Maredia 2013; Nkosi et al. 2014). Gender, marital status, educational level, the number of dependents in the household, and engagement of non-farm economic activities is among the demographic parameters that have been postulated to influence both the adoption of KHG and the household’s food security status in line with many studies (e.g. Muroyiwa and Ts’elisang 2021; Nkomoki, Bavorová, and Banoute 2019; Tesfamariam et al. 2018). For example, we expect that being a women, older adults and more educated respondents to have a positive influence on adoption of KHG and food security levels (Du Toit et al. 2022; Faber, Wenhold, and Laurie 2017). Similarly, the study expects engagement in non-farm economic activities to have a positive influence on both adoption of KHG and food security, while the number of dependents is expected to influence KHG adoption positively but have negative correlation with food security. Households’ assets that were also thought to explain the variation in farmers’ KHG adoption decision and food security status in a positive direction are farmlands, number of livestock, ownership of mobile phone and television or radio (Etana and Tolossa 2017; Ibekun and Adebayo 2020). Social assets may also include farmers’ awareness of home food garden programmes and farmers’ attendance to KHG demonstrations. For institutional factors, we included variables such as households been connected to the national grid (Baliki et al. 2019; Castañeda-Navarrete 2021). The study hypothesised that being connected to the national grid may have a positive influence on food security and KHG adoption.

2.2.2 Measurement of food security status

Food security, or the ability of households and people to obtain food, is a crucial welfare characteristic that poses serious challenges when trying to quantify. Food security is defined as the ability of everyone to have access to enough food to sustain an active and healthy lifestyle (Cafiero et al. 2018). Access to sufficient food remains an essential component of food security despite its multifaceted nature. Food security, because of its multidimensional nature, allows for a wide range of measurement. Food Insecurity Experience scale (FIES), household food insecurity and access scale (HFIAS), food consumption score (FCS), household dietary diversity scale (HDDS), household hunger scale (HHS), and self-assessed food security measure (SAFS), among others, are some of the indicators used for food security proxies. The HFIAS, FIES, and HHS measures are founded on the idea of experiencing food deprivation in the household. The SAFS food security measure is a purely subjective indication based on how the household perceives its own level of food security. Dietary diversity and frequency are measured by the FCS and the HDDS, respectively. A robust food security index, according to Maxwell et al. (2013), is one that captures various aspects of food security and is accurate, dependable and comparable over time and place, albeit, none of the above-mentioned metrics captures all the concepts of food security (Maxwell and Coates 2012). As a result, using a reliable way of gauging food security in households is a vital step in dealing with it successfully.

This study used Food Insecurity Experience Scale (FIES), which is an experiential measure of access to food. FIES assesses the severity of the state of one’s inability to obtain the food necessary to lead a healthy, active, and dignified life. The Food Insecurity Experience Scale Survey Module (FIES-SM; see Appendix I), which is based on responses to an 8-item questionnaire, measures conditions and behaviours arising from a person’s inability to get food owing to a lack of economic or other resources. According to the FIES-SM questions, the target population can be classified as either food secure (to mildly food insecure), moderately food insecure, or severely food insecure based on the ‘Yes/No’ answers. Positive (Yes) response totals eight, whereas negative (No) response totals zero. Households with a score of 0 have no problems with food insecurity. Three levels of food (in) security severity have been established by FIES: mildly food insecurity, moderately food insecurity and severe food insecurity. However, as noted by FAO et al. (2018), those in the mildly food insecurity category are considered as food secure. Hence, two levels of the extent of food insecurity are measured by FIES indicator. A score of 0–3, 4–6, and 7–8 denotes food security, moderate food insecurity, and severe food insecurity, respectively. The FIES is a supplement to other food security metrics such as household dietary diversity score, food consumption score, cost of calories, per capita food expenditure, and self-assessment measures (FAO et al. 2018). The FIES has been confirmed as a reliable tool for assessing food insecurity across different linguistic, cultural, and developmental contexts (FAO et al. 2018). Hence, selected as the best food security indicator with regards to the focus of the study compared with other food security indicators.

In order to estimate the causal effect of KHG adoption on the extent of food security status of the households in Lesotho, we applied endogenous ordered probit (Instrumental Variable ordered probit, IV – ordered probit).

2.2 Empirical strategy

The primary goal of this research is to determine the extent to which the practice of KHG affects food security in Lesotho. When assessing the influence of KHG, one methodological issue to address is potential endogeneity of the treatment variable KHG. If there is endogeneity, KHG estimations may be skewed and fail to capture the underlying impact of KHG participation on food security. Endogeneity can be caused by reverse causality (simultaneity bias), omitted variables (bias from omitted variables) and measurement mistakes (Wooldridge 2010). Given the importance of both food security and KHG, it is entirely plausible that both will be dependent on one another. Furthermore, unobserved variables such as managerial abilities, which can affect both adoption of KHG and food security at the same time, may result in omitted variable bias. This study used self-reported methods to elicit information about households’ food security status, which could also lead to measurement errors. Missing data is another serious issue in impact evaluation. Due to the fact that each household is either KHG adopter or a non-adopter and hence cannot be both. Hence, it is difficult to assess the influence on the same household at the same time. This means that the outcome variable (status of food security for target households) cannot be observed if KHG adoption is not observed at the same time. There is no way to observe the counterfactual for outcome variables because they can only be evaluated in one of the two states: adoption or non-adoption.

We used a two-equation system, specifically an endogenous ordered probit regression technique (Instrumental Variable (IV) – ordered probit model), to evaluate the connection between food security and the treatment variable, KHG. The IV – ordered probit model was used to address endogeneity concerns, and likelihood simulation methods were applied to estimate the parameters. Several studies assessing the effects of an intervention on household food security have divided food security into two categories: food security households and food insecurity households, requiring the use of either a binary probit or logit model. The binary food security indicator is limited as it does not account for households with food security indices ranging from extremely low to very high. The study categorises food security status into three levels: food secure, moderately food insecure and severe food insecurity. These levels are used to rank households based on their food security indices. An ordered probit model is more suitable for this analysis due to the discrete and ordinal nature of the food security measures used.

Food (in)security status, which is ranked from food security to extreme food insecurity, can be used as an ordered discrete variable to explain the treatment assignment outcome. This can be expressed as:

(1) $Y_i=\left\{\begin{array}{c}1\mathit{if}-\mathrm{\infty }<Y_i^{\ast }<{\eta }_1\\ 2\mathit{if}{\eta }_1\le Y_i^{\ast }<{\eta }_2\\ ...\\ \mathit{j}\mathit{if}{\eta }_j\le Y_i^{\ast }<\mathrm{\infty }\end{array}\right\}$(1)

where ${\mu }_1,{\mu }_2\dots .,{\mu }_J$identify the threshold parameters, $\mathit{j}=1,2,\dots .\mathit{J}$denote food (in)security categories and $Y_i^{\ast }$ represent the latent outcome (food security status) for the i^th adopter expressed as:

(2) $Y_i^{\ast }=X_i{\alpha }_1+T_i{\alpha }_2+{\epsilon }_i$(2)

In equation (2)(2) $Y_i^{\ast }=X_i{\alpha }_1+T_i{\alpha }_2+{\epsilon }_i$(2) above, $T_i$ is the treatment variable representing the KHG adoption behaviour of the households. $X_i$and ${\epsilon }_i$ denote a vector of exogenous variables and the error term, respectively. The KHG adoption behaviour of households can be specified as:

(3) $T_i=1\mathit{if}{X}_i{\beta }_1+z_1\mathit{i}{\beta }_2+{\mu }_i\text{<}00\mathit{if}{X}_i{\beta }_1+z_1\mathit{i}{\beta }_2+{\mu }_i\le 0$(3)

where $T_i$ represent adoption of KHG, assigning the value 1 for KHG and 0 for non-adoption of KHG. $Z_i$ is an instrumental variable that was used to identify the adoption equation (3)(3) $T_i=1\mathit{if}{X}_i{\beta }_1+z_1\mathit{i}{\beta }_2+{\mu }_i\text{<}00\mathit{if}{X}_i{\beta }_1+z_1\mathit{i}{\beta }_2+{\mu }_i\le 0$(3) . This instrument was used to improve the identification of equation (3)(3) $T_i=1\mathit{if}{X}_i{\beta }_1+z_1\mathit{i}{\beta }_2+{\mu }_i\text{<}00\mathit{if}{X}_i{\beta }_1+z_1\mathit{i}{\beta }_2+{\mu }_i\le 0$(3) . The principle of identification requires that the instrumental variable should influence the treatment variable $T_i$ but redundant in explaining the changes in the outcome variable ($Y_i$) (Danso-Abbeam et al. 2021; Di Falco and Bulte 2013; Khonje et al. 2018). In this study, we chose farmers’ awareness of the CGP-SPRINGS programme on the basis of intuition. A household who is aware of the programme is more likely to join, resulting in the formation of a sub-population between adopters and non-adopters. One useful measure of farmers’ predisposition towards participation in the KHG programme is their awareness of the programme. Nevertheless, awareness does not directly address the changes in food security. Instead, it is the activities, such as enhancing dietary diversity and increasing crop production, that happened as a result of participation in KHG, which ultimately lead to improvements in food security outcomes. As a result, we anticipated that being aware of the CGP-SPRINGS programme would be the most important way to identify the participation equation. The validity of the instrument was further proven by applying a simple falsification test, as proposed by Di Falco et al. (2011) and used in Khonje et al. (2018) and Danso-Abbeam et al. (2022a), among others, to determine its reliability. The selection instrument was shown to be valid since it influences KHG adoption (equation 2(2) $Y_i^{\ast }=X_i{\alpha }_1+T_i{\alpha }_2+{\epsilon }_i$(2) ) but not food (in) security status (equation 1(1) $Y_i=\left\{\begin{array}{c}1\mathit{if}-\mathrm{\infty }<Y_i^{\ast }<{\eta }_1\\ 2\mathit{if}{\eta }_1\le Y_i^{\ast }<{\eta }_2\\ ...\\ \mathit{j}\mathit{if}{\eta }_j\le Y_i^{\ast }<\mathrm{\infty }\end{array}\right\}$(1) ).¹

2.3. Measuring the average treatment effects

The units of measurement of the response variable fundamentally affect the estimation of the treatment. Continuous outcome variables have been the focus of many empirical studies in impact assessment, with an emphasis on the average treatment and the treatment effect on those who were treated. Taking averages of the treatment impact at the individual level is misleading if outcomes are measured on an ordinal scale, because the difference between outcomes is not defined.

The standard probability estimate of treatment effect is the average treatment effects (ATE), which can be defined as the probability difference between the observation of an outcome with and without treatment. In our case, the potential difference in food security status between KHG adopters and non-adopters. Formally:

(4) $\Delta \mathit{yATE}\equiv \mathit{P}(Y_1=\mathit{y})-\mathit{P}(Y_0=\mathit{y})\mathit{y}=1,\dots .\mathit{J}$(4)

where there are $\mathit{J}$ effects, one for each responses (food security, moderate food insecurity and severe food insecurity). $\mathrm{\Delta y}\mathit{ATE}$is expected to be negative for low $\mathit{y}$ and positive for high $\mathit{y}$ if the treatment has a positive influence on responses, applying the concept that higher outcomes of $\mathit{Y}$are in some manner ‘better’. Probability effects allow for a more extensive estimation of the treatment’s effects in all areas of the result distribution, even if there are no obvious systematic differences in characteristics showing whether the treatment has a substantial impact in practice (Boes 2013).

In a similar vein, the treatment on those households that received it can be characterised as the average effect of treatment on the treated (ATT) parameter. Hence, ATT represents the variation between the treated group (KHG adopters) and the control group (non-adopters) in the response variable (food security outcomes).

The ATT can be specified as;

(5) $\Delta \mathit{yATT}\equiv \mathit{P}(Y_1=\mathit{y}|\mathit{T}=1)-\mathit{P}(Y_0=\mathit{y}|\mathit{T}=1),\mathit{y}=1,\dots .,\mathit{J}$(5)

In the estimation processes, the ‘same scale’ assumption is necessary for both treatment parameters to be robust against the specific values allocated to outcomes. However, this assumption is not unduly restrictive, otherwise it would be impossible to compare the $Y_1$ and $Y_0$ distributions.²

3. Results and discussions

This section presents the results of the study. It begins with the profile of the households who are engaged in KHG (treated group) and those who are not (control group). The second section (Section 3.2) discusses the factors that influence households’ decisions to adopt the KHG technology, while the third section discusses the drivers of household food security/insecurity using the standard ordered probit model. Finally, we present the results of the IV ordered probit model that was applied to estimate the impact of the KHG on food security status in Section 3.4.

3.1. Socioeconomic background and variable description

Table 2 provides an overview of how households’ socioeconomic status varies across the study areas of Lesotho. As shown in the table, we perform statistical significant tests for equality of means between KHG adopters and non-adopters. In this section, we first explain the households’ food security conditions, followed by the independent variables that describe the characteristics of both KHG adopters and non-adopters.

Table 2. Demographic characteristics of the sampled population.

Variables	Description	Full sample	Adopters	Non-adopters	difference
Outcome variables
Food secure	Food secure =1, otherwise 0	0.278	0.280	0.274	0.007
Moderate food insecure	Moderate food insecure = 1, otherwise 0	0.263	0.289	0.192	0.098***
Severe food insecure	Severe food insecure = 1, otherwise 0	0.458	0.429	0.535	0.105***
Independent variables
Gender of household head	Male = 1, otherwise 0	0.488	0.478	0.515	0.036
Marital Status	Married = 1, otherwise 0	0.272	0.27	0.275	0.005
Educational Status	Attain formal education = 1, otherwise 0	0.876	0.845	0.888	0.043
Farm Size	Farm size in hectares (ha)	0.343	0.388	0.224	0.164
Dependency ratio	Proportion of households members dependent of household head	0.405	0.401	0.414	0.013
Number of livestock	Count	2.927	3.215	2.159	1.057
Engagement in non-farm activities	Yes = 1, otherwise 0	0.064	0.076	0.033	0.044
Owned mobile phone	Yes = 1, otherwise 0	0.825	0.861	0.733	0.129
Amount of remittances	Money received from relatives in abroad in Maloti	1554.949	1713.958	1129.569	584.389***
Connected to national grid	Yes = 1, otherwise 0	0.261	0.271	0.232	0.039
Aware of home garden program	Yes =1, otherwise 0	0.367	0.471	0.090	0.379
Attendance of demonstration farms	Yes = 1, otherwise 0	0.164	0.214	0.033	0.181
Maseru	Resides in Maseru = 1, otherwise 0	0.283
Butha -Buthe	Resides in Butha-Buthe =1, otherwise 0	0.141
Leribe	Resides in Leribe = 1, otherwise 0	0.195
Berea	Resides in Berea = 1, otherwise 0	0.176
Mafeteng	Resides in Mafeteng =1, otherwise 0	0.144
Mohale’s Hoek	Resides in Mohale’s Hoek = 1 otherwise 0	0.061
Number of observations	Respondents that participated in the survey	2,014	1466	548

***, ** and * indicate significant levels at 1, 5 and 10% levels.

According to the FIES statistics, approximately 27.8% of Lesotho households are food secure, and there is no statistically significant difference between KHG adopters (28.0%) and non-adopters (27.4%). This indicates that around 27.8% of the overall studied population did not encounter any of the food insecurity scenarios or encountered insufficient or unpleasant meals on rare occasions. Thus, these households did not limit their food consumption, experienced food scarcity, or went an entire day without food. For some other households (classified as moderately food insecure), there was a point when they consumed inferior and unappealing food and reduced the amount of food consumed by household members. Within the severe food insecure household category, KHG adopters (28.9%) differ considerably from non-adopters (19.2%). For households classed as extremely food insecure (45.8%), non-adopters (53.5%) outnumbered adopters by a wide margin (42.9%). These extremely food insecure households either face food shortages, go to bed hungry or go an entire day and night without food.

The table further revealed that a significant difference exists between male-headed households that practice KHG and those that do not. That is, while 51.5% of the KHG-adopters are male-headed households, 47.8% of the non-adopters are male-headed households. Surprisingly, less than 30% of our sampled households were headed by married individuals. For both KHG-adopters and non-adopters, majority (84.5% and 88.8%, respectively) of household heads attain formal education. With regard to landholdings, the average farm size was 0.343 ha. However, the farm size of KHG-adopters was significantly larger than those households with no KHG. A lot of home gardens are built on land that is too small, too remote, or otherwise unsuitable for field crop or forage cultivation (Galhena, Freed, and Maredia 2013). Depending on the household, the size of a KHG varies, but on average, they are smaller than the parcels of arable land that belong to the family. As measured by the number of household members under 16 and those over 65 years old, the dependency ratio is 0.405, 0.401 and 0.414 for the full sample, KHG-adopters and non-adopters, respectively. That is about 40% and 41% of the household members of KHG-adopters and non-adopters, respectively, are considered as dependents (children and the aged).

Household assets such as livestock have always played a vital role in the lives of rural households. On average, households owned about three livestock ranging from poultry, small ruminants (sheep and goats) to cattle. Non-farm income activities have been estimated to have a significant influence on KHG, and consequently food security status of households, particularly in rural areas. Surprisingly, few households are engaged in non-farm income generating activities such as petty trading, formal salaried employee, vocational skill training (masonry, carpentry, etc.), among others. In addition, about 83% of the households owned mobile phone. The amount of money received from relatives and friends in abroad by the KHG-adopters was significantly higher than the money received by non-adopters of KHG. About 26% of the sampled households are connected to the national grid, which provide them with electricity. More households (47%) who participated in KHG had ever been aware of CGP-SPRING programme in the country than the non-adopters of KHG. Likewise, more KHG-adopters had ever attended farm demonstration programmes compared with the non-adopters.

3.2. Determinants of KHG adoption

The results in Table 3 present the estimated results of the factors that drives households’ adoption of KHG. Variables such as educational attainment, land holdings, number of livestock owned, participation in non-farm business activities, ownership of mobile phones, and households’ awareness of home food gardening programmes were important factors that influence the adoption of home food gardening. For limited dependent models like probit, coefficient estimates cannot be directly interpreted in magnitude. Instead, discussions focus on the marginal effects, which indicate the magnitude of change in the adoption of KHG attributed to changes in the explanatory variables in the model.

Table 3. Determinants of keyhole garden’s adoption in Lesotho.

Variables	Coefficient	Std. Err	Marginal effects
Gender	−0.0992	0.0640	−0.0308(0.0198)
Marital status	0.0033	0.0238	0.0010(0.0074)
Educational attainment	0.1989**	0.0956	0.0648(0.0326)**
Farm size	0.2488***	0.0548	0.0772(0.0169)***
Proportion of active labour	−0.0300	0.0852	−0.0093(0.0265)
Number of livestock owned	0.0125**	0.0061	0.0039(0.0019)**
Engagement of non-farm business	0.4820***	0.1476	0.1258(0.0311)***
Owned mobile phone	0.3717***	0.0803	0.1243(0.0286)***
Remittances	0.0004	0.0001	0.0001(0.0000)
Connected to national grid	−0.0737	0.0771	−0.0226(0.0233)
Awareness of CGP-SPRING project	1.0667***	0.1418	0.2419(0.0204)***
Attendance of demonstration field	−0.2017	0.2668	−0.0664(0.0928)
Maseru	0.2226	0.1418	0.0666(0.0408)
Butha-Buthe	0.4614***	0.1575	0.1242(0.0365)***
Leribe	0.1846	0.1498	0.0549(0.0426)
Berea	0.0353	0.1442	0.0109(0.0439)
Mafeteng	−0.2181	0.1468	−0.0712(0.0502)
Constant	−0.2686	0.1949

***, ** and * denote significant levels at 1, 5 and 10%, respectively.

The results indicate that level of the education of the household has a positive and significant effect on households’ probability to practice KHG. The marginal effect of education shows that an additional year of education increases households’ probability to adopt KHG by approximately 6.5%.

Households with large hectares of land parcel have a higher propensity (7.7%) to adopt KHG as indicated by the positive and significant marginal effect. In a similar study, Bahta and Owusu-Sekyere (2019) indicated that households with larger area of land are more likely to participate in a KHG intervention programme.

The quantity of livestock (e.g. poultry, pigs, goats, etc.) owned is positive and significant, suggesting that an additional livestock added to the stocks of animals owned by a household increases its probability of adopting KHG by 0.4%. We also found that households’ engagement in non-farm economic activities increases their likelihood of adopting KHG by 12.5%, all things being equal. Household assets have been predicted by many studies (e.g. Bahta and Owusu-Sekyere 2019; Danso-Abbeam et al. 2021, etc.) to encourage adoption and participation in agricultural intervention programmes. This study, likewise, has a household asset (mobile phone) that has been estimated to have a positive and significant influence on the probability of adopting KHG as indicated by the marginal effect.

In a similar vein, households that were aware of the CGP-SPRING intervention programme had a higher likelihood of adopting KHG compared to those that were not. Rural households in Lesotho can access critical information about the intervention project, including its objectives, benefits, eligibility criteria, and participation processes. Information diffusion clarifies doubts and misunderstandings, allowing households to make informed decisions about their participation. In addition, being aware of an intervention project beforehand increases the perceived importance and worth of the intervention programme among rural households. When households fully understand how the initiative precisely addresses their needs or concerns, they are more likely to recognise its value and become more motivated to join. Martey et al. (2020) found that farmers’ existing knowledge of an intervention programme increases the probability of their involvement in the programme, but it has no influence on their economic outcomes. Finally, residents of Burtha-Burthe are about 12.4% more likely to adopt KHG compared to those located in Mohale-Hoek.

3.3. Determinants of household food security status

Table 4 reports the findings of the standard ordered probit model used to explain the factors influencing food security status measured by FIES. The Chi² test result indicates that the values of the coefficients of the independent variables are not collectively equal to zero (Chi²  = 182.12; p = 0.000). The predicted cut-off points 1 and 2 represent the minimum and maximum values of the unobserved variable that occur when the households’ likelihood of being classified into each food security/insecurity category changes. The cut points shed light on the boundaries at which households are more inclined to transition between categories in response to changes in the explanatory variables. In order to shed light on how variations in independent factors affect the likelihood of various food security outcomes, we predicted marginal effects. An increase in the coefficient of a variable increases the likelihood of falling into a specific category. Similarly, a negative coefficient for a category reduces the likelihood of falling into that category. All significant variables have a corresponding significant marginal effect.

Table 4. Ordered probit model results of the determinants of household food security.

Variables	Food security	Moderate food insecurity	Severe food insecurity
	(Y = 1)	(Y=2)	(Y=3)
Gender	0.0009(0.0172)	0.0002(0.0035)	−0.0011(0.0207)
Marital status	−0.0235(0.0064)^a	−0.0048(0.0014)^a	0.0283 (0.0077)^a
Educational attainment	0.0276(0.0266)	0.0056(0.0055)	− 0.0333(0.0320)
Farm Size	0.0927(0.0132)^a	0.0189(0.0033)^a	−0.1117(0.0158)^a
Dependency ratio	−0.0564(0.0229)^b	−0.0115(0.0048)^b	0.0680(0.0278)^b
Number of livestock owned	0.0023(0.0009)^b	0.0005(0.0002)^b	−0.0027(0.0011)^b
Non-farm business activity	0.1238(0.0349)^a	0.0253(0.0076)^a	−0.1491(0.0420)^a
Ownership of mobile phone	0.1222(0.0237)^a	0.0249(0.0056)^a	−0.1471(0.0286)^a
Remittances	0.0000 (0.0000)^a	0.0000(0000)^a	−0.0000(0.0000)^a
Connected to national grid	0.0543(0.0206)	0.0111(0.0045)	−0.0653(0.0247)
Farm demonstration attendance	0.0104(0.0455)	0.0021(0.0093)	−0.0125(0.0548)
Maseru	−0.0720(0.0386)^c	−0.0147(0.0081)^c	0.0867(0.0465)^c
Butha Buthe	−0.1651(0.0415)^a	−0.0337(0.0092)^a	0.1988(0.0499)^a
Leribe	−0.0163(0.0396)	−0.0033(0.0081)	0.0196(0.0478)
Berea	−0.1196(0.0397)^a	−0.0244(0.0085)^a	0.1440(0.0478)^a
Mafeteng	−0.1086(0.0409)^a	−0.0222(0.0087)^b	0.1309(0.0493)^a
LR Chi^2	182.12
Cut 1	−0.6397 (0.1623)
Cut 2	0.1619 (0.0952)

a, b and c denote significant at 1, 5, and 10 level of significance. Standard errors are in parentheses.

The results indicate that households led by married individuals were less likely to have food security and moderate food insecurity, but more likely to experience severe food insecurity. The findings suggest that households with married couples may have a larger number of dependants, which could lead to negative effects if the increase in size is not accompanied by a corresponding increase in the number of individuals contributing to the household’s welfare. Thus, large household sizes may exert pressure on household resources and limit their ability to access food. Farm size increases the probability of households being food secure and moderately food insecure, while decreases the probability of households experiencing severe food insecurity. That is, a one-hectare increase in farm size results in 9.3% probability of households being food secure, 1.9% of being in the category of moderately food insecure and 11.2% probability of getting themselves out of the severe food insecurity category. This could be because households with larger land sizes may cultivate a broader variety of crops, resulting in more diverse and nutritious foods, as opposed to households with smaller land sizes, which may focus solely on staple crops. Nkomoki et al. (2019) observed landholdings by household as one of the key determinants of food security status in Zambia. The dependency ratio reduces the chances of household being in the category of food security and moderate food insecurity by 5.6% and 1.2%, respectively. However, it has likelihood of increasing households’ experience of severe food insecurity by about 11.2%.

Furthermore, the number of livestock owned by households increases their chances of experiencing food security and moderately food insecurity by just 0.2% and 0.05%, respectively, while decreasing the likelihood of being severely food insecure by 0.3%. Raising livestock increases the amount of high-quality meat, milk, and other dairy products available to households. A second benefit of livestock ownership is that the sales from livestock generate income for the household, thereby increasing their food purchasing power. Similar to our results, Jodlowski et al. (2016), who evaluated the effect of livestock on food security in Zambia, found that livestock ownership increased food dietary diversity. Likewise, Kafle et al (2016) investigated the role of cattle transfer programmes in Zambian households with low levels of economic security. They found that providing training for households in livestock management issues improved their financial capability and food security status.

Another important variable that has been reported by many studies (e.g. Danso-Abbeam et al. 2021; Nkomoki, Bavorová, and Banoute 2019; Zereyesus et al. 2017) to influence household food security status is non-farm activities. Our findings show that households participating in other non-farm economic activities are more likely to be food secure and moderately food insecure by 12.4% and 2.5%, respectively, and have 14.9% chances of reducing severe food insecurity. Non-farm economic activities have the potential to greatly enhance the purchasing power of households, enabling them to acquire a wider range and larger amount of food. Higher disposable income enables households to afford healthy meals and expand the variety of their diets, hence decreasing the likelihood of experiencing food insecurity. In addition, diversifying income sources beyond agriculture and expanding economic prospects for households can enhance food security prospects for those who are vulnerable, especially in rural areas.

Another important variable that has been reported by many studies (e.g. Danso-Abbeam et al. 2021; Nkomoki, Bavorová, and Banoute 2019; Zereyesus et al. 2017) to influence household food security status is non-farm activities. Our findings show that households participating in other non-farm economic activities are more likely to be food secure and moderately food insecure by 12.4% and 2.5%, respectively, and have 14.9% chances of reducing severe food insecurity. Non-farm economic activities have the potential to greatly enhance the purchasing power of households, enabling them to acquire a wider range and larger amount of food. Higher disposable income enables households to afford healthy meals and expand the variety of their diets, hence decreasing the likelihood of experiencing food insecurity. In addition, diversifying income sources beyond agriculture and expanding economic prospects for households can enhance food security prospects for those who are vulnerable, especially in rural areas.

Table 4 results further show that owning household assets such as mobile phones increases the likelihood of households falling into the food secure and moderately food insecure categories. The mobile phone variable, on the other hand, reduces the household’s risk of experiencing severe food insecurity. On a regular basis, mobile phones (which now have in-built radio) provide households with up-to-date information, such as programmes about farming methods and technology. Owners of mobile phones will have improved access to information on how to increase farm output and diversify their livelihoods, potentially reducing food insecurity and poverty in general. Danso-Abbeam et al. (2022b) hypothesised that households with better access to information through information devices are more likely to be food secure. Finally, except for Leribe, households in all other districts surveyed are less likely to be food secure, moderately food insecure, and more likely to fall into the category of severe food insecurity than households in Mohale’s Hoek district (the reference group).

3.4. Impact of home gardens on household food security

The adoption of KHG differed significantly from non-adoption in two categories of food insecurity (moderate food insecurity and severe food insecurity) but no significant difference in food security status between adopters and non-adopters of KHG as shown in the descriptive statistics in Table 1. This disparity, however, ignores selectivity biases caused by observed and unobserved heterogeneities. It was for this reason that the ATT (the effects of KHG on only the treated) on food security was estimated using the ordered probit regression with endogenous treatment. Table 5 presents the full estimation results. In Table 5, the Wald Chi² value of 74.14 is significantly different from zero, implying that the model fits the data well. Moreover, the estimated correlation between the error term of KHG and food security levels is 0.6520 and it is significantly different from zero. Hence, the treatment variable, KHG is endogenous. The positive value suggests that some unobserved factors that affect KHG adoption also tend to increase the chance of farming households being in the food security category.

Table 5. Determinants of adoption of KHG and food security based on KHG adoption status.

Variables	KHG Adoption		Food security = adopters		Food Security = non-adopters
	Coef.	Std. Err	Coef	Std. Err	Coef	Std. Err
Gender	−0.0992	0.0640	0.0397	0.0952	−0.2331	0.2179
Marital status	0.0033	0.0238	0.0364	0.0364	0.1267***	0.0860
Educational status	0.1989**	0.0956	−0.1452	−0.1357	−0.2602	0.2800
Farm Size	0.2488***	0.0548	0.1194**	0.0582	0.1982	0.1399
Active labour	−0.0300	0.0852	0.0348	0.1296	0.5549*	0.3141
Number of livestock owned	0.0125**	0.0061	−0.0011*	0.0038	−0.0219	0.0129
Non-farm business	−0.4820***	0.1476	−0.4123**	0.1785	−0.4222	0.5412
Owned phone	0.3717***	0.0803	−0.4966***	0.1428	−0.4287	0.2425
Remittances	0.0004	0.0001	−0.0000	0.0000	−0.0000	0.0000
Electricity access	−0.0737	0.0771	−0.2256*	0.1253	−0.0357	0.2918
Aware of home food garden programme	1.0667***	0.1418
Demonstration fields attendance	−0.2017	0.2668	0.4038	0.2151	0.3345	0.4273
Maseru	0.2226	0.1418	−0.1880	0.2019	−0.2825	0.3301
Butha_Buthe	0.4614***	0.1575	0.1745	0.2362	0.7268*	0.4589
Leribe	0.1846	0.1498	−0.4433**	0.2118	0.2852	0.3654
Berea	0.0353	0.1442	0.0415	0.2030	0.5085	0.3438
Mafeteng	−0.2181	0.1468	−0.0420	0.2322	0.4964	0.4471
Constant	−0.2686	0.1949
Cut_11			1.1721 0.3131
Cut_01					−0.4108	0.6129
Cut_12			0.5024 0.3323
Cut_02					−1.0454	0.5709
Corr	0.6520	0.2570
Wald Chi^2 = 74.14**
Log Likelihood = −3078 Prob > Chi^2 = 0.000

***, ** and * denote significant levels at 1, 5 and 10%, respectively.

The second and the third column of Table 5 are the coefficient and standard errors of the determinants of KHG adoption. The fourth and the fifth cloumns explain the drivers of food security of the KHG adopters, while columns six and seven explain the factors influencing household food security for the non-adopters of KHG. Because the study has already discussed the determinants of KHG adoption and determinants of household food security (using the standard ordered probit model) in sections 3.2 and section 3.3, respectively, there will be no further discussion of Table 5.

The results of the estimated ATT are presented in Table 6. For our sub-treated population, the findings revealed that households that practiced KHG reaped substantial benefits from doing so. When it comes to KHG specific impact on food security, the ATT shows that it has the potential to raise household food security by 30.60% while reducing moderate and severe food insecurity by 11.02% and 41.621%, respectively.

Table 6. Average treatment effects of adoption of KHG.

Food (in)security status	Average treatment effects (ATT)
Food security	0.3060 (0.0602)***
Moderate food insecurity	−0.1102 (0.0047)***
Severe food insecurity	−0.4162 (0.0726)***

***, ** and * denote significant levels at 1, 5, and 10%, respectively.

Since KHG adoption was associated with enhanced food security, those households who adopted it had 30.60% higher probability of being food secured than if they had not adopted it. Similarly, for adopters in the moderate and severe food insecurity group, adopting KHG resulted in a 11.02% and 41.62% reduction in their moderate and severe food insecurity status, respectively, compared to their status if they had not adopted KHG.

The question of what impact adoption of KHG has on the food security status of adopters has been addressed, and the findings are crucial for policymakers to consider. This finding implies that KHG is an important initiative to reduce food insecurity in rural households. In light of climate change and rising demand for nutrient-dense foods, KHG will remain a critical tool in the fight against global food poverty. Food insecurity in Lesotho is usually linked to malnutrition due to protein-energy deficiency and micronutrient deficiencies (Muroyiwa and Ts’elisang 2021). As a result, the concept of KHG which is already been practiced in some rural communities in Lesotho spearheaded by non-governmental organisations (NGOs) needs to be strengthened. The results of the study add to many pieces of evidence that KHG is a livelihood strategy that can help less developed economies such as Lesotho to fight food insecurity, malnutrition and poverty in general. Although, KHG have the potential to improve food security and generate revenue, they also have social and emotional advantages (Darby, Hinton, and Torre 2020; Porter 2018). For low-income gardeners, Darby et al. (2020) found that pleasure from gardening practice motivates them, while also strengthening social ties and cultural traditions.

3.5. Conclusions and recommendations

Many countries worldwide are challenged with the task of enhancing food production to fulfil the food requirements of their population, ensuring food security, safety, quality, and nutritional demands. Home gardens, especially KHG, are considered an important method for alleviating food insecurity in impoverished homes, especially in developing countries. The study analysis has been in threefolds: 1) examine the factors that influence the adoption of KHG; 2) identify the determinants of household food security status; and 3) quantify the impact of KHG adoption on food security.

Factors that were estimated to have a higher likelihood of influence on households’ adoption of KHG include educational attainment, number of livestock, farm size and engagement in non-farm business. With regards to food security status, factors such as dependency ratio, ownership of livestock, and participation in non-farm activities significantly influence households’ food security in the rural areas of Lesotho. Moreover, the findings of the study indicated that adoption of KHG improves the probability of households being food secure by about 30.6% and reduces the probability of moderate food insecurity and severe food insecurity by 11.02% and 41.62%, respectively.

Since the adoption of KHG significantly enhances household food security status, this study recommends that policymakers and development partners design farm-level strategies that aimed at encouraging households to participate in the KHG programme. The primary worry is the uncertainty surrounding the scalability and sustainability of KHG activities due to the fact that almost all of these programmes have been carried out by non-governmental organisations (NGOs). For future upscaling programmes, it is essential that public personnel acquire adequate training to ensure that the KHG initiative achieves the same results as seen under NGOs. Promoting climate-smart gardens like KHG to rural farmers is essential to boost adoption rates, as they allow year-round production. The inclusion of KG in the mainstream rural planning and development will help to promote the practice of home gardens. Moreover, promoting a large diversity of plants and livestock in the home garden model is one of the channels to minimise food insecurity through KHG. We also recommend that household gardeners be trained on gardening practices that also consider extreme weather conditions, such as drought, frost, flood, etc. It is also suggested that encouraging building of households’ assets such as livestock rearing and engagement in non-farm income activities is critical in enhancing household food security.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Notes on contributors

Oluwasola Ogunleye

Oluwasola Ogunleye holds an MSc in Agricultural Economics from the University of the Free State, Bloemfontein, South Africa. Her research interests include sustainable agricultural practices and the analysis of rural farm household food security.

Abiodun A. Ogundeji

Abiodun A. Ogundeji is an experienced academician and a researcher who has worked for academic, national and international organisations. Abiodun is a Professor of agricultural economics and Director of the Disaster Management Training and Education Centre for Africa (DiMTEC), University of the Free State, South Africa. As a researcher and consultant, Abiodun works with international and local organizations as project leader and principal investigator, conducts research and contributes to publications in different fields of Agricultural Economics, Disaster management and is actively involved in writing research proposals for funding.

Gideon Danso-Abbeam

Gideon Danso-Abbeam is an agricultural economist, agribusiness professional and community development worker with over ten years of professional experience in teaching, research, and community development services for higher learning and development organizations. Currently, Gideon is a post-doctoral research fellow at the Disaster Management Training and Education Centre for Africa, University of the Free State, South Africa, and teaches at the Department of Agribusiness, University for Development Studies, Ghana. Gideon has worked as a consultant for both local and international organizations and has contributed to many publications in diverse fields of agricultural economics.

Notes

1. The preliminary results indicated that the awareness CGP-SPRING project variable used as instrument was significant in the treatment equation (Chi²  = 56.57, p = 0.000) but not significant in the ordered food security equation (F = 2.40, p = 0.121).

2. Note that: 1. The study first estimate the standard ordered probit model to examine the determinants of food security before fitting the stata command ‘eoprobit’ for the endogenous ordered probit model (IV-ordered probit). 2: The ATE and ATT were predicted after fitting ‘eoprobit’.