The urban poor: profile and constraints affecting their participation in allotment gardens

ABSTRACT

Rapid global urbanisation has seen a growing number of urban poor who lack natural endowments to cope with food shortages. Broadening their safety nets merits urgent political attention but requires understanding their profiles. A survey among 88 urban poor in Benin found that they had low educational and income levels, overcrowded and unsanitary housing conditions, and limited access to social services and health facilities; 76 per cent of them were food insecure, influenced by city, gender, ownership of a motorbike, and access to health facilities. Engaging them in allotment gardens requires farming skills, financial capital, and safety issues.

KEYWORDS:

• Urbanization
• poverty
• food security
• urban planning
• urban agriculture
• cities
• Benin

1. Introduction

In 2008, for the firstly time, more than half of the global population, 3.3 billion people, were living in towns and cities (UNFPA 2007) and this figure has continued to rise rapidly since then. From an estimated 4.2 billion people worldwide in 2018, the urban population is projected to reach 6.7 billion by 2050, with almost 90 per cent of this growth happening in Asia and Africa (UN 2019). Hence, at the global level, future population growth will predominantly be in towns and cities. The World Bank (2017) reported that Sub-Saharan African (SSA) countries are undergoing the fastest rates of urban population growth and feared that this may outstrip the capacity of city administrations to provide appropriate services. A specific feature of the urban population growth is that it comprises a great number of poor people (UNFPA 2007). For instance, in 2018, the share of people living in slums in the world and SSA was 29 and 54 per cent, respectively (World Bank 2022). As population growth continues, urban policies are doing little to address the economic and social challenges faced by the urban poor.

Compared to the rural areas, specific aspects of food security in the urban context include the fact that households need to buy most of their food and so are more dependent on the market and on commercially processed food (Armar-Klemesu 2000; Tacoli 2017; Richards et al. 2016). However, many urban residents, especially in SSA countries, have limited purchasing power as most are engaged in insecure and low-paid employment in the informal economy. Food expenditure can, therefore, make up as much as 60–80 per cent of the total income among low-income urban households (Frayne et al. 2010; Maxwell et al. 1998; Tabatabai 1993). Numerous studies have argued that, for the urban poor, it is the dominance of the cash economy over access to such a basic need as food that links urban food systems to poverty and vulnerability to food insecurity (Orsini et al. 2013; Zezza et al. 2008; Mutisya et al. 2016). In addition, the urban poor are highly prone to food insecurity when the price of food goes up (Arene and Anyaeji 2010; Kc et al. 2018; Cohen and Garrett 2010; Ruel et al. 1998), because poorer households reduce what they spend on food or sacrifice important nutritional needs to satisfy their hunger. This situation reduces diet diversity and increases vulnerability to nutrition insecurity. Furthermore, the high costs of housing, transport, and health care further undermine the affordability of sufficient food (Cohen and Garrett 2010). Poor women are at a particular disadvantage because they also face socio-cultural barriers which undermine their capability to escape poverty (UNFPA 2007). Hence, the calls to support the urban poor and increase their resilience and especially their food security are clearly justified and merit political attention at the local, national, and regional levels.

Benin is a typical case in point. With a national urban poverty ratio of 39 per cent in 2019 (World Bank 2020a) and urban food-insecure households estimated at 9 per cent (INSAE and WFP 2017), Benin is one of the poorest nations in Africa. The country is experiencing a rapid urban population growth that brings more challenges than cities can cope with. For instance, from 2009 to 2019, the country’s urban population rose from 43% to 48% while the urban poverty rate at national level increased from 35% to 39% (World Bank 2020a, 2020b). Such a situation indicates that the urban poverty and the related food insecurity are on the rise. Yet, the status of the urban poor in the country is much under-researched and urban policies that aim to tackle their food insecurity require solid information on their profiles, needs, and constraints; in short, ‘who are the urban poor’? Indeed, previous studies on the determinants of poverty in Benin did not clearly explain the phenomenon from a geographical perspective such as urban areas versus rural areas (Alia, Alia, and Fiamohe 2016; Acaha-Acakpo and Yehouenou 2019). Next, they used the monetary approach and household expenditures per equivalent adult as a proxy for living standards, which is a solid approach for comparing households within developing countries (Deaton and Grosh 2000). Nonetheless, this study assumes that poor people in urban areas take matters into their own hands to deal with their food insecurity situation; hence, the need to understand what factors further explain the various levels of food security among the urban poor to inform targeted food security policy interventions in urban areas. This study aims to address these knowledge gaps and adopts, therefore, the following approach. Firstly, it creates a sociodemographic and economic profile of the food-insecure urban poor to facilitate policy targeting. Secondly, it studies the constraints for development of the urban poor by focusing on allotment gardens, a form of urban agriculture, as a specific policy intervention that empowers the poor to improve their own food security situation. This study obtained its primary data from a survey among poor urban households and used descriptive and inferential statistics to identify determinants that explain the food (in)security situation and the urban poor’s constraints for development. The final aim is to embed the results of this study in prevailing poverty alleviation and food security policies.

Concerning the proposed intervention, allotment gardens or allotments are plots of land made available by governments for individuals or families for growing food (Barthel, Folke, and Colding 2010). The land is subdivided from a few or up to several hundreds of parcels that are assigned to individuals or families. Allotments can be a cost-effective intervention to address food insecurity among urban poor, for two reasons (Escaler, Teng, and Caballero-Anthony 2010). Firstly, allotment gardens provide people living in poverty and their households with an opportunity to produce fresh and nutritious food crops, and secondly, selling surplus produce generates extra income to cover other basic needs and improve their living conditions. Despite these potential benefits and learning from the capability approach’s focus on freedom of choice and what individuals are able to do (Sen et al. 1987), engaging in allotment garden projects depends on people’s own choices that necessarily take their personal conditions into account. Thus, independently from their choices, the urban poor might face some constraints that may impede their participation in allotment garden projects; hence, the necessity to study these constraints for further considerations by urban policies. The literature provides some information on typical limitations, such as distance (Teka, Temesgen, and Fre 2018), lack of financial capital, safety, intra-household relations (Teka, Temesgen, and Fre 2018; Arene and Anyaeji 2010; Sonneveld, Thoto, and Houessou 2018), and lack of farming skills (Kc et al. 2018), which the study considers for the interviews.

This paper is organised as follows. Section 2 details the materials and methods that were used to conduct and analyse a survey held among 88 poor urban residents. Section 3 presents the results of the characterisation of the urban poor and the major constraints to their participation in allotment gardens. Section 4 discusses the results in detail. Section 5 concludes and makes policy recommendations.

2. Materials and methods

2.1. Study areas

The research was conducted in two big cities in the southern part of the republic of Benin: Abomey-Calavi and Porto-Novo. These cities provided a good context for this study because of the urbanisation process described below and that is still ongoing in the country. The municipality of Abomey-Calavi covered an area of 539 km² (Mairie d'Abomey-Calavi 2006) and in 2013 had 656,358 inhabitants, a population which had doubled in a decade (2002–2013) (INSAE 2015, 2012). There were 11 public and 90 private hospitals and the schooling rate was over 90 per cent (Mairie d'Abomey-Calavi 2006). The main economic activities were motorbike-taxi, commerce, small artisans (barber, tailor, carpenter, welder, and others), and agriculture. The incidence of human poverty in 2013 was 30 per cent (INSAE 2016) and in 2017 there were between 5 and 10 per cent of food-insecure people (INSAE and WFP 2017). The municipality of Porto-Novo covered 52 km² (Mairie de Porto-Novo 2006) with a population of 264,320 in 2013, up by 18 per cent from 2002 (INSAE 2012, 2015). Christianity, traditional religions, and Islam were the main religious faiths. There were 14 public and more than 43 private hospitals; the schooling rate was 85 per cent (Mairie de Porto-Novo 2006). The economic activities were as follows: commerce, industries, artisans, motorbike-taxi, and agriculture. The incidence of human poverty in 2013 was 31 per cent (INSAE 2016) and in 2017 there were less than 5 per cent of food-insecure people (INSAE and WFP 2017).

2.2. Research design

We derived the data for this paper from the baseline survey of a research project that designed a randomised control trial (RCT) to test the marginal effect of the participation of the urban poor in allotment gardens on their food security and income within the two cities. The RCT was conducted from 2017 to 2019 and set up two pilots of allotment gardens to test if participation of poor dwellers can improve their food security with respect to the control group. Each pilot of allotment garden was one hectare and divided into 20 small parcels that were allocated to participants. Capacity building sessions were provided to participants to support their endeavour. The control group did not participate in any training or gardening activities and was let in on their own to observe how they took matters in their own hands in the absence of policy interventions.

Beforehand, the experiment used pre-defined criteria drawn from the literature (Table 1) to recruit people living in poverty in the two cities. The recruitment followed a public announcement and the people systematically approached were living around the locations where the allotment gardens were set up and were asked if they met the criteria 1–2 and at least one of the criteria 3–5 as a condition of selection. Then, a group of 88 participants (48 in Abomey-Calavi and 40 in Porto-Novo) was recruited and randomly assigned to treatment and control groups; they were interviewed afterwards from April to August 2017. This initial survey on the outcomes assessments was conducted in 2017 at baseline (before developing the allotment gardens). The questionnaire of the survey (for this study) was structured and designed in a spreadsheet format with validated lists in scroll-down menus as a standard response and dedicated fields for open answers to give space for respondents to add more information. The inserted data were stored in a vector format that facilitated further processing. Survey instructions were given to interviewers to guide them in using the hard copy of the questionnaire in the field, using the digital questionnaire to store data, approaching, and gaining trust of the respondents, and dealing with controversial answers.

Table 1. Criteria for selection of respondents.

Nr	Criteria	Respondents	Supporting references
1	High-paid employment	No	(Armar-Klemesu 2000; Orsini et al. 2013; Zezza et al. 2008)
2	Purchasing power	Less than $1.90 a day	(Frayne et al. 2010; Maxwell et al. 1998; Tabatabai 1993)
3	Access to formal health system	No	(Teka, Temesgen, and Fre 2018; Cohen and Garrett 2010)
4	Access to credit	No	(Teka, Temesgen, and Fre 2018)
5	Access to farmland	No	(Teka, Temesgen, and Fre 2018)

The survey covered sociodemographic and economic information, frequency of access to food, and constraints for development (Table 2). The sociodemographic and economic information encompassed variables that were used to study people living in poverty. The frequency of access to food asked questions related to how many times households’ members had zero, one, two, or more meals a day over a period of 30 days. The constraints for development identified potential obstacles that might undermine the respondents to engage in allotment gardens. All these variables were drawn from the international literature on the determinants of poverty and the local context of study areas.

Table 2. Summary of variables of the survey.

Variables	Meaning	Quantification	Range	Supporting references
Sociodemographic and economic information
Age	Age of the respondent (years old)	Real-value number	18 44	(Teka, Temesgen, and Fre 2018; Arene and Anyaeji 2010)
Gender	Gender of the respondent	Category	Women Men	(Teka, Temesgen, and Fre 2018; Arene and Anyaeji 2010)
Schooling	Did you go to school? If so, what level did you reach?	Category	No schooling Primary Secondary or more	(Teka, Temesgen, and Fre 2018; Arene and Anyaeji 2010; Mutisya et al. 2016)
French literacy (reading and writing)	Can you read and write in French?	Dummy	No Yes	(Sonneveld, Thoto, and Houessou 2018)
Marital status	Are you married?	Dummy	No Yes	(Teka, Temesgen, and Fre 2018)
Schooling of spouse	Did your partner go to school?	Dummy	No Yes	(Sonneveld, Thoto, and Houessou 2018)
French literacy of spouse	Can she/he read and write in French?	Dummy	No Yes	(Sonneveld, Thoto, and Houessou 2018)
Presence of children	Do you have children?	Dummy	No Yes	(Teka, Temesgen, and Fre 2018)
School attendance/planning for children	Do they attend or are you planning to let them go to school?	Dummy	No Yes	(Sonneveld, Thoto, and Houessou 2018)
Household size	Number of persons in the household	Real-value number	1 8	(Teka, Temesgen, and Fre 2018; Arene and Anyaeji 2010)
Ownership of motorbike	Do you have a motorbike?	Dummy	No Yes	(Mairie d'Abomey-Calavi 2006; Mairie de Porto-Novo 2006)
Housing	Where do you live?	Category	Own house Rental House of family members Temporal residency	(Teka, Temesgen, and Fre 2018)
Occupation	How do you earn an income to cover your needs?	Category	Employed Motorbike-taxi Commerce Occasional jobs Tailor and barber	(Kc et al. 2018; Oldewage-Theron and Kruger 2011; Legwegoh and Hovorka 2013; Armar-Klemesu 2000; Mairie d'Abomey-Calavi 2006; Mairie de Porto-Novo 2006)
Access to farmland	Do you have access to farmland?	Dummy	No Yes	(Arene and Anyaeji 2010)
Access to credit	Do you have access to credit?	Dummy	No Yes	(Teka, Temesgen, and Fre 2018; Arene and Anyaeji 2010)
Access to formal health Systems	Do you have access to the formal health system?	Dummy	No Yes	(Teka, Temesgen, and Fre 2018; Cohen and Garrett 2010; Arene and Anyaeji 2010)
Income	How much do you earn approximately per month?	Real-value number	> 0	(Teka, Temesgen, and Fre 2018; Cohen and Garrett 2010; Arene and Anyaeji 2010)
Food access indicators
Zero meal	How many times during the last month did any household member not have a single meal during the day? number of times any household member had only one meal during the day; (c) number of times	Real-value number	0 30	(Sonneveld, Thoto, and Houessou 2018; Mutisya et al. 2016)
Single meal	How many times during the last month did any household member have only one meal per day?	Real-value number	0 30	(Sonneveld, Thoto, and Houessou 2018; Mutisya et al. 2016)
Two or more meals	How many times during the last month did household members have two or more meals per day?	Real-value number	0 30	(Sonneveld, Thoto, and Houessou 2018; Mutisya et al. 2016)
Constraints for participation in allotment gardens
Lack of skills	Is the lack of skills a constraint for participation in an allotment?	Dummy	No Yes	(Kc et al. 2018; Mutisya et al. 2016)
Lack of time	Is the lack of time a constraint for participation in an allotment?	Dummy	No Yes	(Sonneveld, Thoto, and Houessou 2018)
Distance to allotments	Is the distance a constraint for participating in an allotment?	Dummy	No Yes	(Teka, Temesgen, and Fre 2018)
Lack of financial capital	Is the lack of capital a constraint for participating in an allotment?	Dummy	No Yes	(Teka, Temesgen, and Fre 2018; Arene and Anyaeji 2010)
Households’ relations	Are intra-household relations a constraint for participating in an allotment?	Dummy	No Yes	(Arene and Anyaeji 2010)
Safety	Is safety a constraint for participating in an allotment?	Dummy	No Yes	(Arene and Anyaeji 2010)

2.3. Data processing and analysis

We transformed the real-value number variables into categorical variables as follows: firstly, the ‘age’ was transformed into two classes: Youth (≤35) and Adults (>35) using the definition of youth adopted by the African Union Commission (African Union 2006) and, secondly, the size of the household was transformed into two classes using the average household size (5) in Benin (INSAE 2015): H1 ≤ 5 and H2 > 5. Next, we calculated the latent variables of food security using the household hunger scale that estimates prevalence of severe food insecurity across context and focuses on the quantity dimension of food access (Leroy et al. 2015). Using a four-week (30 days) period, we asked three items related to the number of daily eating occasions as follows: a) number of times any household member did not have a single meal during the day; b) number of times any household member had only one meal during the day; c) number of times household members had two or more meals during the day. Responses were dichotomised and respondents who experienced items ‘a’ and ‘b’ were assigned ‘1’, else ‘0’; while the response ‘c’ was negated and respondents who experienced such an item were assigned ‘0’, else ‘1’. Then, we calculated a sum of the item scores, ranged from zero to three. Respondents who received a score of zero were categorised as food secure while those who received a score of one were categorised as moderately food insecure. Respondents who received a score of two or three were categorised as severely food insecure.

Data were analysed in Stata version 16 and in four ways. Firstly, we used descriptive summary and chi-square statistics to profile the respondents. Secondly, a stepwise ordered logistical regression was applied to determine the explanatory variables that influenced food security. An ordered logistical regression uses the maximum likelihood estimation to determine probabilities of correct classification of the explained variable. (See additional information in Greene (1980), Maddala (1986), and Davidson and MacKinnon (1993) for a more comprehensive description of discrete choice models). Below we briefly explain the ordered logit model. In the ordered logit model, additive error terms are used, so that the underlying process is given by: (1) $y_i={\beta }^{\mathrm{\prime }}{\mathrm{x}}_{\mathrm{i}}+{\epsilon }_{\mathrm{i}},$(1) where yi represents the food security status, i refers to observation number, β the vector of parameters to be estimated that belong to independent variables xi (city, age, sex, schooling, French literacy, marital status, presence of children, household size, ownership motorbike, housing, occupation, access to farmland, access to credit, and access to formal health systems), ϵi is the disturbance, assumed to be independent across observations. Observed is variable zi of the ordered food security classes (severely food insecure, moderately food insecure, food secure) that is related to yi in that adjacent intervals of yi correspond with qualitative information zi. This relation is given by: (2) $\begin{array}{cc}\mathrm{zi}=1& \mathrm{if}{y}_i<{\mu }_1,\\ z_i=2& \mathrm{if}{\mu }_1\le y_i<{\mu }_2,\\ ⋮& \\ z_i=n& \mathrm{if}{\mu }_{\mathrm{n}\text{-}1}\le y_i.\end{array}$(2) The ordering requires thresholds (µ1,..,µn-1) to satisfy µ1< µ2 < .. < µn-1. Parameters β and thresholds (µ1, … ,µn-1) are simultaneously estimated using the maximum likelihood method, which maximises the probability of correct classifications.

We calculate the probability (Pr) that zi = 1 by: $\begin{array}{cc}\mathrm{Pr}({\mathrm{z}}_{\mathrm{i}}=1)& =\mathrm{Pr}({\mathrm{y}}_{\mathrm{i}}<{\mu }_1)=\mathrm{Pr}({\epsilon }_{\mathrm{i}}<{\mu }_1-{\beta }^{\mathrm{\prime }}{\mathrm{x}}_{\mathrm{i}})=\mathrm{F}({\mu }_1-{\beta }^{\mathrm{\prime }}{\mathrm{x}}_{\mathrm{i}})\end{array},$the probability that zi = 2 by: $\begin{array}{cc}\mathrm{Pr}({\mathrm{z}}_{\mathrm{i}}=2)& =\mathrm{Pr}({\mu }_1\le {\mathrm{y}}_{\mathrm{i}}<{\mu }_2)=\mathrm{Pr}({\mu }_1<{\beta }^{\mathrm{\prime }}{\mathrm{x}}_{\mathrm{i}}+{\epsilon }_{\mathrm{i}}<{\mu }_2)\\ & =\mathrm{Pr}({\epsilon }_{\mathrm{i}}<{\mu }_2-{\beta }^{\mathrm{\prime }}{\mathrm{x}}_{\mathrm{i}})\text{-}\mathrm{Pr}({\epsilon }_{\mathrm{i}}<{\mu }_1-{\beta }^{\mathrm{\prime }}{\mathrm{x}}_{\mathrm{i}})\\ & =\mathrm{F}({\mu }_2-{\beta }^{\mathrm{\prime }}{\mathrm{x}}_{\mathrm{i}})\text{-}\mathrm{F}({\mu }_1-{\beta }^{\mathrm{\prime }}{\mathrm{x}}_{\mathrm{i}})\end{array}$and the probability that zi = n by: $\begin{array}{cc}\mathrm{Pr}({\mathrm{z}}_{\mathrm{i}}=\mathrm{n})& =\mathrm{Pr}({\mathrm{y}}_{\mathrm{i}}\ge {\mu }_{\mathrm{n}\text{-}1})=\mathrm{Pr}({\epsilon }_{\mathrm{i}}\ge {\mu }_{\mathrm{n}\text{-}1}-{\beta }^{\mathrm{\prime }}{\mathrm{x}}_{\mathrm{i}})=\mathrm{F}({\beta }^{\mathrm{\prime }}{\mathrm{x}}_{\mathrm{i}}-{\mu }_{\mathrm{n}\text{-}1})\end{array}.$To meet requirements of a probability model (monotonic-increasing cumulative distribution function and results lie between 0 and 1), the disturbances ϵi are assumed to possess a logistic distribution, leading to a cumulative logistic transformation function Λ¹ that maps the admissible area of y, i.e. (–∞, ∞), to [0,1], with a firstly derivative that is always positive. Thus, the likelihood function for the ordered logit model that consists of (1) and (2) for n = 3 is given by: (3) $\ell (\beta ,{\mu }_1,{\mu }_2)=\prod _{{\mathrm{y}}_{\mathrm{i}}=1}\mathrm{\Lambda }({\mu }_1-{\beta }^{\mathrm{\prime }}{\mathrm{x}}_{\mathrm{i}})\cdot \prod _{{\mathrm{y}}_{\mathrm{i}}=2}(\mathrm{\Lambda }({\mu }_2-{\beta }^{\mathrm{\prime }}{\mathrm{x}}_{\mathrm{i}})-\mathrm{\Lambda }({\mu }_1-{\beta }^{\mathrm{\prime }}{\mathrm{x}}_{\mathrm{i}}))\cdot \prod _{{\mathrm{y}}_{\mathrm{i}}=3}\mathrm{\Lambda }({\beta }^{\mathrm{\prime }}{\mathrm{x}}_{\mathrm{i}}-{\mu }_2).$(3) The function is minimised with respect to the parameters β, µ1 and µ2.

Since reporting and interpreting logistic regression results is not straightforward and is tricky to understand, the odds ratios were reported to illustrate the relationship between outcomes and observations (Peng, Lee, and Ingersoll 2002). The odds ratios are directly derived from regression coefficients in a logistic model and interpreted as the change in the odds of yi given a unit change in xi when all other predictors are held at a constant (Peng et al. 2002). The odds ratio is computed by: $Oddsratio=e^{\beta }$Thirdly, we used chi-square statistics to explore the constraints that the urban poor might face to engage in allotment garden projects. Last, several discussions were held with respondents during the research to better understand the local context and how it might affect the findings. The outcomes were used to augment the discussion.

3. Results

3.1. Socioeconomic characteristics of respondents

Table 3 showed the characteristics of the respondents concerning their sociodemographic and economic background. Overall, there were more women and youths than men and adults, respectively. Less than half of the respondents went to school but very few could read and write French. Most respondents were married and had children. Approximately half of the respondents’ spouses went to school but only a smaller number could read and write French. Their children were attending school or were planned to go to school. Using the country’s average household size as threshold, more than half of the respondents had large families. However, there were significantly more large families in the city of Porto-Novo than Abomey-Calavi. The data also showed that only few respondents had a motorbike. In terms of housing, respondents indicated that they lived in the house of family members, in their own house, in rented accommodation, or in temporary residency, respectively. They did not have a stable employment but were used to practise various jobs (occasional jobs, commerce, motorbike-taxi) to cover their basic needs. They earned on average FCFA 15458 (€ 24) per month; 75 per cent indicated that they earned FCFA 16,000 (€ 24.4) or less per month. On average, respondents of the city Abomey-Calavi were better off than those of Porto-Novo. Furthermore, the urban poor had a low access to social facilities. For example, few respondents indicated that they had access to modern healthcare services, farmland, and credit, respectively. However, there was a major difference between the two cities concerning access to farmland and formal health systems. While in Abomey-Calavi 44 and 56 per cent had access to farmland and formal health systems, respectively, only 5 and 3 per cent had access to farmland and formal health systems, respectively, in Porto-Novo. Finally, few respondents were food secure while almost half were severely food insecure. It is noteworthy to indicate that the food-insecurity phenomenon was significantly worse in Porto-Novo.

Table 3. Socioeconomic characteristics of the respondents.

Variables	Modalities	Frequency / Average
		Abomey-Calavi	Porto-Novo	All
Gender	Women	70.83	87.50	78.41
	Men	29.17	12.50	21.59
Age	Youth (≤35)	85.42	87.50	86.36
	Adult (>35)	14.58	12.50	13.64
Schooling	No schooling	52.08	70.00	60.23
	Primary	35.42	25.00	30.68
	Secondary or more	12.50	5.00	9.09
French literacy (reading and writing)	No	79.17	85.00	81.82
	Yes	20.83	15.00	18.18
Marital status	No	10.42	10.00	10.23
	Yes	89.58	90.00	89.77
Schooling of spouses*	No	29.17	55.00	40.91
	Yes	58.33	35.00	47.73
French literacy of spouses**	No	39.58	42.50	40.91
	Yes	47.92	20.00	35.23
Children	No	8.33	0.00	4.55
	Yes	89.58	95.00	92.05
School attendance / planning for children	No	0.00	2.50	1.14
	Yes	89.58	92.50	90.91
Household size**	≤5	58.33	30.00	45.45
	>5	41.67	70.00	54.55
Ownership of motorbike	No	85.42	85.00	85.23
	Yes	14.58	15.00	14.77
Housing**	Own house	16.67	37.50	26.14
	Rental	37.50	5.00	22.73
	House of family members	39.58	32.50	36.36
	Temporal residency	6.25	25.00	14.77
Occupation*	Low-paid employment	4.17	15.00	9.09
	Motorbike-taxi	6.25	0.00	3.41
	Commerce	33.33	35.00	34.09
	Occasional jobs	41.67	47.50	44.32
	Tailor and barber	14.58	0.00	7.95
Access to farmland**	No	56.25	95.00	73.86
	Yes	43.75	5.00	26.14
Access to credit	No	68.75	82.50	75.00
	Yes	31.25	17.50	25.00
Access to formal health systems**	No	43.75	97.50	68.18
	Yes	56.25	2.50	31.82
Monthly Income**	Mean	18953.45	11062.86	15457.62
	Standard error	2880.17	788.14	1692.33
Food security status**	Food secure	33.33	12.50	23.86
	Moderately food insecure	35.42	17.50	27.27
	Severely food insecure	31.25	70.00	48.86

*Significant at 5%; **Significant at 1%

3.2. Factors associated with the food security of the respondents

As our dependent variable is categorical, we did aim to explain the food-security situation of the respondents by applying a stepwise ordered logistical regression; Table 4 showed the results. The final model was highly significant (p < 0.1%) and four variables were important in explaining the food-security of the urban poor: city, gender, ownership of a motorbike, and access to modern healthcare services. In the city of Porto-Novo, respondents were less likely to be food secure than in Abomey-Calavi. Men were also less likely to be food secure than women. Inversely, respondents who owned a motorbike and had access to formal health systems were more likely to be food secure than those who did not have access to these resources, respectively.

Table 4. Results of the ordered logistic model on food security groups.

Variables	Coef.	Std. Err.	z	P>|z|	[95% Conf. Interval]
City ƚ: Porto-Novo	−1.393924	0.5883425	−2.37	0.018	−2.547054 −0.2407943
Gender ƚƚ: Men	−1.989665	0.6999821	−2.84	0.004	−3.361605 −0.6177253
Motorbike ƚƚƚ: Yes	1.483558	0.7322281	2.03	0.043	0.0484174 2.918699
AHealthSyst ƚƚƚƚ: Yes	1.151264	0.56277	2.05	0.041	0.048255 2.254273
/cut1	−0.5219327	0.4893004			−1.480944 0.4370785
/cut2	1.110711	0.4996811			0.1313538 2.090068

ƚ: The reference was Abomey-Calavi; ƚƚ: Women; ƚƚƚ: Do not own a motorbike; ƚƚƚƚ: Do not have access to health system

Although, there are several R2-like statistics that can be used to measure the strength of the association between the dependent variable and the predictors, they are not as useful as the R2 statistic in regression, since their interpretation is not straightforward; the correct classification rate (hit ratio) may be more important (Norusis 2012). Therefore, Table 5 mapped observed data against model results and showed that 61 per cent of our observations were correctly classified. In 17 cases, the model overestimated food insecurity. In nine out of the 21 cases that observations were food secure, the model predicted a moderate (5) or severe (4) food insecurity; when observations were moderately food insecure the model predicted in eight cases severe food insecurity. More seriously, the model underestimated food insecurity in 17 cases. In 10 cases, the observation was moderately food insecure while the model predicted a food-secure situation. In three and four cases, the model predicted food secure and moderately food secure, respectively, while corresponding observations reported a severe food-insecure situation.

Table 5. Hit ratio of the ordered logistic model.

		Predicted probabilities
Food security status	Observations	Food secure	Moderately food insecure	Severely food insecure
Food secure	21	12	5	4
	23.86	13.64	5.68	4.55
Moderately food insecure	24	10	6	8
	27.27	11.36	6.82	9.09
Severely food insecure	43	3	4	36
	48.86	3.41	4.55	40.91

3.3. Constraints to engage in allotment gardens

Since engaging in allotment gardens depends on people’s own choice, we asked respondents if they were willing to do so. In response, 99 per cent indicated that they were willing to participate in an allotment garden project to diversify their livelihoods; 74 per cent indicating that this choice was motivated by self-consumption and additional income. Afterwards, we explored the constraints they may face in such endeavours in comparison with key sociodemographic characteristics (Table 6). We found that young people were more likely to face a lack of farming skills than adults. Although not significant, both youths and adults indicated the need of financial capital to start the activity. Regarding gender, women were more likely to lack skills and financial capital than men. However, neither education levels nor having children had influence on participants in facing any constraints. Last, large families were more likely to indicate safety issues associated with the commute to gardens and thefts in the neighbourhood.

Table 6. Constraints faced by the urban poor to engage in allotment gardens.

	Lack of skills %		Lack of time %		Distance to allotments %		Lack of capital %		Household’s relations %		Safety %
Constraints Variables	Yes	No	Yes	No	Yes	No	Yes	No	Yes	No	Yes	No
Age
Youth	59 67.05	17 19.32	5 5.86	71 80.68	4 4.55	72 81.82	51 57.95	25 28.41	4 4.55	72 81.82	8 9.09	68 77.27
Adult	6 6.82	6 6.82	1 1.14	11 12.5	1 1.14	11 12.5	10 11.36	2 2.27	0 0.00	12 13.64	3 3.41	9 10.23
	Pr = 0.043		Pr = 0.823		Pr = 0.669		Pr = 0.257		Pr = 0.416		Pr = 0.159
Gender
Women	55 62.5	14 15.91	6 6.82	63 71.59	5 5.68	64 72.73	44 50.00	25 28.41	3 3.41	66 75.00	8 9.09	61 69.32
Men	10 11.36	9 10.23	0 0.00	19 21.59	0 0.00	19 21.59	17 19.32	2 2.27	1 1.14	18 20.45	3 3.41	16 18.18
	Pr = 0.017		Pr = 0.183		Pr = 0.227		Pr = 0.031		Pr = 0.865		Pr = 0.624
Education
No school	41 46.59	12 13.64	2 2.27	51 57.95	2 2.27	51 57.95	33 37.50	20 22.73	2 2.27	51 57.95	7 7.95	46 52.27
Primary	19 21.59	8 9.09	3 3.41	24 27.27	3 3.41	24 27.27	23 26.14	4 4.55	1 1.14	26 29.55	4 4.55	23 26.14
Secondary	5 5.68	3 3.41	1 1.14	7 7.95	0 0.00	8 9.09	5 5.68	3 3.41	1 1.14	7 7.95	0 0.00	8 9.09
	Pr = 0.594		Pr = 0.375		Pr = 0.312		Pr = 0.100		Pr = 0.526		Pr = 0.522
Children
Yes	62 70.45	19 21.59	6 6.82	75 85.23	5 5.68	76 86.36	56 63.64	25 28.41	3 3.41	78 88.64	11 12.50	70 79.55
No	3 3.41	4 4.55	0 0.00	7 7.95	0 0.00	7 7.95	5 5.68	2 2.27	1 1.14	6 6.82	0 0.00	7 7.95
	Pr = 0.052		Pr = 0.456		Pr = 0.498		Pr = 0.900		Pr = 0.197		Pr = 0.297
Household size
Small family	29 32.95	11 12.50	4 4.55	36 40.91	2 2.27	38 43.18	27 30.68	13 14.77	3 3.41	37 42.05	0 0.00	40 45.45
Large family	36 40.91	12 13.64	2 2.27	46 52.27	3 3.41	45 51.14	34 38.64	14 15.91	1 1.14	47 53.41	11 12.50	37 42.05
	Pr = 0.790		Pr = 0.280		Pr = 0.801		Pr = 0.736		Pr = 0.224		Pr = 0.001

4. Discussion

The research clearly used a set of criteria to identify people living in vulnerable situations to understand the factors that were associated with their food security. The study also used a case study to identify possible obstacles that may hinder their participation in allotment gardens initiatives.

Our study found that the urban poor have a very low education level. Indeed, despite the high schooling rates in the two cities (90 and 85 per cent in Abomey-Calavi and Porto-Novo, respectively), fewer than half indicated that they went to school and only a small number (18 per cent) among the urban poor can read and write French. That implies that among those who started school half of them dropped out, which is a cause for concern and a call for policy intervention in poor identified areas within the urban setting. The low level of education among urban poor is also confirmed by Acquah, Kapunda, and Legwegoh (2016) who reported for Gaborone in Botswana that in the capital city’s poorer areas only 21 per cent had completed high school or more. However, it is reassuring that most of respondents let or plan to let their children go to school and achieve higher levels, broadening their chances of overcoming poverty. Indeed, deciding to spend on school fees, school uniforms, and school meals shows parents’ willingness to help their children improve their lives.

The respondents have, on average, five members in their household, which matches the national average (INSAE 2015), but we found a higher number of large families in the city of Porto-Novo. As the urban poor, by definition, have limited finances, they cannot live in a comfortable house. For instance, a small number (26 per cent) built their own house with precarious materials and the rest lived either with family members (36 per cent) or in rented accommodation (23 per cent). To cover their basic needs such as food, the respondents undertake various economic activities such as occasional jobs, commerce, and small paid services (tailor, motorbike-taxi). Such endeavours show that in the absence of a stable income-generating activity, poor urban dwellers take matters in their own hands to earn money, though limited, to cover their crucial needs. However, despite the efforts, their food-security status is worrying: 76 per cent are food insecure, probably because their earnings are very low and do not cover their food needs.

To substantiate this conclusion, we conducted a stepwise ordered logistical regression and found that food security of respondents was influenced by ownership of a motorbike, access to formal health systems, gender, and city. Indeed, owning a motorbike in urban Benin tends to broaden the possibilities for urban residents who are living in poverty to diversify opportunities to secure an income and improve food security. By exponentiating the model’s coefficients, findings showed that the odds of being food secure for respondents who own a motorbike were more than four times greater than among those who do not. The findings also show that the food-security status is correlated by access to formal health systems, a conclusion supported by Teka, Temesgen, and Fre (2018) who find that a relatively lower incidence of poverty is related to access to health facilities. Indeed, discussions with respondents showed that avoiding the traditional medicine that is widely used by poor communities and having access to modern healthcare facilities implies that this group of urban residents has, to some extent, the financial means to cover some basic needs such as food. Taken together, ownership of a motorbike and access to formal health systems give a picture of some financial capacity among people living in poverty in urban areas. Hence, the two indicators may be useful to identify and categorise poor people living in food insecurity in peri-urban and urban areas.

Furthermore, the findings suggest that the odds of being food secure for men are lower than for women which, is surprising and contradicts what is usually conveyed in the literature. For instance, in a study in two informal urban settlements in Kenya which assessed food security over a recall period of 30 days, as in our research, Mutisya et al. (2016) found that men had less likelihood of being food insecure than women. Discussions with respondents showed that as women process the food in the households, they have more opportunities to increase the number of meals per day and therefore more room to improve food security. Men are working or looking for jobs outside the home and have little time to eat (against the trade-off of working and earning money). This plausible explanation is also found in a study in the Indian slums by Lumagbas (2017), namely that women tend to have more access to food than men mostly because they are homemakers.

Last but not least, the findings suggest that the odds of being food secure for poor people in Porto-Novo are lower than for those who live in the city of Abomey-Calavi. The explanation for this might be two-fold. Firstly, learning from the characteristics of the urban poor, respondents living in Abomey-Calavi seem to be better off than those in Porto-Novo, which might have influenced their food-security status. For instance, we found significant effects between respondents of the two cities in terms of income, access to farmland (with benefits regarding access to food) and household size. Respondents living in Abomey-Calavi had higher monthly earnings, more access to farmland and fewer people to feed (leaving more food for household members) compared to those in Porto-Novo. We therefore assume that the variable “city” might have hidden the effects of these variables in the stepwise ordered logistical modelling. In addition, the findings might suggest that giving poor people access to land in the form of allotments might assist them in producing food and improving food security. Secondly, the results of the stepwise ordered logistic modelling suggest that the food security of respondents is influenced by ownership of a motorbike and access to formal health facilities. In the same vein, respondents living in Abomey-Calavi significantly had better access to health facilities and were the ones only involved in motorbike-taxi as compared to those in Porto-Novo, which might explain why there were more food secure people in Abomey-Calavi. In addition, the findings might imply that having assets is necessary but not sufficient to improve livelihoods. Rather, leveraging the potential of these assets, such as an allotment garden, is crucial to an improvement of living conditions and enabling people to overcome poverty.

Improving the livelihoods of poor people requires improving their capabilities that enable them to make own choices. Therefore, we tested whether some identified constraints might impede the respondents to engage in allotment garden projects. We found that, in general, three factors merit further attention: lack of farming skills, lack of financial capital, and safety issues. Nevertheless, and learning from respondents, the magnitude of these constraints is not the same for everyone. For instance, young people were more likely to face a lack of farming skills than adults to participate in an allotment project. This finding may indicate that adults to some extent accumulate life experiences including farming that influence their decisions regarding participation in allotment gardens. However, for both groups, the lack of financial capital is a constraint. The latter is not surprising since the respondents were selected among people living in poverty. The two previous findings are gendered since women were more likely than men to lack farming skills and financial capital. Furthermore, large families were more likely to indicate safety issues associated with the commute to gardens and thefts in the neighbourhood. Indeed, it is most likely that in large families as opposed to small ones, households, and thereby women, may have more children to care for, which may explain why the safety issue is important for such families. For instance, in a qualitative study in Blantyre (Malawi), Riley and Dodson (2016) find that women with small children often find it difficult to carry them up a steep slope or to leave them alone at home to go to the market. Therefore, policy interventions should consider the previous constraints and be gender-sensitive in engaging poor people living in urban areas in allotment gardens. Noteworthy of indication is that beyond these specific policy interventions, it is important to also improve poor-identified areas within cities with basic infrastructure and services as they also contribute to reduce urban poverty (Lynch, Berliner, and Mariotti 2016).

Our sampling methodology mixed purposive targeting and a random assignment to reach the research target group. Based on the purpose of the study and the broader RCT project, this methodology only generated a small sample size that is not representative of the population of the urban poor. Therefore, the conclusions are applicable only to the study sample and cannot be systematically generalised to the cities. However, most of the world’s new population growth is going to be in African cities and, with this process, comes the urbanisation of poverty and food insecurity. Hence, there is an urgent need to understand the people involved, the processes underlying urban poverty and food insecurity and how to design policy interventions to effectively overcome them. To that end, the paper provides valuable insights to better understand the food security of people living in poverty in the poorer cities of a rapidly urbanising world.

5. Conclusion and policy recommendations

Our paper contributes to the body of knowledge on the characterisation and identification of people living in poverty and food insecurity in urban areas. Moreover, the study suggested a concrete policy intervention by focusing on allotment gardens that should improve the food security of the urban poor and learn from this case study to identify possible obstacles that may hinder their participation. The findings showed that the level of education and income is low among respondents, who are living in overcrowded and unsanitary conditions. The high prevalence of food insecurity among them owing to a lack of concrete economic opportunities is a cause for concern, and requires policies to define targeted programs, such as allotment gardens, as a way to offer a social safety net and thus improve their food security. We found that 99 per cent of respondents were willing to engage in allotment gardens to diversify their livelihoods. However, they might face important constraints such as the lack of farming skills, the lack of financial capital, and safety issues, which may hinder them from successfully participating in allotment gardens. Tackling these constraints is not an easy task and needs policy interventions that take into account the following: 1) focus on technical and financial capacity-building programs in the development and management of allotment gardens especially for youth and women who experience more acutely the lack of skills and financial capital; 2) design gender-sensitive programs by considering aspects such as the safety since women who have to take care of children may face difficulties in participation.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by Nederlandse Organisatie voor Wetenschappelijk Onderzoek, File no: 08.260.302 and Nuffic: R/003248.01.

Notes on contributors

Mawuna D. Houessou

Donald Houessou is researcher at ACED, a policy-research institution that generates and shares evidence to inform policymaking in the food security and environment sectors in Benin. He completed a PhD at Athena Institute of the VU University of Amsterdam, Netherlands. His research project was focused on the development of an integrated framework that facilitates the development of urban agriculture in Benin and sub-Saharan Africa at general. He holds an MSc. in agricultural economics from the Faculty of Agricultural Sciences of the University of Abomey-Calavi, Benin. He has extensive research experiences in food security, community resilience, fisheries, and urban food systems issues.

Ben G. J. S. Sonneveld

Ben Sonneveld (1957) is deputy director of the Amsterdam Centre for World Food Studies (ACWFS) of the VU University Amsterdam. He holds an MSc from Wageningen University, Netherlands, with majors in tropical crop science, soil science, and plant physiology. He worked from 1985 to 1987 in Indonesia for the International Institute for Land Reclamation and Improvement and from 1987 to 1992 for the Food and Agriculture Organisation of the United Nations in land and water development projects in Latin America. In 1992 he joined the VU University where he has been involved in food security projects in Ethiopia, Nigeria, Benin, Jordan River Basin, Palestinian Territories, and Senegal. For his PhD, he received the AIID/World Bank Thesis Award. Sonneveld publishes in leading natural resource management and ecological economic journals. A central topic of his research is the modelling of the impact of land and water degradation on agricultural production and eco-services.

Augustin K. N. Aoudji

Augustin Aoudji is lecturer and researcher at the Faculty of Agricultural Sciences of the University of Abomey-Calavi (FSA-UAC), Republic of Benin since 2012. He holds a PhD in Agricultural Economics and has extensive experience in research activities. His research interests centre on issues related to value chain analysis, food security, agribusiness and agricultural entrepreneurship, and project engineering. These have resulted in the publication of several peer reviewed journal articles, occasional reports, and the coordination of projects. He is principal investigator of the project “Enhancing urban food security through development of allotment gardens in and around the cities of Benin” at FSA-UAC, funded under NWO-WOTRO. Augustin Aoudji is driven by success and passion to influence change that can lead to improvements in social outcomes in the community, particularly among marginalised population.

Fréjus S. Thoto

Frejus Thoto is the Executive Director and Lead Researcher at ACED. Frejus has coordinated several policy-research projects in various fields including fisheries, biodiversity, climate change, urban agriculture, agribusiness, and youth employment. He also served as a Knowledge Management Expert for a joint USD 12 million intervention of the African Development Bank and The African Capacity Building Foundation that aimed at promoting results culture for Africa's transformation. Frejus completed his PhD studies on the profiles and formalisation dynamics of agricultural entrepreneurs in Benin.

Denyse J. R. M. Snelder

Denyse Snelder obtained her PhD in 1993 at the University of Toronto, Canada, based on soil erosion research in Baringo, Kenya. Between 1994 and 2009, she worked as assistant professor at the Institute of Environmental Sciences, Leiden University, The Netherlands. Then, she joined as senior specialist in sustainable land and water management at the Centre for International Cooperation, Vrije Universiteit Amsterdam, where she manages various international projects that combine capacity building with research and education: an EU-funded project (2011–2015) on water harvesting technologies in Sub-Saharan Africa; a project (2014–today) on sustainable approaches to livelihood improvement in Kenya; and projects funded by NUFFIC: one on natural resource management and ecotourism in Ethiopia (2011–2018); another one on conflict, conservation and collaboration in natural resource management of Virunga Landscape (2017-today); and one on spatial planning for agribusiness and public policy development in West Kenya (2017–today). Her research interest is in food security, biodiversity conservation, and agroforestry.

Anselme A. Adegbidi

Anselme B. E. A. Adegbidi is Professor in Agricultural economics (PhD at Groningen University, The Netherlands) at the University of Abomey-Calavi. Former Head of the Department of Rural Economics, Sociology and Anthropology, he is now Head of the Centre of Expertise and Consultancy of the Faculty of Agricultural Sciences “FSA”. As the national Coordinator in chief of the IDRC network MIMAP (Micro and Macro Adjustment Policies) for almost ten years, he conducted research on distributive analysis, poverty and inequality profiles and related studies on education and employment. He is author of many publications mainly on farm household’s decision making on soil conservation, farm products marketing, climate change, and farm households’ organisation (innovation platforms).

Tjard De Cock Buning

Prof. dr. T. De Cock Buning is Emeritus Professor at the VU university in Amsterdam, Athena Institute, The Netherlands. His main interests of research are complex challenges for society and government and the mechanisms to deal with them: public engagement, patient participation, multistakeholder platforms, transdisciplinary approaches, communities of practices. In this context he has been involved in the biotechnology debate, predictive medicine, cloning, sustainable agriculture, animal welfare issues, sustainable public health, corporate social responsibility. The last eight years he worked with his PhDs, NGOs and academic partners on issues such as improvement of maternal health services in Burundi, DR Congo, India, and Nepal; Food security and climate change in Ethiopia; and food security in peri-urban areas. He is senior investigator of Amsterdam Centre for World Food Studies with a special focus on system analysis for system wise solutions.

Notes

1 $\mathrm{\Lambda }={\displaystyle \frac{1}{1+e^{-(\bullet )}}}$.