THE EXTENT OF PARTICIPATION IN URBAN AGRICULTURE AND ITS EFFECT 1 ON FOOD SECURITY IN AFRICA AND ASIA: EVIDENCE FROM GHANA AND 2 INDIA 3

This study examined the factors that influence the extent of urban agriculture participation and its 7 effect on food security in Ghana and India. A total of six hundred and fifty urban agriculture 8 farmers were interviewed for this study in Ghana and India. Food security status of urban 9 households was assessed by the use of the Household Food Insecurity Access Scale whereas the 10 determinants of the extent of urban agriculture and its effect on food security were analysed by the 11 use of the heteroskedastic linear regression and the Seemingly Unrelated Regression models, 12 respectively. From the study on average, households in Ghana were mildly food insecure, but that 13 of India was moderately food insecure. The results further revealed that, various demographic, 14 economic, institutional and health and nutrition factors differently influenced urban food security 15 and urban agriculture. Also, the extent of urban agriculture participation positively influenced food 16 security. It is recommended that, Governments and NGOs interested in the reduction of urban food 17 insecurity should aggressively advocate for urban agriculture in urban households through 18 extension education. Interests could be stimulated by gleaning on health benefits of urban 19 agriculture such as producing safe and nutritious food, the opportunity to consume chemical-free 20 food and pursue urban agriculture as a business. The implication for research is that similar studies 21 can be conducted in other emerging urban cities in Africa and Asia for the advocacy for specific 22 urban food security policies and programmes.

1 Food availability at the household level is about the household focusing on food production, reduction of postharvest losses and also ensuring the ability of the household to buy food when needed. At the same time, accessibility is, generally about the household's ability to access the food both physically and economically. This is very prominent in urban areas where there is insufficient access to food due to poverty and high prices of food. With this, in our research we proposed that urban agriculture can have significant influence on accessibility of households. Food utilisation, however, is concerned with the consumption of micronutrients to concerns about only protein and energy as seen to be in the first two dimensions. The three dimensions are not independent of each other. Hence, they should all be stable overtime. embodies the most crucial, preventive variable to UA improvement and sustainability (Orsini et 23 al., 2013). Apart from land unavailability, several metropolitans, municipal and district planning In contrast, the Indian part of the study was conducted in Bihar in the Eastern part of India. 22 Specifically, the study was done in the north (Vaishali, East Champaran, Muzaffarpur) and south 23 (Nawada, Gaya and Patna) of Bihar. The details of the study area can be seen on the map in Figure  1 2017). Following the Yamane (1967) approach, the representative sample of 5% error rate was 2 400 urban households. However, given that, not all urban households will be into UA, we sampled 3 350 households involved in UA through a multistage sampling procedure, but only 300 (86%) 4 were useful for analysis. First, Bihar was purposively selected. Secondly, three districts from both 5 North and South Bihar were purposively selected. UA households were randomly selected at the 6 third stage. It is important to note that, the data was collected by a team of trained enumerators using a 9 structured questionnaire from August to September, 2019 (for Ghana) and November, 2019 to 10 January 2020 (for India). Data was collected on household information and demographics; crops 11 and plants cultivated and livestock owned; marketing channels; household income and 12 expenditures, food security status by the use of the HFIAS measurement tool and the perception 13 of respondents on the importance of UA.  household size, primary job, animal rearing, land size, food expenditure, business, safe food, fear of pesticides, nutritious food, and distance were all hypothesised to influence the extent of UA 1 positively. Likewise, education, experience, location, animal rearing, land size, business, extension 2 contact, and distance were hypothesised to influence food security positively. However, age, 3 household size, primary job, household expenditure, and contract were all expected to influence 4 food security negatively. It should pointed out that the expected sign column in Table 1 represents 5 the various hypothesis used in this study. The hypothesis and the choice of independent variables 6 are grounded on, firstly, literature reviews from various empirical studies across the globe (the 7 supporting references column represents literature that supports the hypothesis for each variable 8 in any of the models, i.e. Model 1, 2 and 3). Secondly, consideration was given to new and 9 important variables that have mostly not been considered in food security and UA empirical 10 studies.
The dependent variables were participation intensity, scores and per capita food 11 expenditure of a household. a follow-up question on frequency of occurrence was asked to know if it happened "rarely (once 1 or twice in the past four weeks), sometimes (three to ten times in the past four weeks) or often 2 (more than ten times in the past four weeks)" (Coates et al. 2007). The highest total score a 3 household could obtain from the scale is 27 and/or 0 for a minimum. The higher the value for a 4 household, the higher the severity of food insecurity. In the study, the total cumulative score was 5 used for the regression analysis; however, for the description of the food security status of    Table 6 and 7 for details), all other explanatory variables were similar in 3 the two equations. Given these two equations, efficient estimates of the coefficients are derived 4 from the model because it allows for the error term relative to the two equations to be correlated, 5 which could not be the same for OLS. It is instructive to note that, the higher the correlation 6 coefficient between the two equations, the better the efficiency (Zellner,1962).
The linear SUR equation is specified as follows: where t = 1,…, R, i = 1, food security (HFIAS) score for each household and the household per and variance-covariance matrix  . The two linear SUR equations are: and The variance-covariance matrix is as follows: to Table 1 for details) in the model. Table 5

here, extent of participation in urban agriculture in Ghana and India)
The determinants of the extent of participation in UA in Ghana and India are presented in Table 5.  (2015) who reported that UA is increasingly becoming a source of livelihood for many urban dwellers due job loss, economic downturn as well as the limited number of employment 1 opportunities in the formal sector. Further, a household that rears animals is less likely to increase In Table 7, the independent variables explain the dependent variables in the two equations as 20 presented by a statistically-significant (p<0.01) chi-square values. Also, the results reveal that there 21 is a correlation (approximately 0.11) between the errors in the two equations suggesting that

22
Seemingly Unrelated Regression is a fit model for the analysis. As a consequence, it is revealing to note that, in Table 7, except for the variable on business, all the factors that were significant in 1 the subjective food security measurement (HFIAS sored) and the objective tool (per capita food 2 expenditure) had the same direction of influence on food security in the urban households.

3
(Insert Table 7 here, factors influencing the food security status of households in India) 4 5 From Table 7, the results show that, an increase in age decreases food insecurity in India among