Urban Food and Nutrition Security Resilience through Urban Farming: A Circular Economy Approach

This study uses a randomized control trial to investigate the impact of urban farming and using black soldier fly frass fertilizer (BSFFF) from recycled household waste for urban farming on food and nutrition security. The proposed interventions include two treatments: (1) multi-storey gardens (MSG) for planting vegetables including starter material and training on how to use (T1), and (2) MSG combined with BSFFF from recycled household waste, including starter material and training on use (T2).

2023-07-01

2023-10-31

The outcomes of interest include the household’s dietary diversity score measured as the number of food groups consumed by a household out of 12 food groups seven days before the survey; food security, per capita food consumption expenditure; income, measured as the total household consumption expenditure; and household crop production

Gendered effects of the intervention, labor supply, and time devoted to home production of food

Randomization will be performed at the enumeration area level because MSGs are relatively large and can serve several households. First, we identify all N enumeration areas within 15 villages in Kibera. We will randomly divide a sample of M (< N) enumeration areas into three. M/3 enumeration areas will receive MSG (group 1), M/3 enumeration areas will receive MSG and FF (group 2), and the remaining M/3 will be the comparison group (group 3). We will administer surveys to assess whether treatment and control groups have similar characteristics. We will implement MSG in every household in a treatment enumeration area (groups 1 and 2) and BSFFF will be available for use by all households within the treated enumeration area (group 2).

Randomization will be performed at the enumeration area level because MSGs are relatively large and can serve several households. First, we identify all N enumeration areas within 15 villages in Kibera. We will randomly divide a sample of M (< N) enumeration areas into three. M/3 enumeration areas will receive MSG (group 1), M/3 enumeration areas will receive MSG and FF (group 2), and the remaining M/3 will be the comparison group (group 3). We will administer surveys to assess whether treatment and control groups have similar characteristics. We will implement MSG in every household in a treatment enumeration area (groups 1 and 2) and BSFFF will be available for use by all households within the treated enumeration area (group 2).

Within treatment enumeration areas, we will offer the chance of every household installing MSG to grow their own vegetables. Some households might refuse treatment, in which case, they will not have access to MSG and will not be able to grow their own crops.
Randomization of MSG across enumeration areas allows estimation of the overall effect of MSG on dietary and economic outcomes by comparing treatment and control enumeration areas, even in the presence of within-enumeration area externalities. See, for instance, Duflo and Saez (2002), Miguel and Kremer (2002; 2004). This is the intention-to-treat approach.

The question is whether we can decompose the effect for treatment enumeration areas into a direct effect on treated household and an externality effect on untreated households within treated enumeration areas. This would allow us to get at the effect on treated. The assignment to treatment may be used as an instrumental variable for actual compliance provided that the exclusion restriction applies: changes in dietary and economic outcomes depend on having MSG alone, though not in being given the option of adopting MSG. If externalities are deemed important (because treated households share food with non-treated households, for instance) we will have to understand what explains selection into the program (socioeconomic characteristics, departing dietary outcomes, etc.), so that we can determine a suitable instrumental variable to correct for self-selection. We will also ask that people list people in their network to control for network intersection. Again, randomization of BSFFF across enumeration areas allows estimation of the overall effect of MSG on dietary and economic outcomes by comparing treatment and control enumeration areas.

The implementation of BSFFF will have to be such that those who refused MSG do not have access to frass fertilizer since it will be applied to random MSGs. If this implementation correction is cared for, the same reasoning for the correction of self-selection in MSG alone applies.
We will be able to identity cross- enumeration area externalities (the impact of MSG and BSFFF for households in enumeration areas located near treatment enumeration area) using exogenous variation in the local density of treatment enumeration households generated by the enumeration area-level randomization.

Done in office by a computer.

Enumeration areas

Yes

150 enumeration areas including 5-10 households

900 households

50 enumeration areas control, 50 enumeration areas with multi-storey gardens only and 50 enumeration areas with multi-storey gardens and frass fertilizer.

Based on a conservative power calculation, a sample size of 900 households will be used in this study with 600 household being assigned to the treatments (300 in T1 - multi-storey gardens and 300 in T2 - BSFFF) and 300 households being assigned to the control group. The power calculation assumes a 5 percent level of statistical significance and a statistical power of 80 percent. The power calculation further assumes a monthly per adult equivalent food expenditure of KShs. 5705. This benchmark value for the outcome indicator is from the 2015/16 Kenya Integrated Household Budget Survey. This sample size will enable the study to detect a minimum of KShs.1500 ($12) increase in monthly per adult equivalent food expenditure.