The (Re) Production of Gendered Positionality in Botswana's Commercial Urban Agriculture Sector

Abstract

Urban agriculture is increasingly touted as a key element to achieving urban productivity and sustainability, particularly in the developing world. Understanding the role and potential of commercial urban agriculture in addressing food security and economic growth requires an assessment of the agricultural systems themselves, along with the net outcomes they generate. Limitations of past research on urban agriculture make these assessments difficult. Specifically, studies tend to aggregate data such that they mask differential experiences of men and women farmers and fail to explain adequately the influence of location and human-environment relations on production systems. People's ability to create productive and sustainable urban agricultural systems is premised on who they are, where they are located, and how they interact with the environment in that location. This article presents an empirical investigation of the effects of gender on commercial urban agriculture in Greater Gaborone, Botswana. It employs a conceptual framework that bridges sociospatial and human-environment traditions in geography, and highlights gendered environments to facilitate this convergence. The investigation reveals that gender clearly influences the quantity and type of foodstuffs produced for the urban market. Gender matters because men and women enter into agricultural production, and participate within this urban economic sector, on unequal terms based on socioeconomic status, location, and interactions with the environment. If urban agriculture is to contribute to food security and economic growth, as well as urban sustainability more generally, gender relations of power, as produced and reproduced through sociospatial and human-environment relations, must inform understanding of this phenomenon.

Key Words:

• Botswana
• gender
• human-environment relations
• urban agriculture
• urban geography

Acknowledgments

I would like to thank Susan Hanson and Dianne Rocheleau for their assistance and encouragement throughout this research endeavor. Thank you to Billie Lee Turner, Luc Mougeot, Susanne Freidberg, Laila Smith, Robin Roth, Mazen Labban, Samuel Ratick, Andrew Schiller, and Mary Thomas for their insightful comments. I acknowledge the Government of Botswana and Department of Environmental Science at the University of Botswana for their support, especially Masego Mphathi, Richard Segodi, A.C. Mosha, Daphne Keboneilwe, and Moses Samson. I am grateful for the wealth of information provided by all of the farmers, entrepreneurs, and others who took the time to participate in my agricultural survey of Greater Gaborone. The research was funded by the National Science Foundation, Social Sciences and Humanities Research Council of Canada, and the AgroPolis Award from the International Development Research Centre.

Notes

^aExchange rate of US$1.00=P5.00

^aExchange rate of US$1.00=P5.00

^aExchange rate of US$1.00=P5.00

^bFigures in parentheses are numbers of producers

^aExchange rate of US$1.00=P5.00

^aExchange rate of US$1.00=P5.00

^aExchange rate of US$1.00=P5.00

1. For the purpose of this article, the male/female co-owned category has been omitted from the analysis.

2. Given the difficulty in collecting self-reported income levels, all efforts were made to increase the degree of accuracy of such data. Specifically, gross earnings from agricultural enterprises were used as a “minimum” monthly income and combined with information provided about other income-generating activities. Where possible, stated amounts of income-generating activities were cross-referenced with typical salaries in given employment categories (based on government statistics) and verified with participants.

3. Although a direct question was posed during interviews related to labor in kind, it did not figure prominently in the operation of men's or women's agricultural enterprises.

4. Even in the case of tribal land, which is free of charge, applicants must demonstrate that they have enough capital to establish and sustain their agricultural enterprise to the satisfaction of the land board.

5. Cumulative rent was calculated as the lease price per month times the number of months engaging in agricultural production while occupying the particular piece of land.

6. Discriminant analysis involved gender as the dependent variable, run against a number of independent variables based on socioeconomic, locational, and environmental groupings. Discriminant functions (based on both “enter” and “stepwise” calculations) were assessed by identifying those variables with significant individual Wilks' Lambda scores and high loadings, assessing the Wilks' Lambda value and significance of the Wald statistic, comparing the group centroids/means, and noting the percentage of original grouped cases correctly classified (using a minimum threshold of 70 percent). Logistic regression was used to verify the discriminant results given the relative sensitivity of discriminant analysis to assumption violations, as well as to the issue of combining metric with non-metric (dummy) independent variables. The strength of logistic regressions was based on the −2 Log Likelihood and related R ²s, Hosmer and Lemeshow Test of chi-square, and percentage of correctly classified observations.

7. There were a total of twenty-eight variables selected. Given the recommended ratio of five observations for one variable (Hair et al. 1998), multiple regressions were run individually for each group of variables, namely socioeconomic, locational, and environmental; these were then run in combination with each other to see whether results were consistent. Results from each multiple regression analysis were interpreted consistently according to the following criteria: relatively high and significant adjusted R ², comparison of Pearson coefficients (r) with zero order correlation coefficients (beta coefficient) to check for suppression, and looking at the condition indices and variance proportions to assess collinearity. Variables that emerged as the best explanations of the variation in the dependent variable were then combined and run together in a single regression to create the strongest possible regression equation/model. This and other cross-verification procedures were used in this multivariate analysis and confirm the validity, stability, and strength of the results. There are two further notes regarding cross-verification procedures: First, there was a degree of suppression particularly evident when the environmental variables were run in combination with either socioeconomic or spatial variables. Here, the independent variable of soil and water tended to make other variables stronger predictors of the dependent variable. Nevertheless, the individual, combined, and final regression models revealed consistent results. Second, multiple regressions were run with cases missing data excluded pairwise rather than listwise so as not to “throw away data” by decreasing the n for the overall sample. Given that there was a minimal amount of missing data, plus the fact that listwise results did not significantly alter results, the pairwise selection was justified.

8. A similar within-class gender pattern is evident, albeit not as statistically significant, among low-income producers. Specifically, low-income men (n=7) have higher levels of income than low-income women (n=20) (average P7,076 with a median of P5,800 compared to average P2,956 with a median of P2,100), invest greater amounts in fixed assets (average P9,465 with a median of P5,504 compared to average P1,630 with a median of P650), and have higher gross monthly earnings (average P3,560 with a median of P4,005 compared to P3,540 with a median of P1,623). And there are differentials in gross monthly earnings per hectare, with men earning an average P471,591 with a median of P639,130 and women earning an average P841,076 with a median of P404,600.

9. Middle-income women on tribal land generate average yields of P331,040/hectare with a median of P62,494/hectare. This is significantly higher than the middle-income group as a whole, which yields on average P150,517/hectare with a median of P27,033/hectare.

10. Long recognizing women's socioeconomic marginalization, the Government of Botswana has made special provisions for women via FAP, offering easier eligibility terms than men in qualifying for business grants, often with larger sums of funds allocated.

11. Figures for the high-income group contain a considerable amount of variation. Given confidentiality agreements and the small n in this category, however, mean and median figures are provided to show the significant differences between this group and those producers found in middle- and low-income categories.