Читать книгу Hidden Hunger and the Transformation of Food Systems - Группа авторов - Страница 51

Table 1 summarizes findings by Demmler et al. [21] using the panel dataset on adults from Kenya (collected in 2012 and 2015) and panel regression models including fixed effects estimations (a form of statistical differencing), which allow for causal inference. The table depicts the effects of supermarket shopping on adults’ energy consumption from different food groups, the share of energy from highly processed foods, and the effects on nutritional outcomes. Supermarket shopping is measured in terms of a dummy variable that takes a value of 1 if any of the food consumed was purchased in a supermarket, and 0 if all foods were obtained from traditional sources. All estimates are net effects of supermarkets after controlling for confounding factors, such as income, age, education, and physical activity levels, among others.

Table 2. Effects of supermarket shopping on adults’ health

While supermarket shopping significantly reduces the amount of daily energy from unprocessed staples (–112 kcal) and fresh fruits and vegetables (–124 kcal) consumed by adults, it increases the consumption of meat and fish (+24 kcal), dairy and eggs (+9 kcal), and vegetable oils (+60 kcal). Furthermore, supermarket shopping increases the share of energy from highly processed foods by 3 percentage points. Demmler et al. [21] clearly showed that supermarkets contribute to dietary changes towards more animal-source products and foods with higher energy density and higher processing levels. At the same time, supermarkets reduce the consumption of unprocessed foods and fresh fruits and vegetables.

Supermarkets also have significant effects on adults’ nutritional outcomes in Kenya. Demmler et al. [21] found robust evidence that shopping in supermarkets significantly increases adult BMI by 0.64 and significantly increases the likelihood of being overweight/obese by 7 percentage points (Table 1). While the effects on BMI estimated with the panel data are somewhat smaller than those estimated in earlier studies with cross-sectional data [14, 17, 22], they are more robust and reliable as they better control for possible endogeneity bias.

Fig. 1. Prevalence of child and adolescent stunting and underweight depending on household shopping behavior. *** Difference between children in households shopping and not shopping in supermarkets significant at the 1% level. Authors’ presentation based on data from Debela et al. [24].