| dc.description.abstracteng | Many developing countries are currently undergoing a rapid nutrition transition. This transition is characterized by changes in dietary habits towards more energy-dense, often processed foods with high fat and sugar contents, and more sedentary lifestyles. As a result, overweight and obesity rates have increased. Worldwide obesity has more than doubled since 1980. In 2008, 34% of all adults were overweight or obese. For children under five, an estimated 6.6% were either overweight or obese in 2011, an increase from 4.5% in 1990. At the same time, undernutrition rates are still high. Globally, about 26% of all children under five were stunted, while 16% were underweight in 2011. This coexistence of undernutrition and overweight/obesity, referred to as the dual burden of malnutrition, has been observed in many developing countries, sometimes even within the same households. The nutrition transition is driven by demand-side factors, such as rising incomes and urbanization, as well as supply-side factors, such as globalizing food systems. The food retail sector is becoming more and more modernized, and supermarkets are playing an increasing role. In fact, some developing countries have witnessed a ‘supermarket revolution’, depicting a rapid spread of supermarkets within a short period of time. The retail format has an influence on the types of foods offered, as well as on sales prices and shopping atmosphere, which may affect consumer food choices.
This dissertation comprises three essays. In the first two essays, we analyze whether the spread of supermarkets in developing countries has an effect on food consumption patterns, and whether it contributes directly to rising overweight and obesity. We address this question using cross-section household and individual level data collected in Kenya in a quasi-experimental setting. Kenya has recently witnessed a rapid spread of supermarkets that now account for about 10% of national grocery sales. In addition, over 25% of the women in Kenya are overweight or obese. In the third essay, we analyze the effect of the nutrition transition on child malnutrition indicators using a cross-country regression approach with secondary panel data.
In the first essay, we analyze the effect of supermarkets on dietary behavior. The available literature suggests that supermarkets affect dietary patterns, although hard evidence is scarce. A study in Guatemala found that supermarket shoppers consume more processed foods with adverse nutritional outcomes. This is in contrast to a study in Tunisia showing that supermarket shoppers had higher dietary quality. We contribute to this limited literature by collecting and using detailed household level data on food purchases from different retail formats. We account for potential endogeneity of supermarket purchases in the regression models using instrumental variable techniques. We find that supermarket purchases increase the consumption of processed foods, both in terms of expenditure shares as well as calorie shares. An increase in the share of supermarket expenditure by one percentage point increases the share of expenditure on processed foods by 0.38 percentage points. In addition, a one percentage point increase in the share of supermarket purchases increases calorie consumption by 0.85%. For average consumers that currently do not purchase any food in supermarkets, a switch to supermarkets would translate into an additional daily consumption of 200 kilocalories, everything else held constant. This effect is partly driven by lower prices per calorie. We conclude that supermarkets affect dietary behavior, thus contributing directly to the nutrition transition.
In the second essay, we extend this analysis to examine whether supermarkets are a causal factor of overweight or obesity. Research on the impact of supermarkets on consumer nutritional status in developing countries is rare; we are only aware of one study in Guatemala. In that study, it was found that food purchase in supermarkets increases the body mass index (BMI) of consumers. However, the research for Guatemala builds on a household living standard survey that was not specifically designed for analyzing the nutritional impact of supermarkets. Hence, a few variables of interest, such as food quantities purchased in different retail outlets, were not properly captured. Moreover, the impact on BMI was analyzed for all individuals in the sample above 10 years of age, an approach that masks possible differences between adults and children. BMI is only a suitable indicator of nutritional status for adults. We use the household level data from Kenya, which we supplement with individual level anthropometric measures. To deal with the likely endogeneity of supermarket purchase variable, we employ instrumental variable regressions. We analyze the impact of supermarket purchase on nutritional status, separately for adults and for children and adolescents. We also examine impact pathways through which supermarkets affect nutritional status by estimating a system of structural equations. Controlling for other factors, buying in a supermarket increases the BMI of adults by 1.7 kg/m2 and raises the probability of adult overweight or obesity by 13 percentage points. For children and adolescents we do not find a significant impact on overweight. Instead, buying in a supermarket tends to decrease child undernutrition through a positive impact on height-for-age z-scores and a negative effect on severe stunting. For both adults and children, the nutrition impacts of supermarkets occur through higher calorie consumption and changes in dietary composition.
In the third essay, we study the effect of the nutrition transition on childhood nutritional status. One point of criticism against the use of underweight as an indicator of child undernutrition is that it is likely affected by the nutrition transition, indicating a reduction in undernutrition although nutritional quality may not really have improved. As an alternative, stunting is viewed as a more reliable indicator. It has been argued that stunting is less affected by the nutrition transition, although this has hardly been studied. We analyze the effect of the nutrition transition on childhood underweight, overweight, and stunting, employing a cross-country regression approach. We use fat consumption, share of modern retail in grocery sales, and the prevalence of overweight women as indicators of the nutrition transition. Pooling data from Demographic and Health Surveys, Planet Retail, FAOSTAT, and World Development Indicators, we estimate fixed effects and random effects panel models. Our results show that the nutrition transition has an effect on child weight, as hypothesized previously. It significantly and consistently reduces underweight rates, while the effects on child overweight are less clear. In contrast to widely held beliefs, we also find clear and consistent evidence that the nutrition transition reduces child stunting.
We derive several general conclusions. Among other factors, supermarkets are drivers of the nutrition transition in developing countries, contributing to dietary changes among consumers. Supermarkets are causing consumers not only to eat more calories, but also to get a bigger share of their calories from processed foods. The direct impact of supermarket purchase on nutrition outcomes varies by age cohort and initial nutritional status. The results for Kenya show that supermarkets increase adult BMI and overweight, whereas for children the effect is a reduction in undernutrition. Based on the cross-country analysis, we find that the nutrition transition reduces both child underweight and stunting, while the expected impact on child overweight is not so clear. Hence, the primary and secondary data analyses suggest that the nutrition transition has positive effects in terms of reducing child stunting. These results challenge the general view that the nutrition transition would only have undesirable health effects in developing countries. Of course, more research is needed to confirm these results, but our findings indicate that simplistic conclusions may not be justified. | de |