Public policy and social determinants of nutrition behaviour and outcomes – quasi-experimental evidence from low- and middle-income countries

As of today approximately 3.19 billion people worldwide, i.e. 42 percent of the world’s population, are malnourished. Out of them 811 million are undernourished and 2.38 billion people are overweight or obese. Both undernutrition and overnutrition are a health risk for the affected individuals, and lower their productive capacities and labour market perspectives. This thesis provides evidence on how public policy can create an incentive architecture which is conducive to healthy nutrition behaviour in low and middle income countries (LMICs). The first paper analyses whether the conditional cash transfer programme Bolsa Familia in Brazil has influenced food consumption and nutritional outcomes among its beneficiaries. The results show that the bulk of the cash transfers is spent on food, with a disproportionate increase in the consumption of dairy and sugary products, but no overall impact on overweight and obesity. The second paper investigates whether the free health insurance programme Seguro Popular in Mexico has altered nutritional choices and outcomes among lowincome families in Mexico. The analysis suggests that the programme has increased obesity among those who were already overweight at baseline, and that beneficiaries have reduced the consumption of carbohydrates in favour of meat. The third paper focuses on the importance of gender norms in determining nutritional outcomes and describes the growing disparities in obesity rates between women and men. It shows that female empowerment leads to lower gender obesity gaps in a worldwide sample of countries, but that this effect is entirely driven by the MENA region. The fourth paper focuses on peer effects and social learning. It assesses the impact of a behaviour change campaign to reduce child malnutrition in Mozambique. The paper shows that the programme did not only improve nutritional practices among the programme’s participants, but also among untreated neighbours, suggesting the presence of social learning effects