Relations between socio-economic factors and nutritional diet in Vietnam from 2004 to 2014: New insights using compositional data analysis.

This paper contributes to the analysis of the impact of socioeconomic factors, like food expenditure level and urbanization, on diet patterns in Vietnam, from 2004 to 2014. Contrary to the existing literature, we focus on the diet balance in terms of macronutrients consumption (protein, fat and carbohydrate) and we take into account the fact that the volumes of macronutrients are not independent. In other words, we are interested in the shares of each macronutrient in the total calorie intake. We use compositional data analysis (CODA), adapted to deal with the relative information contained in shares, to describe the evolution of diet patterns over time, and to model the impact of household characteristics on the macronutrient shares vector. We compute food expenditure elasticities of macronutrient shares, and we compare them to classical elasticities for macronutrient volumes and total calorie intake. The compositional model highlights the important role of many factors in the determination of diet choices and we will focus mainly on the role of food expenditure. Our results are consistent with the rest of the literature, but they have the advantage to highlight the substitution effects between macronutrients in the context of nutrition transition.

Decomposition of changes in the consumption of macronutrients in Vietnam between 2004 and 2014.

Vietnam is undergoing a nutritional transition like many middle-income countries. This transition is characterized by an increase in per capita total calorie intake resulting from an increase in the consumption of fat and protein while the carbohydrate consumption decreases. This paper proposes to highlight the sociodemographic drivers of this transition over the period 2004-2014, using Vietnam Household Living Standard Survey data. We implement a method of decomposition of between-year differences in economic outcomes recently proposed in the literature. This method decomposes the between-year change in various indicators related to the outcome distribution (mean, median, quantiles, etc.) into (1) the effect due to between-year change in the conditional distribution of the outcome given sociodemographic characteristics, or "structure effect", and (2) the effect due to the differences in sociodemographic characteristics across years, or "composition effect". In turn, this last effect is decomposed into direct contributions of each sociodemographic characteristics and effects of their interactions. The composition effect, always positive, generally outweighs the structure effect when considering the between-year changes in distributions of per capita calorie intake or calorie intake coming from protein or fat. The effects of changes in the composition of the Vietnamese population thus overcome the effects of changes in preferences of the same population. This finding is reversed in the case of carbohydrates. Food expenditure and household size appear to be the main contributors to the composition effect. The positive effects of these two variables explain well most of the between-year shifts observed in the calorie intake distributions. Urbanization and level of education contribute negatively to the composition effect, with the noticeable exception of fat where the effect of urbanization is positive. But these two variables effects are negligible compared to those of food expenditure and household size.

Smooth conditional distribution function and quantiles under random censorship.

We consider a nonparametric random design regression model in which the response variable is possibly right censored. The aim of this paper is to estimate the conditional distribution function and the conditional alpha-quantile of the response variable. We restrict attention to the case where the response variable as well as the explanatory variable are unidimensional and continuous. We propose and discuss two classes of estimators which are smooth with respect to the response variable as well as to the covariate. Some simulations demonstrate that the new methods have better mean square error performances than the generalized Kaplan-Meier estimator introduced by Beran (1981) and considered in the literature by Dabrowska (1989, 1992) and Gonzalez-Manteiga and Cadarso-Suarez (1994).