Résumé
Recent research on calorie intake and income relationship abounds with parametric models but usually gives inconclusive results. Our paper aims at contributing to this literature by using recent advances in the estimation of generalized additive models with penalized spline regression smoothing (GAM). These semi-parametric models enable mixing parametric and nonparametric functions of explanatory variables and enlarge the distribution of the response variable. The revealed performance test (Racine and Parmeter, 2014), supported by simulation data, shows that GAM models outperform the parametric models. Using data from CHNS in 2006, 2009 and 2011, we find a positive and statistically significant relationship between household calorie intake and household income for the poor. Then the impact of increasing income on calorie consumption slows down for the middle class and the rich. In addition, we find that income-calorie elasticities are generally small, ranging from 0.07 to 0.12.
Mots-clés
Calorie intake and income; generalized additive models; CHNS data; revealed performance test; cross validation procedure;
Codes JEL
- C14: Semiparametric and Nonparametric Methods: General
- C15: Statistical Simulation Methods: General
- C30: General
- C52: Model Evaluation, Validation, and Selection
Référence
Michel Simioni, Christine Thomas-Agnan et Thi-Huong Trinh, « Calorie intake and income in China: New evidence using semiparametric modelling with generalized additive models », Vietnam Journal of Mathematical Applications, vol. 14, n° 1, 2016, p. 11–26.
Voir aussi
Publié dans
Vietnam Journal of Mathematical Applications, vol. 14, n° 1, 2016, p. 11–26