An almost ideal demand system analysis of non-durable consumption categories

Abstract

This is a theoretical and empirical exploration into the estimation of demand systems based on quarterly and annual macroeconomic data from the National Accounts for Norway. This data set provides a unique source to portray the composition of household consumption and the changes in it over time. The starting point for the study − which the title is also alluding to − is the AID system due to Deaton and Muellbauer (1980), which pays heed to microfoundations as the preferences underlying the demand system allow exact aggregation across households. The report analyzes and compares various versions of the AID system: its basic static version, its linear approximation and dynamic versions with and without habit formation. The report also gives a comprehensive review of the existing literature in this field. An econometric model is specified and each version of the AID system is tested for a set of overidentifying restrictions by means of system methods, e.g. homogeneity, Slutsky symmetry and downward sloping compensated demand functions (also known as negativity conditions). These restrictions are frequently rejected, which is in line with previous findings. In a final chapter the forecasting performance of the preferred model is evaluated − again adopting advanced econometric tests. The macroeconometric models of Statistics Norway – MODAG and KVARTS – contain a somewhat simpler demand system (LES) as an integral part of the consumption block of the models, see section 5.4 in Boug and Dyvi, eds. (2008). This study goes beyond the scope of the LES and gives a backdrop towards which one can evaluate the incumbent system. It is an important source of information about the size and stability of demand and price elasticities and supplements in this respect a previous study on Norwegian data by Raknerud, Skjerpen and Swensen (2007) published in Empirical Economics.