关键词: 消费差异, 多阶模型, 场景变量

Abstract: When the differences between groups of samples are significant, we should use multilevel models and introduce contextual variables. This paper explores the 1600 households’ consumption data, including 11 cities, 7 counties and towns in Guangdong province. The conclusions are that significant differences exist in the consumption level, marginal propensity to consume among different cities, counties and towns. This paper introduces per capita disposable income of a family (PDI) and whether having a house (HOUSE) in micro-lay to explain the consumption difference among households, and introduces the status of development (D) and per capita disposable income (GPDI) as contextual variables to explain the consumption difference among different cities and countries. Empirical study shows that variables in micro-lay can explain more than 48 percent about intra-class consumption difference and the two contextual variables can explain more than 99 percent about the between-group difference.

Key words: Difference of Consumption, Multilevel Models, Contextual Variables