Abstract: The identification of heterogeneous consumers provides a reference for the formulation and evaluation of macroeconomic policies aimed at releasing the consumption potential of residents. However, in the current related studies, there is no objective automatic identification of heterogeneous consumers through the characteristics of data. In this paper, based on the two-stage consumption decision model, a panel structure model with group structure is constructed using the micro data form China Family Panel Studies (CFPS). Based on the C-Lasso method, information criterion is introduced to verify the existence of group structure, and the number of groups and tuning parameters are determined, so as to realize the automatic identification of heterogeneous consumers according to the characteristics of data. The empirical results show that: (1) Chinese consumers can be divided into two groups. Compared with the traditional method and the clustering method in machine learning, the information criterion value of the proposed method is smaller, indicating a higher accuracy in identifying heterogeneous consumers. (2) There are differences in the effects of temporary income changes and household assets on consumption among different groups of consumers, shows that the identification method is scientific and objective. For the group with higher household assets, temporary income change and illiquid assets significantly promote consumption, but liquid assets to inhibited consumption; The temporary income changes and illiquidity assets of the group with lower household assets significantly inhibited consumption, while liquid assets promoted consumption. (3) There are significant differences in household assets, education level and urban and rural categories among different groups of consumers. The group with higher household assets has longer years of education and has a higher proportion of urban household registration. The group with lower household assets has shorter schooling years and had a higher proportion of rural household registration. (4) Different liquidityof household assets allocation may be the cause of consumer heterogeneity. A new idea and method for automatic identification of heterogeneous consumer group structure is proposed, which provides a reference for the formulation and evaluation of macroeconomic policies to release the consumption potential of different types of consumers.

Key words: panelstructuremodel; C-lasso; groupstructures; heterogeneousconsumers; householdasset liquidity