Beyond Industrial Classification: The Four-Layer Functional Model and a New Coordinate for Economic Structure Analysis

Abstract：Mainstream economic classifications divide economic activities according to final product form or industrial sector affiliation. The dispute over “Factoryless Goods Producers” (FGP) in statistical circles has already exposed the distortion of this classificatory coordinate: Apple, Nvidia, Nike, and similar firms are counted as manufacturing, although they themselves do not engage in large-scale manufacturing. However, the focus of the FGP literature is “whether such firms should be classified into manufacturing or wholesale”—that is, finding a new cell within the old coordinate, without questioning the coordinate itself. Research on functional specialisation in the trade literature has demonstrated the feasibility of dividing economic activities by function, but its object of division is occupations and trade flows; it has not extended to firm classification and national economic structure. This paper connects the two literatures and proposes a functional classification framework for economic structure analysis—the Four-Layer Functional Model: instead of dividing by final product form, it divides economic activities according to the form of their core output and their functional position in the production system into Basic Resource Function (V1), Hard-Entity Production Function (V2), Soft-Entity Production Function (V3), and Service Realization and Organization Function (V4). It provides executable classification rules based on the principle of “one’s own core output,” rules for handling composite types, and rules for boundary cases. Exploratory applications at the firm, industry, and national levels show that the framework can identify systematic classification mismatches invisible under traditional classification. The paper further derives five falsifiable propositions from the framework and provides a staged mapping scheme from firm level to national level. This paper is a theory-building study; cases serve to illustrate classification and explain mechanisms, not to perform statistical inference.