方法对比

	Demand System Estimation	Discrete Choice Demand Model
领域	经济学	经济学
方法族	Regression model	Regression model
起源年份	1954	1995
提出者	Richard Stone (linear expenditure system); developed by Deaton, Muellbauer, Theil, Barten	Daniel McFadden (logit); Berry, Levinsohn & Pakes (random-coefficients aggregate demand)
类型	System of structural demand equations estimated jointly	Characteristics-based discrete-choice model of demand for differentiated products
开创性文献	Stone, R. (1954). Linear expenditure systems and demand analysis: an application to the pattern of British demand. The Economic Journal, 64(255), 511–527.	McFadden, D. (1974). Conditional logit analysis of qualitative choice behavior. In P. Zarembka (Ed.), Frontiers in Econometrics. Academic Press. ISBN: 9780127761503
别名	Consumer Demand System, System of Demand Equations, Complete Demand System, Demand System Analysis	Discrete Choice Demand, Random-Coefficients Logit Demand, BLP Demand Model, Characteristics-Based Demand Model
相关	3	3
摘要	Demand system estimation jointly models how a consumer or population allocates a budget across a complete set of goods, estimating a system of equations — one per good — that relate each good's expenditure share or quantity to all prices and total expenditure. Unlike a single-equation demand curve, a demand system imposes the cross-equation restrictions implied by consumer theory: adding-up (shares sum to the budget), homogeneity (no money illusion), and Slutsky symmetry (consistency of cross-price effects). Classic functional forms include Stone's Linear Expenditure System, the Rotterdam model, and the Almost Ideal Demand System, and the system is estimated with seemingly unrelated regression or full-information methods.	Discrete-choice demand models estimate the demand for differentiated products — cars, cereals, computers — by modeling consumers as choosing the single product that maximizes their random utility, where utility depends on the product's observed characteristics and price plus an unobserved quality term and an idiosyncratic taste shock. Aggregating individual choice probabilities yields predicted market shares, which are matched to observed shares to recover preference parameters. The framework spans the simple multinomial and nested logit of McFadden to the Berry-Levinsohn-Pakes (BLP) random-coefficients model that uses aggregate market data, allows flexible substitution, and instruments for price endogeneity.