Aggregation and simple dynamics
证明，仅从宏观加总数据的动态行为就能完全识别经济个体中科伊克滞后或AR(1)系数的分布，并提供了加总层面的可检验含义。美国消费数据支持部分消费者为随机游走、其余为ARIMA(1,1,0)的假设。
The koyck (geometric) lag or AR(1) specification is a commonly proposed behavioral model, sometimes after differencing. The distribution of koyck lag or AR(1) coefficients across agents in an economy is shown to be completely identified just from the dynamic behavior of aggregate (macroeconomic) data. Aggregate testable implications of an economy composed of agents having koyck lags or AR(1) models are provided. Extensions to higher-order and time-varying lags are discussed. Aggregate U.S. consumption data are shown to support the hypothesis that some consumers have random-walk consumption, while the rest have ARIMA (1, 1, 0) consumption with widely varying AR coefficients. Copyright 1994 by American Economic Association.