Empirical Puzzles or Aggregation Problems - A View of Agent-based Model

Traditionally, most literatures supposed that the interactions between heterogeneous individuals had relatively limited influence on explaining macroeconomic phenomena, hence simplified the whole economy into only representative agent, or even its multiple. For recent decades, notwithstanding the representative agent model has remarkably progressed in its theorem completeness and mathematics algorithm, some empirical puzzles remains unsolved. Taking Hall(1978) ,the random walk of consumption model, as an example, there still has some debates in excess sensitivity. Therefore, in this paper, we desire to figure out two principal potential drawbacks of Hall’s model as followed: (1) the official consumption is often calculated according to the expenditure amounts, instead of total consumption, (2)If agents are heterogeneous and interactive, we can’t use central limit theorem(CLT) rashly. In other words, the standard econometric procedures (e.g. cointegration, Granger-causality, impulse-response functions of structural VARs) may generate spurious evidence in aggregate equations. Nevertheless, agent-based model simulation could be a data-generation mechanism (DGM), it not only can help us avoid using imperfect data but also help us examine the behavior of various econometric methods to see whether they actually well behave when the data is indeed the aggregation over interacting heterogeneous bounded rational individuals.