如何理解宏观经济学中的校准（calibration）？

以前上 Minnesota macro 的时候做过几道跟校准（calibration）有关的习题，班门弄斧两句。这个东西其实一开始是 Ed Prescott 搞出来的，简单来讲就是用现实中的数据（矩）来拟合理论上（parsimonious and misspecified）的模型，从而获得一些经济学上的 insights。

校准的思路跟结构计量经济学很像*，不同的地方是 Prescott 彻底放弃了数理统计这条路，而是采用了“毛估估”（eyeballing）的方式来评判模型与数据拟合的好坏程度，所以校准得到的结果通常都不会附有标准误一类的统计量。

~~~~~~~~~~~~~~~~~

* Dawkins et al. (2001):

Calibration is estimation, estimation is calibration... If calibration is the setting of the numerical values of model parameters relative to the criterion of an ability to replicate a base case data set as a model solution, and estimation is the use of a goodness of fit criterion in the selection of numerical values of model parameters, the two procedures are closely related. In both cases a selection of model parameter values which is thought to be reasonable (or best) relative to some criterion applied to data is involved. In one sense, both procedures lead to identical outcomes.

(cf. Christina Dawkins, T.N. Srinivasan, John Whalley, "

," In: James J. Heckman and Edward Leamer, Editor(s), Handbook of Econometrics, Elsevier, 2001, Volume 5, Pages 3653-3703.)

~~~~~~~~~~~~~~~~~

(Source: Heller-Hurwicz Economics Institute)

相关历史信息和更详细的科普可参见 Kevin D. Hoover 的这篇文章：

。