这已经不是什么秘密了，前沿实证经济研究中reduced form model的地位每况愈下，取而代之的是structural model. 找到感兴趣的X和Y，分别放在回归式两端做回归的方式，在经济学意义上受到广泛质疑；甚至一般的IV方法在structural modeling的角度看来也难以将模型的阐释意义提升。

下文是斯坦福大学P E T E R C . R E I S S & F R A N K A . WO L A K 合作的一章，主要利用Industrial Organization的研究例子，介绍传统reduced form的缺陷以及结构模型的定义、方法，对于想要学习结构模型方法的同学想必是很有用的。下面引用少许段落：

一个可靠的结构模型基本的要求：
(1) flexible statistical descriptions of data; (2) respectful of the economic institutions under consideration; and, (3) sensitive to the nonexperimental nature of economic data. When, for example, there is little economic theory on which to build, the empiricist may instead prefer to use nonstructural or descriptive econometric models.

结构模型有特定的适用条件，与reduced form最大的区别在于对经济理论直接的阐释，但是在研究方法的须谨慎选取。Understanding when and why structural modelers must make compromises, and that structural modelers can disagree on compromises, is important for understanding
that structural modeling is in part “art”.

文中对于IV的解释也许会改变你固有的了解，会让你意识到一般的IV选择方法是多么草率。“Look for an exogenous variable that is uncorrelated with the ’s but at the same time correlated with the right-hand side endogenous variable y2”. While these two approaches are not necessarily incompatible, the second approach does not seem to involve any economics. (This should sound a warning bell!)