因果推理中的科学模型--反事实、选择性偏差与赫克曼结构计量经济学模型

A Scientific Model of Causality

因果推理中的科学模型起源于赫克曼对于经济学中样本选择性偏差的研究,它代表了经济学对于因果关系的理论思考,也是对统计学提出的因果理论的回应。为了解决选择性偏差问题,赫克曼没有像统计学一样试图通过随机实验而使得样本选择对于干预结果没有影响,而是求助于科学研究,去发现导致选择性偏差的真实原因,对偏差进行估计,并且将这一选择性偏差模型化,形成了有关选择的模型和结果的模型的两个模型。这种将选择与结果分别建立模型,并且强调它们之间内在逻辑关系构成了因果推理的结构计量经济学路径,也是因果推理中科学模型的核心思想。赫克曼的科学模型不仅对于社会科学研究有非常重要的意义,而且对于政策评估有突出的价值。更多的因果模型不应该是非此即彼的选择,而应该是在相互竞争中共同学习、共同成长和共同演化。

The scientific model of causality originated from the research of the sample selection bias by Heckman.It was a representative model of economics,which is different from the causal inference in statistics.To solve the selection bias problems,statistics proposed the randomization to reduce the influence of assignment mechanisms.Not from the statistical perspective for causality,Heckman tried to find out the reasons of selection bias and model the selection bias from the perspective of scientific theories.Finally Heckman propose an econometric approach or structural econometric approach to the causality,which comprises the selection equations and outcome equations.Heckman's causality is not only very important for social sciences but also for policy evaluation.We should not make“one best choice”ong many models of causality.On the contrary,different models could serve different purposes and they can also learn from each other,co-development and coevolution.