■ Financial Econometrics Chapter 3.Causal Inference framework 1 0 Rubin Causal Model Jin Ling School of Finance,Zhongnan University of Economics and Law
Financial Econometrics Chapter 3. Causal Inference framework 1 Rubin Causal Model Jin Ling School of Finance, Zhongnan University of Economics and Law 1
Outline ·Causal effect 。Rubin Causal Model The Applications for Rubin Causal Model 2
• Causal effect • Rubin Causal Model • The Applications for Rubin Causal Model 2 Outline
Causal effect 。Cause of effect:: ·Economic growth 。Stock return. Investment efficiency ·Effect of cause: The effect of short sell on price efficiency. The effect of national team rescue on market stability. The effect of stimulus program on firm investment. The relationship between cause of effect and effect of cause. Financial econometrics mainly focuses on cause of effect or effect of cause? 3
• Cause of effect: • Economic growth. • Stock return. • Investment efficiency. • Effect of cause: • The effect of short sell on price efficiency. • The effect of national team rescue on market stability. • The effect of stimulus program on firm investment. • The relationship between cause of effect and effect of cause. • Financial econometrics mainly focuses on cause of effect or effect of cause? 3 Causal effect
Causal effect ·Causal effect: The effect of treatment on outcome variables. The difficulty to explore causal effects: 表1.】Yule-Simp50n悼纶 总样本 生还 死亡 生还率 干预组 20 20 50% 控制组 16 24 40% 男性样本 生还 死亡 生还率 干预组 18 -12 60% 控制组 3 70% 女性样本 生还 死亡 生还率 干预组 8 20% 控制组 9 30% 表1.2伯克利分校的研究生录取率 男生 女生 专业 申请人数 录取数 录取率 中请人数 录取数 录取率 专业1 200 40 20% 400 120 30% 黄业2 800 400 50% 100 70 70% 合计 1000 440 44% 500 190 38% 4
• Causal effect: • The effect of treatment on outcome variables. • The difficulty to explore causal effects: 4 Causal effect
Causal effect Credibility revolution: Rubin Causal Model (Neyman-Rubin Model). Randomized experiments. THE SVERIGES RIKSBANK PRIZE IN ECONOMIC SCIENCES IN MEMORY OF ALFRED NOBEL 2021 David Joshua Guido Card D.Angrist W.Imbens “for his empirical "for their methodological contributions to labour contributions to the analysis economics of causal relatorships THE ROYAL SWEDISH ACADEMY OF SCIENCES 5
• Credibility revolution: • Rubin Causal Model (Neyman-Rubin Model). • Randomized experiments. 5 Causal effect
Causal effect The characteristic of credibility revolution: ·Experiment design. Randomized experiments as foundations. Model potential outcomes instead of observable outcomes. Separate potential outcomes and assignment mechanism. Data illustrates causal effect without any artificial intervention. 6
• The characteristic of credibility revolution: • Experiment design. • Randomized experiments as foundations. • Model potential outcomes instead of observable outcomes. • Separate potential outcomes and assignment mechanism. • Data illustrates causal effect without any artificial intervention. 6 Causal effect
Causal effect The limitation of credibility revolution: Randomized experiments is hard to reach in reality. The conflicts between the requirement of randomized experiments and the purpose of economic and financial researches. Quasi natural experiment. Not limited to strict randomized experiments. Revolution or Burden? 7
• The limitation of credibility revolution: • Randomized experiments is hard to reach in reality. • The conflicts between the requirement of randomized experiments and the purpose of economic and financial researches. • Quasi natural experiment. • Not limited to strict randomized experiments. • Revolution or Burden? 7 Causal effect
Causal effect The debate about credibility revolution: THE SVERIGES RIKSBANK PRIZE IN ECONOMIC SCIENCES IN MEMORY OF ALFRED NOBEL 2019 Abhijit Esther Michael Banerjee Duflo Kremer "for their experimental approach to alleviating global poverty" THE ROYAL SWEDISH ACADEMY OF SCIENCES
• The debate about credibility revolution: 8 Causal effect
Causal effect The debate about credibility revolution: 但是，很多学者也担心.仅关注"小问题可能并下能从根本上解决盆国问题。世界银行经济学家Rvi05指出，回答小问题对于填补重大的知识空白和服务政府决策作用有限并且如果存在盆困陷阱依据对 小问题研究得出的结论作决策可能只是让穷人陷入低水平均衡。要真正解决大多数人的贫国问题，还是需要从制度、经济发展等大问题着手。一个典型的例子是如何解释我国和其他国家与地区在减蓝进程上 的差异451。过去30多年尽管世界盆困人口显善减少，但是减贫进程却存在巨大的区域和国别差异。我国是世界上减盆人口最多并率先实现联合国千年发展目标的国家，但是，在撒哈拉以南的非洲国家 盆困的发生率仍然很高，并且时有反弹「12】。为什么有的国家成功减盆？有的国家却进展缓慢？很多经济学家都认为，我国的政治经济制度、国家治理能力以及在1970年代末开始的改革开放带来的经济发展无疑对 减贫贡献很大但这些因素都不是三位诺奖得主关注的”小问题的范斯。 9
• The debate about credibility revolution: 9 Causal effect
Causal effect The debate about credibility revolution: (二)RCT方法简M 相对于德一类方法RCT的一个主要突被是通过随机分配得到了因甲效应（减处理效应]的无偏估计量，本文侣鉴DuO等人1列举的评估敢称书对学生学业表现的影肩案例来说明莱用RCT方法进行因单效应 估计的基本原阻，￥表示实际现家到的学校的学生的平均成捷用Y表示如奥得到免费教科书，学校的学生的平均成绩用表示设有得到免费敢科书学校的学生的平均成姨。我们感兴爆的是这得免费敢科书 对学校的影响即Y:,Y。但是如前所达我们并不能对同一所学校同时现察到两个结平，对总体而言我们示望知道平均处里效应(Average Tre8 tment Efect,ATE)EYYC,在大样不的博况下，平均处理效 应将收敛于对项目参与者的平均处理效应(Average Treatment E什eion1 ne Treated,ATT这里用D(试1)表示， D=YT得到列免贵教科书了 一印段有得到晚要较科书】 =EYTITJ-EYC C7 =EYTTJ -EYOTJ-EYC]+EYGT] =EYT -YCIT]+(EIYCIT] -EYCCJ) (1) 其中第一项Y一Y军T7表示项目整与者的平均处理蚊应第二项(YT】一YC)度量了选择病误即千横组学校和控制增学校在下挨受于孩时的结果上的差异， RCT的方法款是强过划机分配构研究对象穗机分配到干疫组或者控制绍宋减少或者两除可能的选择性偏娱，常用的惠机分配的方法包括过度招幕后随机分配(0 versubscnptio,题机逐物推进项目（「ndam ized phase-n.组内随机(wt曲in-group randomizat1on和随机鼓励参与(encourag8 nent des5gn等，关于这些方法的详细介绍可以参考Dutlo等人的介绍，国随机分配生成的干预组和控制组在没有干预的 情况下的结果变最理论上是没有显管整异的，即式(1]中的第二项等于0。分配亮成后，干预组的学校接受干预《得到免费教材]，控制组的学校不接受干预设有免费校材)。通过比较两组在干预后均值的差异可以得 到改策或者项日的因甲效应的无偏估计量（试(2》. D=EIY.IT-EIY.IC] (2) 10
• The debate about credibility revolution: 10 Causal effect