计量经济学及工具变量.pptx

单击此处编辑母版标题样式,单击此处编辑母版文本样式,第二级,第三级,第四级,第五级,单击此处编辑母版标题样式,单击此处编辑母版文本样式,第二级,第三级,第四级,第五级,单击此处编辑母版标题样式,单击此处编辑母版文本样式,第二级,第,三级,第,四级,第,五级,计量经济学及工具变量,计量经济学及工具变量,1,、计量经济学是用来干什么的？,评价经济政策、观察效果、验证经济理论、寻找逻辑关系、预测,如何评价？如何检验？如何预测？,总体：,样本：,2,、计量中我们要做什么？,A,、我们对某个或几个变量的回归系数感兴趣,B,、写出计量模型：模型设定,变量取舍,C,、得到这些系数的无偏、有效、一致估计量,3,、什么因素会导致,OLS,回归无法获得无偏估计量,？,Endogeneity,2007,年,5,月,22,日,现代快报,：,有教育界人士指出，高中阶段学生的抽象思维能力和议论能力不强，写作为时动不动就“啊”“呀”地抒情，很难写出非常有条理、很严谨的议论，这与基础教育中现有教师队伍的性别构成偏向与女性有关2007,年,5,月,23,日,青年报,：,学生思辩能力弱怎能全怪女教师,作为中学语文教师，笔者对这种“动辄就抒情”的文风深有同感，但对于“女教师比例过高导致抽象思维能力差”的说法却不敢苟同。

孩子的成长过程中女教师比例偏高,有可能造成学生性格上和意志能力上的缺陷，但女教师多却与学生抽象思维能力和议论能力的形成没有必然的联系：一个人的抽象思维能力和议论能力强不强，除了天赋和性别的因素，也与后天影响有关学生的阅读面及社会思辩氛围对学生抽象思维能力的提高具有很大的促进作用而我们目前的教育机制偏重于培养学生的形象思维能力，从小学到初中，一直训练学生写记述文，训练学生的情感表达和抒发学生的课外阅读，也自然偏重于趣味性比较强的“形象阅读”更主要的是，社会没有给孩子提供利于思辩能力提高的氛围譬如电影电视，过多强调引人入胜的故事性和情感性，即使是说理性的电视节目，也要借助于形象思维作媒介像央视最火的“百家讲坛”，实际上是将说理给故事化和趣味化了，这在一定程度上会转移学生对理性思辩的注意度因此，学生思维能力的提高，依赖于教育者尤其是语文教师的针对性训练，社会理性氛围的形成也至关重要少年不识愁滋味，爱上层楼爱上层楼，,为赋新词,强说愁,而今识尽愁滋味，欲说还休欲说还休，却道新,(,天,),凉好个秋Whats Endogeneity?,Gauss-Markov Theorem:,A1:Linear in parameters,A2:Random sample,A3:Zero conditional mean,E(u|x)=0,A4:Sample variation in the independent variable,(unbiasedness of OLS),A5:Homoskedasticity,(BLUE of OLS),when,an independent variable is correlated with error term,then zero conditional mean assumption does not hold,it is said to be an,endogenous variable,;,A.Omitting variables,B.Measurement error,C.Simultaneity,Omitting an important factor that is correlated with any of will cause endogeneity;,What we have learnt on endogeneity caused by omitting variables?,A.,Proxy,B.,FD,C.,FE,Only when,panel data,is available,FD and FE can be used to estimate the effects of,time-varying independent,variables in presence of,time-constant omitted variables,or we only have cross sectional data,But if we have,time-varying omitted variables,or we are interested in the effect of,time-constant variables,or we only have cross sectional data,Ch15.Instrumental Variables,I.Introduction,1.Whats IV for endogenous variables?,(1),where ability is omitted,and ability is correlated with education,so,edu,is an endogenous variable;,(2),(3),We call,z,an IV for,edu,z,is,exogenous in equation(1),.,A,.IV should has no partial effect on dependent variable;,(IV,不能对,Y,产生直接影响,),B,.IV should be related to the endogenous independent variable;,(IV,能够对内生变量产生直接影响,),C,.Assumption(2)cannot be tested,yet we can test assumption(3)by regressing endogenous variable(eg.edu)on IV in reduced form equation:,2.Why do we need IV?,From equation,We can get,Then,we can get the IV estimator of,i.e.,If,z=x,we can see that IV estimator is simply the OLS estimator;,In terms of population correlations and standard deviation,we have the probability limit of IV estimator,where and are standard deviations of u and x in the population,So,we can see from this equation that:,A,.,Even if Corr(z,u)is small,the IV estimator will be inconsistent if Corr(z,x)is also small;,B,.,According to the law of large number,the IV estimator is consistent,provided that assumptions for IV are satisfied.,In,small samples,the IV estimator can have a substantial bias,so large samples are preferred when IV method is adopted.,C,.,IV is preferred to OLS,on asymptotic bias grounds when,Corr(z,u)/Corr(z,x),Corr(x,u),plim of OLS estimator:,3.How to find an IV?,Generally,we need to find an,exogenous,variable which has,no direct effect,on dependent variable,is,not related to,the omitted variable,and is,correlated,with the endogenous variable.,4.Some examples,IV for education(ability is omitted),A,.The last digit of an individuals identity card number(bad IV),B,.Mothers or fathers education(not good enough),C,.Number of siblings(it looks good),IV for,skipping classes,on,final exam score,(ability is also omitted),(4),IV:,distance,between living quarters and campus,distance is,exogenous,in equation(4),5.More examples for IV:,A,.,IV for education when ability is omitted,A dummy variable that is equal to 1 if a man is born in the first quarter of the year,(,Angrist and Krueger,1991,QJE,Vol.106).,B,.,Acemoglu,Johnson,and Robinson(NBER Working paper 7771,2002),Question:Why Africa and Australia,New Zealand,Canada,US has different performance?,Omitting Variable Bias,Employing,mortality rate,as,IV for institutions,High mortality rate not suitable for settle extractive state worse institution,Low mortality rate suitable for settle Neo-Europes better institution;,“,坐寇”与“流寇”,Hall and Jones(QJE,Vol.114,1999):,Why does output per worker vary enormously across countries?,Used the,distance from the equator,and the,extent to which the languages of Western Europe are spoken as a mother tongue,as IV for infrastructure.,a.,Western Europe expansion from the sixteenth to nineteenth century has influence around the world;,language,(English,Spanish,French,German,Portuguese),b.,Western Europeans were more likely to settle in areas that were broadly,similar in climates to Western Europe,which was far from equator,;,c.,Wes。