研究生中级计量经济学老师课件

应用计量经济学Instructor:FredEngst(阳和平)(orjustFred)Bestwaytocontactmeisbyemail:[e-mail],[e-mail]第一课EconometricsintheNewsThePurposeofEconometricsTypesofDataTheArtofEconometrics（计量的艺术性）应用计量经济学EconometricsintheNews清华经管学院李宏彬等在《美国科学院院刊》(2013年7月8日)发表论文阐释空气污染对预期寿命的影响研究者收集了1981年到2000年间、淮河南北90座城市每日总悬浮颗粒物浓度的数据，并通过中国疾病监控系统的数据，计算出1991年到2000年上述城市的各年龄段死亡率、预期寿命和死于心肺疾病的数量。结果发现，在中国，随着由南至北的纬度变化，空气污染水平也以复杂的相关关系平滑地变化着。其中，在淮河附近，空气污染水平出现了巨大的“跳跃”。这是由于以淮河为界，中国南北方的供暖政策不同。冬季北方大量燃煤供暖，使得淮河北岸空气中的总悬浮颗粒物比相邻的南岸跳跃性地高出200微克/立方米。图为淮河以北TSP上升247.5微克/立方米这份研究发现，长期暴露于污染空气中，总悬浮颗粒物（TSP）每上升100微克/立方米，平均预期寿命将缩短3年。图为淮河以北预期寿命下降5年。按照北方地区总悬浮颗粒物的水平，这意味着中国北方5亿居民因严重的空气污染平均每人失去5年寿命，污染的代价巨大。ThePurposeofEconometrics从数据中找规律性的关系Typesofdata:Tocorrectlyidentifythetypesofdata,weneedtounderstandthedifferencebetweenentityandvariable,inwhicheachentityhasinfinitevariables.Cross-sectionaldata(截面数据)e.g.recordingsofeverystudent’sweightfortodayTimeseries(时间序列)e.g.theweightrecordofapersonoverayear.Paneldata(面板数据)—thecombinationofcross-sectionandtimeseriese.g.theweightrecordsofallthestudentshereforeachdayandforoverayear.Theidentificationofdatatype:Q:thedatainthepublished2012StatisticalAbstractofChinaistypicallyofwhatkind?Q:Whatkindofdataisthepublishedstockmarketactivities?地区年均人数（万）平均劳动报酬北京51439,684天津19527,628河北50116,456山西36618,106内蒙古24318,382辽宁49819,365吉林26616,393黑龙江49715,894上海33337,585江苏67923,657浙江61127,570安徽33817,610福建42719,424江西28315,370山东89819,135A)CrossSectionB)TimeSeriesC)PanelD)NotSureWhatkindofdataisthis?Whatkindofdataisthis?年GDP＝（养殖业＋制造业＋其他）19783645.21018.41745.2881.619804545.61359.42192.0994.219859016.02541.63866.62607.8198610275.22763.94492.73018.6198712058.63204.35251.63602.7198815042.83831.06587.24624.6198916992.34228.07278.05486.3199018667.85017.07717.45933.4A)CrossSectionB)TimeSeriesC)PanelD)NotSure收入张三李四。。。王麻子2000年2314543251652342001年253894623967341。。。2009年301255239570128A)CrossSectionB)TimeSeriesC)PanelD)NotSureWhatkindofdataisthis?ATruePanelvs.APooledCrossSectionOftenlooselyusethetermpaneldatatorefertoanydatasetthathasbothacross-sectionaldimensionandatime-seriesdimensionMoreprecisely,atruepaneldataisonlydatafollowingthesamecross-sectionunitsovertimeOtherwiseit’sapooledcross-sectionHowisthisdifferfromaPanelData?食堂伙食评价学生1学生2。。。学生982000397学生A学生B学生C。。。学生10520018576。。。。。。学生甲学生乙学生丙学生丁。。。学生79200959362A)CrossSectionB)TimeSeriesC)PanelD)NotSureWhatkindofdataisthis?The“Art”ofEconometricsTheArtofEconometricsAgeneralapproachtovariableselectionandmodelspecification:(回归变量的选择)Specifya“base”or“benchmark”model.(基准设定)Specifyarangeofplausiblealternativemodels,whichincludeadditionalcandidatevariables.Doesacandidatevariablechangethecoefficientofinterest(1)?(改变因子集对基准设定的影响)Isacandidatevariablestatisticallysignificant?Usejudgment,notamechanicalrecipe…变量的选择应是以理论为指导的，而不是机械的CrossSectionalExample1LetsworkonCATestScoreDatasetVariables:DIST_CODE:DISTRICTCODE;READ_SCR:AVGREADINGSCORE;MATH_SCR:AVGMATHSCORE;COUNTY:COUNTY;DISTRICT:DISTRICT;GR_SPAN:GRADESPANOFDISTRICT;ENRL_TOT:TOTALENROLLMENT;TEACHERS:NUMBEROFTEACHERS;COMPUTER:NUMBEROFCOMPUTERS;TESTSCR:AVGTESTSCORE(=(READ_SCR+MATH_SCR)/2);COMP_STU:COMPUTERSPERSTUDENT(=COMPUTER/ENRL_TOT);EXPN_STU:EXPENTITURESPERSTUDENT($’S);STR:STUDENTTEACHERRATIO(ENRL_TOT/TEACHERS);EL_PCT:PERCENTOFENGLISHLEARNERS;MEAL_PCT:PERCENTQUALIFYINGFORREDUCED-PRICELUNCH;CALW_PCT:PERCENTQUALIFYINGFORCALWORKS;AVGINC:DISTRICTAVERAGEINCOME(IN$1000'S);Example1:Whatdodatasayabouttheclasssize/testscorerelation?TheCaliforniaTestScoreDataSetAllK-6andK-8Californiaschooldistricts(n=420)Variables:5thgradetestscores(Stanford-9achievementtest,combinedmathandreading),districtaverageStudent-teacherratio(STR)=no.ofstudentsinthedistrictdividedbyno.full-timeequivalentteachersQ:Whatkindofdataisthis?Initiallookatthedata:(Youshouldalreadyknowhowtointerpretthistable)Istherearelationship?Scatterplotoftestscorev.student-teacherratioDodistrictswithsmallerclasseshavehighertestscores?Whatdoesthisfigureshow?VariableswewouldliketoseeintheCaliforniadataset:(我们希望得到的变量)Schoolcharacteristics:(学校因素)student-teacherratio(学生-教师比)teacherquality(教师质量)computers(non-teachingresources)perstudent(教学设备)measuresofcurriculumdesign…(教材质量)Studentcharacteristics:(学生因素)Englishproficiency(英语程度)availabilityofextracurricularenrichment(课外活动)homelearningenvironment(家庭环境)parent’seducationlevel…(家长教育水平)VariablesactuallyintheCaliforniaclasssizedataset:(我们实际得到的)student-teacherratio(STR)(学生-教师比)percentEnglishlearnersinthedistrict(PctEL)(英语程度)percenteligibleforsubsidized/freelunch(领得到午饭补贴的%)percentonpublicincomeassistance(得到政府贫困救济的%)averagedistrictincome(教育区的人均收入)AlookatmoreoftheCaliforniadata学英语者%午饭补贴的%贫困救济的%Digression:presentationofregressionresultsinatable(展示多个回归结果的方法)Listingregressionsin“equation”formcanbecumbersomewithmanyregressorsandmanyregressionsTablesofregressionresultscanpresentthekeyinformationcompactlyInformationtoinclude:variablesintheregression(dependentandindependent)（变量）estimatedcoefficients(系数)standarderrors(标准误)resultsofF-testsofpertinentjoin