三重差分法運行和示例

三重差分法及運行/s?__biz=MjM5OTMwODM1Mw==&mid=2448053255&idx=1&sn=54fe4a6f5a9382df9bdb20f0b1abda83&chksm=b32389298454003f233cda8e53bbbfc6d1554f9482ccc1385952b6739aee41edcf4e22f726d3&scene=0#rd(SyntheticControlMethod)Abadie&Gardeazabal(2003)(14)(Difference-in-differences-in-differences,DDD)( )/

□□□□□□□□□□□□曲事用T由啊dunwiyM，早t= =2沖?沖加可而qris*(T■S)冷力?■1日dswy曜ri就看后日匕時巾礎則S=T=1Alinouighitlanotahownng^or-cugiYhetffl_Illum&ounnaflhB&sbmatatinthlamodelare正如上面的二重差分法實際上運用的是 做的回歸，我們之前說過，倍差法是相當于,里面包括個體效應時間效應，而對于一個這樣的 ，我們可以運用通過添加個體和時間虛擬變量來回歸，或者運用回歸來消除個體和時間效應，再加上那些交互效應后就可以像其他回歸一樣。treatmentcontroltreatmentcontrolhttp://econ.lse.ac.uk/staff/spischke/ec524/evaluation3.pdf1992接下來，我們來區(qū)分一下二重差分和三重差分在4.25的運行過程5.051992接下來，我們來區(qū)分一下二重差分和三重差分在4.25的運行過程5.05CardandKruegerftefulltimeemploymenttreated=1t=11992 111992 2bk=1BurgerKingkfc=1 roys=1RoyRogers wendys=1Wendy'scode Stata讀取數(shù)據(jù):use"http://fmwww.bc.edU/repec/bocode/c/CardKrueger1994.dtaidtreatedftebk_dLff10NJ311aa0a1NJ4a1000120NJ1310000Z1NJ1Z,510a013DNJ12.50100□3 1NJ7.50100-040NJ160ai0□ANJ2口0a14050PJQQ1q051NJ250010060NJ30di0-口6 IIHJ600i00D0NJ■00010B1HJ27.50□01□簡單的二重差分:difffte,t(treated)p(t)

DtFFERENCE- MCE￡口3TTMM10KRE￡ULT5NumberofobservationsintheDIFF-IN-DIFF:801EeforeCentMl:78Treated:326404After1132。397155646Outcomevar.fteB.Err.It|P>lt|BeforeControlT工EtedDiff(T-C)19,94917,065-2.SB41.135-2.540.011**AfterControlTreated.Diff(T-C)17.54217-5730-0301.1430.030.979Diff-in-Diff2.9141.6111+810.071DDDDDDR-square: 0.01.MeanaandStandardErrorsjareestimatedbylinearregreasionInference：土土★p<0.01j**p<0,05;tp<0.1* DD10%簡單的三重差分（快餐品牌 作為第二個處理組）：difffte,t(treated)p(t)ddd(bk)TRIPLEDIFFERENCE(DDD)ESTIMATIONRESULTSNotationofDDD:ControlControlTreatedTreated(A)(B)伍)(B)ControlControlTreatedTreated(A)(B)伍)(B)treated—0treated.=0treated=1treated=1and.andandand.bk=1bk=0bk=1bk=0BeforeAfterControl㈤:343569Control(B):444286Tredted(A):133132265Treated㈤:193188381404397NumberofobservationsintheDDD:801Outcomevar.fteS.Err..|t|P>|t|BeforeControl(A)25,654Control(B)15.540Treated(A)IB.547Treated(B)16,044Diff(T-C)-7.6122.2063.450.001***A