实验八内生解释变量问题

《实验八内生解释变量问题》由会员分享，可在线阅读，更多相关《实验八内生解释变量问题（7页珍藏版）》请在金锄头文库上搜索。

1、实验八 内生解释变量问题实验目的】掌握工具变量法、两阶段最小二乘法的使用以及内生性检验与过度识别约束检验实验内容】利用美国各州的数据为样本观测值，建立香烟需求模型（见教材138页例 4.3.4）实验步骤】表4311995年美国48个州人均香烟消费、收入与对香烟的课税地区人均香烟消 费Q （盒）人均收入Y （千美兀）香烟平均销售价 格P （美分/盒）香烟平均消费税Tax （美分/盒）香烟平均的特别消费 税Taxs （美分/盒）AL101.0912.92103.9226.570.92AR111.0412.17115.1936.425.49AZ71.9513.54130.3242.876.21CA5

2、6.8616.07138.1340.039.04CO82.5816.32109.8128.870.00CT79.4720.96143.2348.568.11DE124.4716.66108.6631.500.00FL93.0715.43123.1737.996.97GA97.4714.59102.7423.620.94IA92.4013.90125.2639.375.96ID74.8512.88117.8734.125.61IL83.2716.83130.2344.627.37IN134.2614.33101.4025.924.83KS88.7514.36114.9731.505.47KY17

3、2.6512.6195.7917.725.42LA105.1812.82110.1028.874.23MA76.6218. 41142.4649.216.78MD77.4717.65122.0739.375.81ME102.4713.28129.4240.037.33MI81.3915.73158.0464.968.95MN82.9516.13144.5947.249.46MO122.4514.50103.1726.900.91MS105.5811.28111.0427.567.26MT87.1612.31102.5027.560.00NC121.5414.4098. 4219.033.79N

4、D79.8112.52126.1544.627.14NE87.2714.56119.5438.065.69NH156.3416.41109.3432.150.00J80.3719.21133.2641.997.54NM64.6712.37115.5829.535.50NV93.5316.93135.5638.718.87NY70.8218.19145.5852.495.60OH111.3815.02108. 8531.505.18OK108.6812.73111.6430.845.32OR92.1614.87124.8740.680.00PA95.6415.58115.5936.096.54R

5、I92.6015.78147.2852.499.64SC108.0812.78100.2720.344.77地区SDTNTXUTVAVTWAWIWVWY人均香烟消费Q （盒）97.22122.3273.0849.27105.39122.3365.5392.47115.57112.24人均收入Y千美兀）13.0214.3014.1212.3716.0514.0215.6714.8111.7514.12香烟平均销售价 格P （美分/盒）110.26109.62130.05118.75109.36115.25156.90132.14109.26104.03香烟 Tax30.8424.2842.653

6、3.1417.3928.8752.8240.6826.9023.624.248.127.365.655.215.4910.266.296.180.00资源来源：根据 Introduction to Econometrics (2ndedition) 整理。根据商品需求函数理论，对香烟的人均消费需求Q与居民的收入水平Y及香烟的销售价格P有 关；虑到在市场均衡时香烟的销售价格也同时受香烟的需求量的影响，则Q与P之间存在着双向因 果关系， P 为内生解释变量，考虑到香烟价格中包含政府对烟草的课税，而香烟的人均消费量本身 不会直接影响政府对香烟的课税政策，因此香烟的消费税可能是价格的一个适当的工具变量

7、。一、建立对数线性模型：In Q二B +P In Y +P In P +卩0 1 2对原模型做回归分析结果如下：| Proc | Object | | Print | Na m e | Freeze Esti m ate Fore cast Stats Re d |Dependent记1曰口1七：LNQI?1 eth dl_S quare sD ate: 10/26/1 6 Time: 1 ：56Sample: 1 d!8includedi o&servsitlons:Va riauieCo eTTl cl e ntstd. Errort-statisticPro&.C10.340941.02

8、263-310.112070-0000l_N0.344032 .S34JS8 31.4046960.1 500LNP-1.40E376 .25129 4-5-. 59 6 5260.0000R-squared0.432759Mean dependent var1 o t M :n）QiLtT .THL. L.i mL ted. I itf osmat i :-tl Mut i mum L.i kLi hood.and. ZKCl a=匚Hm-RZG 匚 c a. n-L e- k-Li g占 Rm 舌*- a s s i dkiJLRjCH A-uX or g icr s x veC jl d

9、i i :-rLal Hq t git o sk et i e iEITTARYCHc-i -34 CLoei t. Fr obi t. Ercti-em*口RZ3HHZI一 口匸咼= B a 匚h口1 cs在 Instrument list 中写入工具变量，没有工具变量的直接把解释变量包括常数项写上，有工具变量 的就直接把工具变量写上，如本例lnP的工具变量是tax,如下图，点击确定结果如下图：=Equation; UNITTLED Workflle: UNTITLED;EntitledX_ S XI View Proc Object Print hlam 吧 尸 reeze Estimat

10、e 尸 ore-cast Stats Res IdsDe pendent variable:匸1Meth o d: Two-Stage Least Square sDate: lQ/.2Sfl6 Time: 11 ：26&呂 Eplu 1 48Included observations: -4SIn strum e nt sspeuificsiticiri: C LNY TAXVariableCoefficientStd. Errort-StatisticProb.C10.022961.0B171S9.265775o.oaaoLNY0.2969530.2403541.24300 30.2200

11、L_hJR-14950.270991-.S51072R-squared0.431077Ms an dip-e ndmnty/siF.530839Adjusted R-squared0.+05792S.D. dlep-endent var0.243351S. E. of regression . 1S7BS7Sum, squared resld1.5S34LQBF-stati stic-13.267-12Durbin-Watson slat1.942391Fro b(F-statistic) .000030Second!-Stage SSRi.sgsia-stsiti stic0.000000Instrument rank3有上述结果