計量經(jīng)濟學試驗2

1、實驗2-3實驗?zāi)康模篈RMA模型識別及估計與應(yīng)用(ADF檢驗、Q統(tǒng)計量、ACF、PACF)、ECM模型 建模、VAR模型建模檢驗與應(yīng)用、離散選擇模型建模估計與檢驗案例分析案例1中國支出法GDP的ARMA( p,q)模型估計中國支出法GDP是非平穩(wěn)的，但它的一階差分是平穩(wěn)的，即支出法GDP是1(1)時間序列?？梢詫?jīng)過一階差分后的GDP建立適當?shù)腁RMA(p,q)模型。GDP單整性檢驗首先檢驗19782000年間中國支出法 GDP時間序列的平穩(wěn)性，即原序列的平穩(wěn)性。用 Eviews同時估計出 上述3個模型的適當形式。只要其中有一個模型的檢驗結(jié)果拒絕了零假設(shè)，就可以認為時間序列是平穩(wěn)的， 當3個模

2、型的檢驗結(jié)果都不能拒絕零假設(shè)時，則認為時間序列是非平穩(wěn)的。Eviews中，GDP平穩(wěn)性的ADF檢驗卞II型3、2、1的檢驗結(jié)果如下:Edit Object View Proc Quick Options Window Help.V近刃.Pmc. 0bject|Properties! .Print| .Stmplc|Genr|Bhcet.GraphStatsAugmented Dickeyuller Unit Root Test on GDPNull Hypothesis GDP has a unitrustExogenous: Constant. Linear TrendLag Length-

3、 2 (Automatic based on SIC, MAXLAG=4)t-StatisticProb/Aijpmnted 口i&依丫-Fu帕r test statistic口 31459S。9972Test criticml values: 1 % level-4 4983075% level-3.659410% level-3 268973*MacKinnon (1996) one-sided p-valuesAugmented Dickey-Fuller Test Equation Dependent Variable: D(GDP) Method. Least SquaresDate

4、 04 17/13 Time: 19:55Sample (adjusted) 1981 2000Included observations 20 after adjustmentsCoefficientStd Errort-StatisticProbGDP(-1)0 0092970 0295510 314596。了 574D(GDP(-1)1.4992970.1675728.9471640 0000D(GDP(-2)-1 0071780203389-4 9519730 0002C-1010.949805 4085-1.2552000 2286TREND(1978)229 1858120 137

5、81 9076910 0758R-squared0 941778Mean dependent var4228 060Adjusted R-squared0.926252S.D dcpEndcrrt va 1r3774.652S E of negression1025 064Akaike info citeion16,91622Sum 領(lǐng)間印resid15761356Schwarz criterion17.164115Log likelihood-164 1522Hannan-Quinn crit6r1696381MiawObjectProperties |Print.Hama|FraHza|

6、|Sample雇nr|shet.Graph|StateAugmented Dickey-Fuller Unit Root Tt on GDPNull Hypothesis GDP has a unit rootExogenous: ConstantLag Length 2 (Aulomatic based on SIC. MAXLAG=4)t-StatisticProb.*Augmented Dickey-Fuller test statistica3845321.0000Test critical va ues1 % level5% level 10% level-3.906546-3.02

7、0686-2.650413MacKinnon (1&96) one-sided p-values.Augmented Dickey-Fuller Test EquationDependent Variable D(GDP)Method: Least SquaresDate 04/1W3 Time IS 56Sample (adjusted): 1981 2000Included observations: 20 after adjustmentsCoefficientStd. Errort-StatisticProb.GDP(*1)0 0571300.0168803.384532U.003SD

8、(GDP-1)16521430.15885310.400460 0000D(GDP(-2)-1 15026B0204051-5 637156OQOOQC357.3266395.46400.9035630379GR-squared0.927652Mean dependent var4228.0OAdjusted R-squared0.914067S.D dependent var3774.652S E of r地曰 sig1106 384Akaike info criterion17,03244Sum squared restd195853的Schwarz crrtenor17 2315gLog

9、 likelihood-166 3244Hannan-Quinn enter1707131F-ststistic晚3&蝴6Durbin*Watson stat2.241361Yisw J Proc ObjBcti1ropBrtiBS | | Print | Hama | Frame | SampleGenr | Sha團gph隔ateAugmnt&d Dick&y-Fuller Unit Root Test on GDPNull Hypothesis: GDP has a unit rootExogenous NoneLag Length: 2 (Automatic based on SIC.

10、 MAXLAG-4)(StatisticProb *Auqmenled Dickev-Fuller test statistic4 1472210 9999Test critical values: 1% level-2.6857185% level-1 95907110% level-1.607456,MacKinnon (1996) one-sided p-values.Augmented Dickey-Fuller Test EquationDependenl Variable D(GDP)Method: L&ast SquaresDale 04/17/13 Time 19 59Sa m

11、 pie (adj usted): 1991 2000Irvluded observatione 20 after adjustmerrtsCoefficient Std Error t-Statistic Prob.GDP(-I)D(GDP(-1) DfGDP 0.0634231 701430 1.1910440 0152934.147221Q1433B611.466210197143605C7 貂0.00070 00000.0000R-squared0.923961Mean dependent var4228 060Adjusted R-squared0 915015S.D. depnde

12、nt var3774.652S.E. of regression1100.3MAkaike info criterian1690221Sum squared resid20584734Schwarz criterion17.13156Log likelihood-106 3221Hannan-Quinn criter17 01136Durbin-Watson stat2.250949根據(jù)3個模型檢驗結(jié)果。統(tǒng)計量都大于臨界值(左側(cè)單尾)，因此在行0.05的顯著性水平下，接受原假 設(shè)，即GDP序列存在單位根，是非平穩(wěn)序列。進一步對一階差分后的序列檢驗判斷GDP是否是一階單整序列，即1(1)。對GD

13、P 一階差分后序列進行 ADF檢驗，首先采用模型 3進行檢驗，檢驗結(jié)果為：Augmented Dickey-Fuller Unit Root Test on 口(GDPNull Hypothesis D(GDP) has a unit rootExogenous Constant Linear TrendLag Lengtti： 1 (Automatic based on SIC. MAXLAG =4)t-StatisticProb 7AuqmateJ 口ickey-Ful& 3t statistic-5 儂232口。睦Test c rttical values:1 % level4.4g的0

14、75% level-3.656446m level-3.2669731 MacKinnon (193S) one-sided p-valuesAugmented Dickey-Fuller Test EquationDependent Variable: D(GDP:2)Method Least SquaresDate- 04/17/13 Time: 20:05Sample (adjusted): 1981 2000Included observations： 20 3fter adjustmentsCoefficient Std. Error l-Statistic ProbD(GDP(-1

15、)-0.4940930.095480-5.1B32320.0001D(GDP(-1),2)Q965G310 1502625 4263210Q000C-1177 234590.3404-1.9941610 0635TREND(1978)261 2M261,762844 2299590.0006Rsqjared0.750206Mean dependent var2 的.1050Adjusted R-squared0 7033695 D dependent var1828 339S.E. of regression99E.7S33Akauke info cnterion16 32179Sum squ

16、ared resid15865MgSchwarz criterion17,02094Log likelihood-164 2179Hannan-Quinn criter16 96067F-statistic16 01755Durbin-Watson stat2.212567由于。=-5.183232p時，rj漸近服從如下正態(tài)分布：rk*N(0,1/n),因此，如果計算的 rk*2以后，r；| r&ca或 | State Resids；m- Equation; UNTITLED Workfile: CONS_GDP;lJLLL o_View Proc Object PrirU MartieFre

17、e Estitnaba | Forecast StMs ；Resids.Dependent Variable GDPDlMeUiod: Least SquaresDate. 04/17/13 Time, 2,0:41Sample (adjusted. 1981 2000Included observations. 20 after adjustmentsCorwergence achieved after 3 iterationsDependent Variable. GDPD1Method. Least SquaresDate 04/17/13 Time 20:42Sample (adjus

18、leJ- 1991 2000Included dbsetvaticins 20 after adjustmensCoefficient Std Error ( Statistic PnobCoefficient Std Error t-Statistic PrabAR(1)1 5927010 2013227.9116140 OOMAR(2)4)652666O2OJ633-3.2051140 0049C909.5563457.79741.9B6B10O.OGSSGDPD1(-1)1,4947640.1930327.743S900.0000GDPDl (-2)-0.6779650.13919395

19、a34620 0023R-squaredO.B47029Mean dependent var422S.060Adjualed R-squaredO.B3B531S.D. dependent var3774.652S.E. o( regression1516.776Akaike irtfo criterian17.58120Sum squared resid41410964 Schwarz critenori17.6807?Log likelihood-173.SI 20 Hannan Quinn criter.17,6006Durbin-Watson sial 1 14 9667R-squar

20、ed0,375S56Mean dependent vsr4228.060Adjusted R-squared0.861251S.D. dependent var377.652S.E of regression1406 023 AXaik&info criterion17 47240Surn squared resid33607319Schwarz criterion17 62176Log UkelihQQd-171.724QHann即-Quinn criter17 5(HMF-st?tistic59.96078Durbin-Watson 煢磯1 220600Prob(F-siaiistic)O

21、.DOOOOOInverted AR Roots80- 14iSQ+ Hi令3個模型的殘差序列分別為：e1、e2、e3,檢驗是否為白噪聲序列: Series: El Workfile: CONS_GDP:UntitledVtew procobjl1ropwttiM| 卜忖hMftGraph stCorrelogram of E1Date 04/17/13 Urne- 20 6Sample: 197S 2000Included observations. 20LJ Series: E2 Workfile: CONS_GDP:UntitledVewjproc| ObjectProperties |

22、Fnnttame|Freeze| |Sampta|Gmu|5hwtGraph|每向Corrvlogram of EfDate: 04717/1S Time: 20:4Sample 1978 2000Included obsen/stions 20AutocorrelationPartial CorrelationAC PAC Q-Stal Prob1口1n1 0.3fi2 0.382 3.嫻R 0.066111匚12 0.013 -0.155 3.3S921匚11 13 -0.133 -0.096 3.8473 0 2781匚11 二14 -0.342 -0.299 7.0557 0.1331

23、匚II115 -0.171 0.079 7 9122 0.1611ZI 11二 16 0.253 0340 9.9297 0.1281 11匚17 0 H4 -0 180 10.635 0.1S511 i1 18 0 057 -0.063 10,753 0.2161li1 19 -Q 019 -Q 045 10 768 Q 2921匚I1口 iw -Q 146 0啊 11 705 口 3051匚H1匚i11 -0232 -0 189 14 342 02151n1l12 -0(M9 -0 044 14 473 0 272AntacorrelatiQn Partial CorrelaitonAC

24、PAC Q-Srat Probii匚 i匚11匚111111 1 11 11Zl 13 113 iIP )i i 口i I 11111 1111 1111 025B Q25B 1 &403 0215 2 -Q139 -0 220 2 0106 03C6 3 -024 -0164 3.5721 0312 4 0 529 4X511 11 273 0024 5 -0300 -0.204 13.916 0.016 6 0,270 0213 16.215 0.013 70,158-0.24717.0610.01780,115-0.18017.550002590097-0.14317 923003610

25、-0.0360.18117 9790.05511 -0.136 -0103 1&8& 0.063 12 0,064 -O.ODB 19.113 0.066口&祀-M，17； 13 Time- 2046Sample. 1978 2000Included observations 20Autocorrelation Partial Correlation AC PAC Q-Stal Prob1 口 1l 1 0.257 0257 1.5290 0.216I (1 112 -OgQ -0 113 1 5674 0 157 1113 -0.059 -0.020 1.8583 0.646ICZ11 二1

26、4 -0 328 -0 337 46240 0 32B1 c1115 -0.1* 0 026 5 25M 0 3S21n1Il6 0,345 0 398 9.036A 0 1721 11 t17 0155 -0 070 9 8515 0.197111 018 0076 -0 037 1 0.064 0 2611i1 19 0,011 -0 089 10069 0 3451匚l1 1IQ -0.123 0 180 10 733 0 3791匚i1匚111 230 -0 161 13 325 0 2731l i112 -0 012 -0 061 13 333 0.345觀察Q統(tǒng)計量和相應(yīng)的概率值發(fā)

27、現(xiàn)，模型（1）與模型（1）的殘差項接近于一白噪聲，但模型（2）存在4階滯后相關(guān)問題，Q統(tǒng)計量的檢驗也得出模型 2拒絕所有自相關(guān)系數(shù)為零的假設(shè)。因此：模型1與3可作為描述中國支出法 GDP 一階差分序列的隨機生成過程。ARMA（ p,q）模型應(yīng)用用建立的AR（2）模型對中國支出法 GDP進行外推預(yù)測：模型（1）可作如下展開：GDP _GDPt=?（GDPt一GDPy） +%（GDPtq -GDPt）GDR =（1+*）GDR4+ d2 平1）GDP %GDPt 二如果已知t-1、t-2、t-3期的GDP時，就可對第t期的GDP作出外推預(yù)測。模型（3）也可展開，但多出一項常數(shù)項。Eviews中，對

28、樣本外一點的預(yù)測，如果該樣本點已超過workfile的樣本范圍，則需要調(diào)整樣本區(qū)間,操作如下：點擊工作文件的 Proc選項卡，選擇Structure/Resize Current Page”選項，宜 EViews| File_Edit Object View Proc Quick Options Window Hseries e3-residO Workfile： CONS_GDF - (c:utersthinkpadde. I o F-a |j|wwIproc|；QbjMt| |pnnHsawejptallat/-：忖how |?而旬又附心值想anlM cc1口出由出，ir3Fe,E 曰出出

29、Rss回0回回回0回回回回=1回回Set Sample.Stmcture/Resize Current Page. Append to Current Page Contract Current Page., Reshape Current Page Copy/Extract from Current Page Son Current ”Laad Workfile Page 5ave Current Page.Rename Current Page.Delete Current PageImport txportsidDisplay Fit會出現(xiàn)如下對話框:點擊OK確定，則:隹I EViews

30、File Edit Object View Proc Quick Options Window 卜 senes e3=resid Workfile: CONS_GDP - (c:usersthinkpadde. gview Iftqc 匕國ct. Print 58聘網(wǎng)&佰鏟 show fetch 5用胎| elate GenDisplay FRange 1979 2001 - 24 obsSample 1978 2001 - 24 obs回 eq_ci=eq_ecrnE gdp=gdp_acff3g S-1 2 3d deee6 21eii gdp_d1_acf_pacf 0gdpdl-l=-

31、 gdpdl _1_acf_q 世 gd 口 dLZacf.q 世 gdpdlafq =1= gdpdl _acf_pacf =gdpdl ar工作文件樣本區(qū)間擴大到 2001年。模型（1）的預(yù)測:國 EViewsFile Edit Object View Proc Quick Options Window Helpseries e3=residseries gd pf_1 =2 239 gdp (-1 HO,442+1,239)*gdp(-2)+0.442*gdp(-3)Series: GDPF_1 Workfile: |viewpnc|obj&dpnoperfes fFrinRange 1

32、978 2001Sample: 1979 2001-24 obs-24 0bsDisplay Fil,回c回 eq_ci國 grojpO2S cons=eq ecm0 Icons回 cons_gdp_4S gdp01gdpHdconsigdp_adf30 residSdgdpgdp_dl_acf_pacf國 dgdp_a3 gdpdtSeikii gdpd7 1 acf qSe2Wk qdpdl 2 acf qSe3建 gdpdl 3 acf q0eci耀 gdpd1_acf_pacfSecil回gdpdl ar回 eqOlSgdpd1f3叵|eq02s5EM國 eq030 gdpf_3H e

33、q04回 graupOlO Workfile: CONS_GDP (c:usersthinkpadde. | 0口bjSCt ； Det祈信+廣j 5 Mw 除岫 :gletg |Genr* h Untitled 卜 Wvv Pagv/19858253.0581&8610327 69198711074 49193S13236,81198917590 87199017358.791&9119837 19199224129 711&9330233.691G94431761219955797豆 56199667767.25199775272.901&9878636 35199931193 28200

34、085403 7620D195468 86模型（3）的預(yù)測: WorkHle: CONS_GDP - (c:usersthinkpddde.,.Q1國、Viev Proc ObjectPrintSaveDetoils+/- Shovr | Fetch StoreDelete |Genr |sampjTPD1F3 Wo eRange: 1978 2001 - 24 obsSample. 197S 2001 - 24 obsDispla_ _ L 二 Ik/r二1 Series: GDPF 3 Work1國c=eq_ci0 lea回 eq_ecmS Igd_4S gdp3 res=gdp_adf3

35、端 gdpd1_acf_pacf0 gdpdlgdpdl搟gdpdl _3_acf_q 耀 gdpdCacf pacf 畫 gdpdl_arnsP idView Proc ObjectProperties=lc0ns_g(0 d8nsi S dgdp =dgdp ar Se10 e30 eci0 ecil國 eq01回 I eq02 國 eq03 回 I eq04* 卜 Untitli19858792J0。198610132 8019g7117g4 70198S14704 00198916466 00199018319 500 gdpdlf319912120040199225063 70ibi

36、groupu i 圄 group02199334500 70199446690 70md Z199558510.50New Paae /199668330 40Durbin-Watson stat6r19199774894 20Inverted AR Roots.80-.14iS0+.14i20199879003 3020199982673 10200089112 50200197159 44因此有:對2001年由國支出法GDP的預(yù)測結(jié)果(億元)預(yù)測值實際值誤差模型19546995933-0.48%模型39716095933128%應(yīng)用ARIMA(p,d,q)模型建模過程：對原序列進行平穩(wěn)性檢

37、驗，如果序列不滿足平穩(wěn)性條件，可以通過差分變換或者其它變換，如對數(shù) 差分變換使序列滿足平穩(wěn)性條件；通過計算能夠描述序列特征的一些統(tǒng)計量( ACF和PACF),來確定ARMA模型的階數(shù)p和q,并 在初始估計中選擇盡可能少的參數(shù)；估計模型的未知參數(shù)，并檢驗參數(shù)的顯著性，以及模型本身的合理性；進行診斷分析，以證實所得模型確實與所觀察到的數(shù)據(jù)特征相符。以上第3、4步，需要一些統(tǒng)計量和檢驗分析在第2步中的模型形式選擇是否合適，所需的統(tǒng)計量和檢驗如1)檢驗?zāi)Ｐ蛥?shù)顯著性水平的t統(tǒng)計量；2)為保證ARIMA(p,d,q)模型的平穩(wěn)性，模型的特征根的倒數(shù)皆小于1;3)模型的殘差序列應(yīng)當是一個白噪聲序列，用檢驗

38、序列相關(guān)的方法如Q統(tǒng)計量檢驗。用經(jīng)過居民消費價格指數(shù)調(diào)整后的 構(gòu)建誤差修正模型。11)檢3C cons和inc的平穩(wěn)性案例2 ECM莫型建模估計1978-2006年中國居民總量消費(cons)與總量可支配收入(inc)的數(shù)據(jù)Null Hypohiesis D(CONS,2 has a unit rootExogenous, constant, Linear TrendLag Lenglh 0 (Aulomal r ba on SlC MAXL.AG=6)Null Hypothesis D(lMC.2) has a unil routEjogenous constant. Lmear Trend

39、Lag Length- 0 AutDma.tic based o.2) C TREND(197fl)-1.161657-10448115,805630 200398-5.547606207 3544-0 50275312,226041.2927840 00OD 0.6199 o.2oaaR-squared0.572788Mean dependent var31.64615Adjusted R-squan&d0 53563SS D Oependenrvar673 7445S.E. of Feges5icm459.1167Akaike info criterion15,20465Sum squar

40、ed resrt48481翡Scnwvarz criterion15 34962Log likidihaod-194.6605Hannan-Quinn cliter.15.24645F-staiisiic15 4T872OurtHn-W3t$oiri 與tat2045250Prc*(F-slaU5bc)0.000057,-SlatislKProb/Augmented Dickev-Fuller test statistic-4 9aB5130.0024rest cribcai values1 % level5% level 1。呢 level-4 356068 -3.595026-3 2334

41、56MacKirinon one-suled p-vaiuesAugmenl&d DicKey-Fuller Test EquationDependent Vanatue- D(INC 3)Method. Least SquaresDate 12/3Q/14 Time，2T2啟sample (adjusted) 1S81 2006Included oiiservatians 26 after ad|U5trnenisCDefficieniSid Error l-StatisticProbD(1NC(-1).2)-1.0387330 208225-4.9885130.0000c-417 6950

42、461 7033-0.9046830.3750TREND(1978)54,2482528,4514Q1.9066920.0691R-squa.red0.519713Mean dependent var60.2153SAdjusted R-squaned0 47794BS D dependent var1385 472S.E. of regression1001.047Akaike info criterion1676365Sum squared resid248191Schwarz entenon16.90881Log likeliliDod-214.9274Hanrian-auinn tri

43、ler.16.80545F-statishc12 44399Durbin-Wat sw stat1 98G4D9Prob(F-s(atistic!0 000217Hull Hypothesis: D(LN_CONS) has a unit root Exogenous con&tanl-Llnear TrendLag Length 0 (Automatk based on SIC. MAXLG=6)t-StatisticProb*Augmented Dickev-Fuller lest statistic-3.7691070 0345Test ent re sii values-1% leve

44、l5% level1。 level-4 339330-3.5B7527-3 229230MacKinntin one-sided p-valuesNull Hypolhesis: DfLNJNC has a unit root Exogenous Con&tam-Linear TrendLag Length. 0 (Automatic based ort SIC. MAXLG=6)t-stausiicProb*Auamented Dickev-Fuller lest statistic-4.1109560 0166Test Eiitical values1% level5% level 10%

45、 level-4 339330-3 5B7527-3.229230*MacKinnon (1996) one-sided p-vlue5.Augmented Die Key-Fuller Test Equation dependent Variable D(LW_CONS.2) Method Least Squares Date. 12/30/14 Time. 20.34 Sample (aajusleti) 1980 2006 lnclu(led observalions 27 after adju&tmentEAugmented OicK&y-Fulier Test EquationDep

46、endent variable, dclnjmcjMethod Leasl SquaresDale 12/30/14 Time 20.34Sample (adjusled) 1980 2006included oDB&rvaltons 27 after adjustmentscoeTTicient std. Error t-stallstlc ProbD(Lh_CONS(d) CT RE ND(1?78-0 740151 0.05B115 5 29E-D50.19636837691970.0225572.5763950 WXJB53 D 062090Q.D009 0.0166 0 961QR-

47、5quareci0 377136Mean oepencJent var-0 000473Adjusiefi R-squared0.325231s D. dependeni var0 041629S E ot regression0 0341196Akaike info errterion-3 808975sum squared resic!0.02&D65sc rmarz criterion-3.664993Log likeliliQad54 42118Hannan-Quinn enter-3.76E161F-statlstlc7.265852Dufbin-walson stal1 83577

48、8C-.-LPL -it -上：:上： _ *nrkCoenicierirSM Error t-5t3llstlcProbD0-H INC(-1J)-0 Bi48230198205-4 11095C0 0004c0 0538010.0244892.1969020.0379TREMD(1S780 0013060 0011371.1464010 3621R-squared0 418272Mean dependent var0.000565AdjDBled R-squared0 369795s D deperdenl var0057427S.E of regression0 045589Akaike

49、 info criterion-3.233883sum squared nesia0 049079Scmwarz criterion-3 0B9901Log likelihood46,65742Hannan-Quinn criter.-3.191070F-statistice,、g628201r jn. ,n j e- n nOurtun-Watson star1 87652C*ni經(jīng)檢驗發(fā)現(xiàn)cons和inc序列都是I(2)時間序列，而取對數(shù)后的ln(cons)和ln(inc)序列都是1(1)時間序歹U, 根據(jù)經(jīng)濟理論擬構(gòu)建ln(cons)和ln(inc)的長期均衡模型。檢33 ln(cons

50、)和ln(inc)的協(xié)整關(guān)系Depeodenl Variable. LN_CONSMetncKl Least SquaresDtc 12/30/14 Time- 20 38Sample 1078 2M6included observations. 29CoefficientSid Errort-StabsUcPnot)C0.5872550.1427994.11245B0.0005LNJMC0.8800210.01421961,391950.0000R-四V吊red口993001Men dependent v3r0401233Adjusled R-squared0 992742S.D. depe

51、ndenl var0.6fi5fi0s.E of regression0.056790AH卻 Ke info criterion-2.832439sum squared residQ .08 7078Schwarz cfflefion-2.730143Log liKeiinocxi心 07037Hannan-Quinn enter-2 SO29O7F-slatistic3S30 614Durtoinr-Watson stat0.415198Prob(F-slatistit)0.000000Augmented Dickey-Fuller unrt Root rest on ecmNull Hyp

52、olhesis ECM has m unit root Exogenous, constant. Linear Trend Lag Length: 6 (Auiomatic Eased on sic. maxlag=6)t-SltisticProb *Adornented Dickey Fuller 乳 stali3M7.8194180.0000Test cnttcal values1% level-4 4407395峪 level-3&S28961G% level-3 254671MacKinnon (1996) one-sided p-values.Augmented Dickey-F j

53、llfer Test Equalion Dependent Variable. D(ECM) Melhod: Least Sa wares* ill*殘差ecmt序列的AEG檢驗,t 統(tǒng)計量=-7.819 1 111 0.159 0.055 16.9S0 0.10B:1*12 0O56 0 043 17 36B Q 1弼經(jīng)檢驗殘差序列平穩(wěn)，且不存在序列相關(guān)。 誤差修正模型為：ln(const) = 0.794 ： ln(inct) - 0.241ecmu-10 -案例3 VAR模型建模、估計、檢驗與應(yīng)用凱恩斯學派認為貨幣供給量變動對經(jīng)濟的影響是間接地通過利率變動來實現(xiàn)的。貨幣政策的傳遞主要 有

54、兩個途徑：一是貨幣供給與利率的關(guān)系，即流動性偏好途徑；二是利率與投資的關(guān)系，即利率彈性途徑。 根據(jù)凱恩斯的理論，當貨幣供給量增加時，貨幣供給大于貨幣需求，供給相對過剩，利率下降，刺激投資， 促進國民經(jīng)濟增長。當然他假定利率變動是由市場調(diào)節(jié)的，與貨幣供給量呈反方向變動。在我國利率是基 本固定的，但是仍然可以利用政策手段直接調(diào)整利率或投資，同樣可以達到經(jīng)濟宏觀調(diào)控的目的。但貨幣 學派主要強調(diào)貨幣供給量對經(jīng)濟的短期影響，而中長期，貨幣數(shù)量的作用主要在于影響價格以及其他貨幣 表示的量，而不能影響實際國內(nèi)生產(chǎn)總值。為了研究貨幣供應(yīng)量和利率的變動對經(jīng)濟波動的長期影響和短期影響及其貢獻度，采用我國1995年

55、1季度2007年4季度的季度數(shù)據(jù)，并對變量進行了季節(jié)調(diào)整。設(shè)居民消費價格指數(shù)為CPI_90(1990年1季度=1)、居民消費價格指數(shù)增長率為CPI、實際GDP的對數(shù)ln(GDP/CPI_90)為ln(gdp)、實際M1的對數(shù)ln(M1/CPI_90)為ln(m1)和實際利率rr(一年期存款利率 R-CPI)。利用VAR(p)模型對ln(gdp)、ln(m1)和rr3 個變量之間的關(guān)系進行實證研究，其中實際GDP和實際M1取對數(shù)差分后平穩(wěn)，出現(xiàn)在模型中，而實際利率是平穩(wěn)序列，沒有取對數(shù)。變量平穩(wěn)性檢驗Augmented DickeyFuller Unit Root Test an D(LN M1

56、 P SA)Null Hypothesis: D(LH_M1_P_SA) has a unit root Exogenous Constant, Linear TrendLag Length 0 (Aulomalic based on SIC. MAXLAG-10)tStattsttcProt).aAugmented Dickey-Fuller test static he-8 075925oooooTesl c riitic al values.1 % level5% level 10% level-4 J 52511 0502373 -3.180699tMacKinnon (1996) o

57、ne-sided p-valuesAugmented Dlclcey-Fulier Test EquaUon Dependent Variable D(LN_M1_P_SA,2) Method Least Squares Date 12/3014 Time. 21.39sample (adjusied). 1S95Q3 2007Q4 included observations 50 after adjusunemsAugmented Dickey-Fulltr Unit Root Tst on D|LN GDP P SAhull Hypothesis. d(ln_gdp_P_5A)has a

58、unit rootErogenous Constarrt. Linear TrendLag Length 0 (Automatic based on SIC MAXLAG=10)I-StatislicProb?Auamenieo Oictev-Fuiier tesl statistic-11 BS914oooooTest critical11% level-4 1525115% level-3.50237310% level-3.1806&9MacKinnon (1996) one-siaed p-vaiuesAugnienl&d Dickey-Fuller Test Equation Dep

59、endent Variable D(LN_e啊.|prn omectl Pnnt Hare I Fneeie Eme IIlirpuPe | 融金7|RepresentationsEstimation OutputResiduak，on Estinwica196550.216726J5842)(0 J 5974)M60Sj 1 35E744Endogenous TableEndogenous Graph297220 258082電368)(0J62M)aniQ7i r iLag StructureAR Roots TableAR Roots GraphGranger Causality/Blo

60、ck Exogeneity TestsLag Exclusion TeitsResidual Test占卜Cointegration Test.mpulse Response.Variance Decomposition.Lag Length Criteria-Label038S20.409763Sum sq. re&ids26.35B500.0067460.002573 Var VARI Worlcfile: R MlVAR Lag Order Selection CriteriaEndogenous variables RR DLOG(M1_P_SA) DLOG(GDP_P_SA) Exo