最新計(jì)量經(jīng)濟(jì)學(xué)論文--中國糧食總產(chǎn)量多因素分析

1、中國糧食總產(chǎn)量多因素分析專業(yè)年級(jí)：13金融(2)班 學(xué)號(hào): 201312030140 姓名： 謝昊摘要：本文選取1990年到2013年的相關(guān)數(shù)據(jù)，應(yīng)用計(jì)量經(jīng)濟(jì)學(xué)所學(xué)知識(shí)對(duì)根據(jù)經(jīng)濟(jì)理論選取的影響我國糧食產(chǎn)量的各因素進(jìn)行分析、檢驗(yàn)，并對(duì)其影響程度的大小進(jìn)行定量分析， 進(jìn)一步明確和完善相關(guān)的 經(jīng)濟(jì)學(xué)知識(shí)。關(guān)鍵詞：糧食產(chǎn)量糧食播種面積農(nóng)用機(jī)械總動(dòng)力有效灌溉面積農(nóng) 業(yè)化肥使用量一、文獻(xiàn)綜述農(nóng)業(yè)作為我國最基礎(chǔ)的產(chǎn)業(yè)，農(nóng)產(chǎn)品的每年的產(chǎn)量直接關(guān)系著我 們的民生，故而糧食的產(chǎn)量一直是我們最關(guān)心的。影響因素的分析首先，糧食作為農(nóng)作物，其產(chǎn)量肯定會(huì)受到農(nóng)用化肥施用量條件的影響其次，我認(rèn)為糧食的播種面積對(duì)于糧食產(chǎn)量也

2、有一些影響最后，農(nóng)業(yè)機(jī)械總動(dòng)力也是影響糧食產(chǎn)量的一大重要因素二、數(shù)據(jù)收集與模型的建立(一)數(shù)據(jù)收集1983年一200處中國糧食生產(chǎn)與相關(guān)投入的資料(表 1)年份糧食總產(chǎn)量Y糧食耕種面積(x1)農(nóng)用化肥施用 量(x2)農(nóng)業(yè)機(jī)械總WJ (x3)1990446241134662590.3287081991435291123142805.1293891992442641105602930.2303081993456491105093151.9318171994445101095443317.9338021995466621100603593.7361181996504541125483827.9385

3、471997494171129123980.7420161998512301137874083.7452081999508391131614124.3489962000462181084634146.4525742001452641060804253.8551722002457061038914339.457930200343070994104411.6603872004469471016064636.6640282005484021042784766.2683982006498041049584927.7725222007501601056385107.8765902008528711067

4、935239821902009530821089865404.4874962010546481098765561.7927802011571211105735704.2977352012589581112055838.81025602013601941119565911.9103907（二）模型設(shè)計(jì)為了具體分析各要素對(duì)我國糧食產(chǎn)量影響的大小，我們可以用糧食總產(chǎn)量（y）作為衡量，代表糧食發(fā)展；用糧食耕種面積（x1）、農(nóng)用化肥施用量（x2）以及農(nóng)業(yè)機(jī)械總動(dòng)力（x3）。運(yùn)用這些數(shù)據(jù)進(jìn)行回歸分析。采用的模型如下:y= B 1+ B 2x1+ B 3x2+ B 4x3+u其中，y代表糧食總產(chǎn)量，x1代表

5、糧食耕種面積，x2代表農(nóng)用化肥施用量，x3代表農(nóng)業(yè)機(jī)械總動(dòng)力，u代表隨機(jī)擾動(dòng)項(xiàng)。我們通過 對(duì)該模型的回歸分析，得出各個(gè)變量與我國糧食產(chǎn)量的變動(dòng)關(guān)系。三、模型估計(jì)和檢驗(yàn)(一)模型初始估計(jì)(表二)Dependent Variable: 丫Method: Least SquaresDate: 12/21/15 Time: 16:27Sample: 1990 2013Included observations: 24VariableCoefficientStd. Errort-StatisticProb.C-44644.146601.867-6.7623500.0000X10.6841160.0531

6、1312.880430.0000X24.0429710.9747514.1476970.0005X30.0310320.0383520.8091310.4280R-squared0.966281Mean dependent var49317.62Adjusted R-squared0.961223S.D. dependent var4867.060S.E. of regression958.4155Akaike info criterion16.71945Sum squared resid18371206Schwarz criterion16.91579Log likelihood-196.6

7、334F-statistic191.0450Durbin-Watson stat1.534928Prob(F-statistic)0.000000回歸函數(shù)為：.A丫 =Y4644.14+0.684116X1 +4.042971X2+0.031032X3(6601.867)T= (-6.762350)R2 =0.966281(0.053113)( 0.974751 )(0.038352)(12.88043)(4.147697)(0.809131 )2R =0. 961 223 F= 1 9 1. 045（二）多重共線性檢驗(yàn)X1相關(guān)系數(shù)矩陣（表三）X1X21-0.267566314901X3-0.

8、23239867238X2-0.2675663149010.977074961235X3-0.232398672380.977074961235根據(jù)多重共線性檢驗(yàn)，解釋變量之間可能存在著線性相關(guān)。為了進(jìn)一步了解多重共線性的性質(zhì)，我們可以做輔助回歸。（表四）被解釋變量可決系數(shù)R2的值方差擴(kuò)大因子X10.090191.09913X20.95640922.9405X30.95558322.6398由上表可以得知，輔助回歸的可決系數(shù)很高，經(jīng)驗(yàn)表明，方差擴(kuò)大因子VIFj >=10時(shí)，通常說明該解釋變量與其余解釋變量之間有嚴(yán)重的多 重共線性，這里的x2、x3的方差擴(kuò)大因子遠(yuǎn)大于10,表明存在嚴(yán)重 的

9、多重共線性問題。為了進(jìn)一步篩選并剔除引起多重共線性分變量， 需要采用逐步回歸的 方法。分別作Y對(duì)X1、X2、X3的一元回歸，意愿回歸結(jié)果如下表（表五）變量X1X2X3參數(shù)估計(jì)值0.3696284.0710710.162556t統(tǒng)計(jì)量1.4728006.7542466.867695R20.0897480.6746520.681921R20.0483730.6598630.667463（表六）X1X2X3R2X1、X30.641034(9.246298)0.186325(16.84505)0.937277X2、X31.5875860.1009490.686571(0.558181 )(0.8936

10、59)通過采用剔除變量法，多重共線性的修正結(jié)果如下：剔除X2。（表七）Dependent Variable: YMethod: Least SquaresDate: 12/25/15 Time: 10:06Sample: 1990 2013Included observations: 24VariableCoefficientStd. Errort-StatisticProb.C-31636.647732.436-4.0914190.0005X10.6410340.0693299.2462980.0000X30.1863250.01106116.845050.0000R-squared0.93

11、7277Mean dependent var49317.62Adjusted R-squared0.931303S.D. dependent var4867.060S.E. of regression1275.661Akaike info criterion17.25679Sum squared resid34173555Schwarz criterion17.40404Log likelihood-204.0814F-statistic156.9019Durbin-Watson stat1.001388Prob(F-statistic)0.000000修正后方程為A0.186325X2(77

12、32.436)T= (-4.091419)(0.069329)(9.246298)(0.011061)(16.84505)R2 =0.937277R2 =0.931303 F =156.9019Y' = -31636.64 + 0.641034X1（三）異方差檢驗(yàn)（表八）ARCH Test:F-statistic0.037667Probability0.847978Obs*R-squared0.041181Probability0.839189Test Equation:Dependent Variable: RESIDEMethod: Least SquaresDate: 12/24

13、/15 Time: 18:58Sample(adjusted): 1991 2013Included observations: 23 after adjusting endpointsVariableCoefficien Std. Error t-Statistic Prob.tCRESIDA2(-1)1280357.0.041531504218.42.5392910.2139870.1940810.01910.8480R-squared0.001790Mean dependent var1341173.Adjusted R-squared-0.045743S.D. dependent va

14、r1852594.S.E. of regression1894492.Akaike info criterion31.82974Sum squared resid7.54E+13Schwarz criterion31.92848Log likelihood-364.0420F-statistic0.037667Durbin-Watson stat1.986528Prob(F-statistic)0.847978由上表可以得知，(n-p) R2 =0.041181 ,給定顯著性水平為0.05,查72分布表得臨界值72. (p) =5.9915> (n-p) R2,則接受原假設(shè)，表明模型中的

15、隨機(jī)誤差項(xiàng)不存在異方差O(四)自相關(guān)檢驗(yàn)(表九)Dependent Variable: YMethod: Least SquaresDate: 12/25/15 Time: 10:06Sample: 1990 2013Included observations: 24VariableCoefficientStd. Errort-StatisticProb.C-31636.647732.436-4.0914190.0005X10.6410340.0693299.2462980.0000X30.1863250.01106116.845050.0000R-squared0.937277Mean de

16、pendent var49317.62Adjusted R-squared0.931303S.D. dependent var4867.060S.E. of regression1275.661Akaike info criterion17.25679Sum squared resid34173555Schwarz criterion17.40404Log likelihood-204.0814F-statistic156.9019Durbin-Watson stat1.001388Prob(F-statistic)0.000000Y' = -31636.64 + 0.641034X1

17、+0.186325X2(7732.436)(0.069329)(0.011061)T= (-4.091419 )(9.246298)(16.84505)22R2 =0.937277R2 =0.931303 F =156.9019查DW夜可知，dl=1.188 , du=1.546,模型中DW<d顯然有自相關(guān)r65000-60000- 56000-50000-45000-4000090 92 94 96 98 00 02 04 06 08 10 12Res id 向Aduzl Fitted |(表十)殘差的變動(dòng)有系統(tǒng)模式，連續(xù)為正和連續(xù)為負(fù)，表明殘差項(xiàng)存在一階 自相關(guān)。對(duì)模型進(jìn)行BG僉驗(yàn)，

18、用Eviews分析結(jié)果如下: （表H-）Breusch-Godfrey Serial Correlation LM Test:F-statistic2.642994Probability0.097113Obs*R-squared5.223742Probability0.073397Test Equation:Dependent Variable: RESIDMethod: Least SquaresDate: 12/24/15Time: 19:24VariableCoefficientStd. Errort-StatisticProb.C1247.5647284.9980.1712510.86

19、58X1-0.0114660.065346-0.1754620.8626X30.0001740.0102860.0169140.9867RESID(-1)0.5170860.2287092.2608930.0357RESID(-2)-0.1401580.230260-0.6086960.5499R-squared0.217656Mean dependent var-4.21E-12Adjusted R-squared 0.052952 S.D. dependent var 1218.937S.E. of regression1186.225Akaike info criterion17.177

20、99Sum squared resid26735479Schwarz criterion17.42342Log likelihood-201.1359F-statistic1.321497Durbin-Watson stat1.906529Prob(F-statistic)0.297918由上表顯示LM=TR2=5.223742 ,其p值為0.073397,表明存在自相關(guān)。對(duì)模型進(jìn)行處理：對(duì)原模型進(jìn)行科克倫-奧克特迭代法做廣義差分回歸，用 Eviews進(jìn)行分析所得結(jié)果如下：Dependent Variable: YMethod: Least SquaresDate: 12/24/15 Time

21、: 19:38Sample(adjusted): 1991 2013Included observations: 23 after adjusting endpointsConvergence achieved after 10 iterationsVariableCoefficientStd. Errort-StatisticProb.C-39779.1510686.99-3.7222020.0014X10.7219490.0973507.4160450.0000X30.1791840.0184359.7199340.0000AR(1)0.4884590.1835412.6613060.01

22、54R-squared0.956230Mean dependent var49521.70Adjusted R-squared0.949319S.D. dependent var4870.329S.E. of regression1096.434Akaike info criterion16.99428Sum squared resid22841163Schwarz criterion17.19176Log likelihood-191.4343F-statistic138.3617Durbin-Watson stat2.032169Prob(F-statistic)0.000000Inver

23、ted AR Roots.49(表十二)由圖表知 DW=2.03216何以判斷 du=1.543, dl=1.168 , du<DW<4-du,說明無自相關(guān)。AY' = -39779.15+0.721949X1 +0.179184X2rA,八 cmi12 AR(1)= 0.488459 (10686.99)(0.097350)(0.018435)T= (-3.722202) (7.416045)(9.719934)R2 =0.956230 R2 =0.949319 F =138.3617(五)時(shí)間序列的平穩(wěn)檢驗(yàn)：(表十三)ADF Test Statistic -3.2302

24、771% Critical Value* -2,67565% Critical Value-1.957410% Critical Value-1.6238*MacKinnon critical values for rejection of hypothesis of a unit root.Augmented Dickey-Fuller Test EquationDependent Variable: D(E)Method: Least SquaresDate: 12/24/15 Time: 19:47Sample(adjusted): 1992 2013Included observati

25、ons: 22 after adjusting endpointsVariableCoefficientStd. Errort-StatisticProb.E(-1)-0.6891400.213338-3.2302770.0042D(E(-1)0.0983720.2016690.4877870.6310R-squared0.364916Mean dependent var136.8152Adjusted R-squared0.333162S.D. dependent var1264.449S.E. of regression1032.551Akaike info criterion16,803

26、96Sum squared resid21323242Schwarz criterion16,90315Log likelihood-182.8436Durbin-Watson stat1.989449經(jīng)檢驗(yàn)，表明殘差序列不存在單位根，是平穩(wěn)序列。經(jīng)濟(jì)意義檢驗(yàn)：所估計(jì)的參數(shù)以日2、久分另U為-39779.15、0.721949、0.488459從經(jīng)濟(jì)學(xué)意義上來說，我國糧食產(chǎn)量 y與農(nóng)業(yè)農(nóng)用機(jī)械總動(dòng)力x3正 相關(guān)，與糧食耕種面積(x1)成負(fù)相關(guān)。1、擬合優(yōu)度檢驗(yàn)(eviews表在附錄里最后一張表)2可決系數(shù)R =0.956230, R2 =0,949319,這說明所建模型整體上對(duì)樣本數(shù)據(jù)擬合較好

27、，即解釋變量“糧食耕種面積”和“農(nóng)用化肥施用量”對(duì)被解釋變量“糧食總產(chǎn)量”的絕大部分差異作了解釋。2、F檢驗(yàn)針對(duì)H0: B 1 = 8 2=0,給定顯著性水平 =0.05,在F分布表中查出 自由度為k1 = 2和n k=20的臨界值F% (2,20) =3.49,由表 中得到 F= 138.3617>F% (2, 20) =3.49,應(yīng)拒絕原假設(shè) H。p 1 = (3 2=0,說明回歸方程顯著，解釋變量“糧食耕種面積”和“農(nóng)用化肥施用量”對(duì)被解釋變量“糧食總產(chǎn)量”有顯著影響。3、t檢驗(yàn)針對(duì)H。月=0 ,和H0B2=0,由上表可以看出，t (瓦)=-3.722202 , t (久)=7.4

28、16045, t(息)=9.719934,取 a=0.05 ,查表 t。.(20) =2.086.因?yàn)閠 (久)> t0.25 (20),所以拒絕H0:日1 =0,因?yàn)閠(P2) > t0.025 (20),所以拒絕H0：P2=0 ,因?yàn)?t( P1) <t0.025(20),所以接受H0 :1=0。對(duì)斜率系數(shù)的顯著性表明，解釋變量“糧食耕種面積”和“農(nóng)用化肥施用量”對(duì)被解釋變量“糧食總產(chǎn)量”有顯著影響。四、結(jié)論分析和政策建議(一)主要結(jié)論1)從模型可以看出農(nóng)民對(duì)化肥的投入量，即模型中的化肥施用 量，是影響糧食總產(chǎn)量增產(chǎn)的最顯著因素，說明我國目前農(nóng)業(yè)生產(chǎn)中， 農(nóng)民對(duì)農(nóng)業(yè)的投入

29、所產(chǎn)生的效益最大。2)從模型可以看出，糧食作物耕種面積也是影響糧食總產(chǎn)量的重要因素之一，擴(kuò)大糧食作物耕種面積無疑是可以使糧食增產(chǎn)的。3）農(nóng)業(yè)機(jī)械化是農(nóng)業(yè)現(xiàn)代化的重要內(nèi)容和主要標(biāo)志之一，而通過對(duì)模型的回歸分析，可看出我國的農(nóng)業(yè)機(jī)械化程度是較低的， 對(duì)我 國的糧食總產(chǎn)量增產(chǎn)貢獻(xiàn)十分低下。（二）政策建議1）首先，在短期內(nèi)為緩解糧食供應(yīng)緊張，應(yīng)提高農(nóng)民種糧的積 極性擴(kuò)大糧食耕種面積，這是增加糧食總產(chǎn)量的唯一辦法。 農(nóng)民積極 性主要取決于種糧食的收益及其預(yù)期，收益則是賣糧收入與成本的差 額。因此，應(yīng)該雙管齊下，穩(wěn)定并提高糧食價(jià)格，控制農(nóng)用物資價(jià)格 的過快增長，在涉農(nóng)物資上實(shí)行嚴(yán)格的價(jià)格管制，控制種糧的成

30、本。在提高農(nóng)民積極性的同時(shí)，也得以增加了化肥的施用量，在一定程度 上，影響糧食總產(chǎn)量的增產(chǎn)。但是，由于我國土地后備資源有限，且糧食耕種面積已占耕地總 面積較大比例（75%）,其調(diào)整幅度不大；在一定程度上是一個(gè)既定的前 提。從我國糧食生產(chǎn)的發(fā)展來看，總產(chǎn)量的增長主要取決于單位面積 產(chǎn)量的提高。而單位面積產(chǎn)量直接決定于農(nóng)戶的資本和勞動(dòng)投入，即農(nóng)戶的種糧積極性；同時(shí)受經(jīng)濟(jì)體制和政策、科技進(jìn)步狀況和市場環(huán) 境等強(qiáng)有力的影響。因此，我們一方面要堅(jiān)持最嚴(yán)格的耕地保護(hù)制度， 控制非農(nóng)業(yè)占 地，建立基本農(nóng)田保護(hù)區(qū)，確保基本農(nóng)田總量不減少、質(zhì)量不下降。 一方面要加強(qiáng)對(duì)現(xiàn)有耕地的開發(fā)，通過進(jìn)一步改進(jìn)耕作制度和應(yīng)用

31、優(yōu) 良品種，保持相對(duì)穩(wěn)定的糧食作物耕種面積，提高耕地利用效率。2 ）受邊際效益遞減規(guī)律的影響，化肥投入在糧食增產(chǎn)方面的能力逐漸下降；施肥方法落后、偏施和過施現(xiàn)象普遍存在，盲目增加化 肥施用量并不能從根本上使糧食增產(chǎn)，關(guān)鍵是要提高化肥的利用率。3）我國現(xiàn)在農(nóng)業(yè)機(jī)械化程度遠(yuǎn)遠(yuǎn)不能滿足現(xiàn)代農(nóng)業(yè)發(fā)展的需求， 要實(shí)現(xiàn)農(nóng)業(yè)現(xiàn)代化，必須在以下各方面積極穩(wěn)妥地推進(jìn)農(nóng)業(yè)機(jī)械化的 發(fā)展：要把主要農(nóng)產(chǎn)品生產(chǎn)過程機(jī)械化和產(chǎn)業(yè)化經(jīng)營有機(jī)結(jié)合起來；對(duì)農(nóng)業(yè)機(jī)械化進(jìn)行結(jié)構(gòu)性調(diào)整；因地制宜，有重點(diǎn)的推薦地區(qū)農(nóng)業(yè)機(jī)械化；大力促進(jìn)農(nóng)業(yè)技術(shù)進(jìn)步，重視農(nóng)村的基礎(chǔ)教育；建立與農(nóng)業(yè)機(jī)械化相適應(yīng)的農(nóng)村經(jīng)濟(jì)體制?？v觀中國農(nóng)村現(xiàn)狀，與其他產(chǎn)業(yè)相

32、比，農(nóng)業(yè)的發(fā)展一直處于較低 的狀態(tài)。擴(kuò)大耕作面積，提高單產(chǎn)，實(shí)現(xiàn)機(jī)械化、規(guī)?；a(chǎn)是保證 我國農(nóng)業(yè)健康發(fā)展的必有之路?！緟⒖嘉墨I(xiàn)】1、龐皓，計(jì)量經(jīng)濟(jì)學(xué)，西南財(cái)經(jīng)大學(xué)出版社，2014年6月第三版2、周四軍，對(duì)我國糧食生產(chǎn)影響因素的計(jì)量分析，統(tǒng)計(jì)與決策，2003年3、趙慧江，基于回歸分析的糧食產(chǎn)量影響因素分析，懷化學(xué)院學(xué)報(bào)，2009年4、呂美巧、馬廣，農(nóng)業(yè)機(jī)械化發(fā)展影響因素分析與評(píng)價(jià)，農(nóng)機(jī)化研究，2008年5、李妍，中國糧食生產(chǎn)影響因素及地區(qū)差異分析，經(jīng)濟(jì)研究導(dǎo)刊，2009年附錄1Dependent Variable: X1Method: Least SquaresDate: 12/22/15 T

33、ime: 09:24Sample: 1990 2013Included observations: 24VariableCoefficientStd. Errort-StatisticProb.C118379.68270.69614.313140.0000X2-3.5891343.927502-0.9138470.3712X30.1022010.1559860.6551940.5195R-squared0.090190Mean dependent var108857.3Adjusted R-squared0.003541S.D. dependent var3944.710S.E. of reg

34、ression3937.719Akaike info criterion19.51106Sum squared resid3.26E+08Schwarz criterion19.65832Log likelihood-231.1327Hannan-Quinn criter.19.55013F-statistic1.040870Durbin-Watson stat0.291206Prob(F-statistic)0.3706662Dependent Variable: X2 Method: Least SquaresDate: 12/22/15 Time: 09:14Sample: 1990 2

35、013Included observations: 24VariableCoefficientStd. Errort-StatisticProb.C3217.3121300.5632.4737840.0220X1-0.0106560.011661-0.9138470.3712X30.0384110.00186020.646100.0000R-squared0.956409Mean dependent var4360.633Adjusted R-squared0.952257S.D. dependent var981.9691S.E. of regression214.5609Akaike in

36、fo criterion13.69153Sum squared resid966764.0Schwarz criterion13.83879Log likelihood-161.2984Hannan-Quinn criter.13.73060F-statistic230.3753Durbin-Watson stat0.212818Prob(F-statistic)0.0000003Dependent Variable: X3 Method: Least SquaresDate: 12/22/15 Time: 09:15Sample: 1990 2013Included observations

37、: 24VariableCoefficientStd. Errort-StatisticProb.C-69567.6734359.53-2.0246980.0558X10.1960100.2991630.6551940.5195X224.812071.20178020.646100.0000R-squared0.955583Mean dependent var59965.75Adjusted R-squared0.951353S.D. dependent var24724.62S.E. of regression5453.262Akaike info criterion20.16228Sum

38、squared resid6.24E+08Schwarz criterion20.30954Log likelihood-238.9474Hannan-Quinn criter.20.20135F-statistic225.8983Durbin-Watson stat0.198222Prob(F-statistic)0.0000004Dependent Variable: YMethod: Least SquaresDate: 12/22/15 Time: 16:20Sample: 1990 2013Included observations: 24VariableCoefficientStd

39、. Errort-StatisticProb.C9080.94527337.040.3321850.7429X10.3696280.2509701.4728000.1550R-squared0.089748Mean dependent var49317.62Adjusted R-squared0.048373S.D. dependent var4867.060S.E. of regression4747.883Akaike info criterion19.84844Sum squared resid4.96E+08Schwarz criterion19.94661Log likelihood

40、-236.1813F-statistic2.169140Durbin-Watson stat0.116875Prob(F-statistic)0.154969Dependent Variable: YMethod: Least SquaresDate: 12/22/15 Time: 16:27Sample: 1990 2013Included observations: 24VariableCoefficientStd. Errort-StatisticProb.C31565.182691.44711.727960.0000X24.0710710.6027436.7542460.0000R-s

41、quared0.674652Mean dependent var49317.62Adjusted R-squared0.659863S.D. dependent var4867.060S.E. of regression2838.531Akaike info criterion18.81962Sum squared resid1.77E+08Schwarz criterion18.91779Log likelihood-223.8354F-statistic45.61984Durbin-Watson stat0.431896Prob(F-statistic)0.0000016Dependent Variable: YMethod: Least SquaresDate: 12/22/15 Time: 16:39Sample: 1990 2013Included observations: 24VariableCoefficientStd. Errort-StatisticProb.C39569.811530.63325.851930.0000X30.1625560.0236706.8676950.00