【作者】 Julian Witjaksono；

【导师】 袁有禄；

【作者基本信息】 中国农业科学院 ， 农业资源与环境， 2014， 博士

【摘要】 转基因(GM)棉花自1995在美国开始商业应用,1996开始种植后,得到了迅速而广泛的应用,但有关其利敝的争论仍在进行中。自1996在美国的棉农第一次种植后,转基因棉花种植面积到2011达到24.7百万公顷。由于棉花价格降低,2012年全球转基因棉花的种植面积略有下降,也达到了24.3百万公顷。在全球3000万公顷的棉花种植面积中,有81%或2430万公顷是转基因棉花,分布在15个国家。种植面积超过100万公顷的三个最大的棉花生产国包括印度、美国和中国,种植面积分别为：印度1080万公顷、美国440万公顷、中国400万公顷。一般认为,且有许多科学家声称,有大量的证据表明转基因棉花提供的经济收益主要来自于产量的增加,减少农药使用和增加农民收入。本研究结合了系统回顾和Meta分析,证明了其对社会经济,特别是在经济指标中的影响。采用STATA12.1对系统性回顾进行Meta分析。此外,为了评估影响经济效益的因素,采用了SPSS20进行了多元回归分析。资料收集是从1998到2012之间的经过同行评议和没经过同行评议的数据中作为分析的数据库。成功地收集了129篇文章,这些文献至少包括一项经济指标(产量、净收益、种子成本,农药成本,管理和劳动力成本和收益)。从中我们又选出了53篇论文,作为Meta分析的数据库(46篇已成功地用于产量的Meta分析,25篇用于种子成本的Meta分析,28篇用于农药成本的Meta分析,20篇用于管理和劳动力成本的Meta分析,29篇用于净收益的Meta分析)。在做回归分析时,利用微软Excel2003进行数据制表,16个样品(制表的数量)的平均数据,包括所有的经济指标(产量,种子成本,农药成本,管理和劳动力成本和净收益)都进行了回归分析。把所有的国家(中国,印度,美国和澳大利亚)的数据作为一个数据库,利用Meta分析进行了综合的系统性回顾。该Meta分析提供的证据表明GM棉花和非GM棉花两处理间的差异显著性可通过I2的百分比表示出来。本文展示了产量、种子成本、农药成本、管理和劳动力成本和净收益,其I2的百分比分别是97.2%,78.4%,99.8%,97.17%和98.28%。这意味着试验之间由于研究方法、抽样方法、农民的特点、地域的依赖性、气候条件、和站点特异性等不同而使结果的差异非常大。这项研究对所有的经济指标在固定效应方法下,其结果在统计上是显著的,反映在P值<0.05。在回归分析中,总的结果表明,R2值为0.439,表明转基因棉花的净收益(因变量)和预测指标(产量,种子成本,农药成本,管理和劳动力成本)的线性关系并不非常显著。同时我们发现常规棉R2值0.88,表明两者之间显著的线性关系。方差分析(ANOVA)表明,常规棉的P值为0(P<0.05),达到极显著水平；而在转基因棉花P值0.032(P<0.05),这也表明,转基因棉花和其对应的常规棉间不存在多重共线性问题。通过Durbin Watson (DW)值也检测了自相关的问题,表明转基因棉和非转基因棉花没有自相关的问题,其值分别为1.446和2.1。利用另一项测试Vif(方差膨胀因子)也检测了多重共线性问题,总体结果表明转基因棉花模型和非转基因棉模型的Vif值都小于10,说明两个模型没有多重共线性的问题。转基因棉花的56%的线性关系,可以由异常的解释变量来说明,而常规棉的12%的线性关系,可以由异常变量来解释。从统计上推论,本研究表明,转基因棉花的风险高于常规棉,因此,除了经济指标以外,种植GM的农民也应该考虑所有其它的影响因素。Meta分析表明转基因棉和常规棉间的产量和净收入差异具有统计学上的显著性,因此,本项研究表明,我们相信转基因棉与常规棉在经济利益方面相比,还是可以给棉农带来可观的效益。更多还原

【Abstract】 The benefits of genetically modified (GM) cotton continue to be disputed, despite rapid and widespread adoption since their commercial introduction in the United States in1995and first planted in1996. Since its1996debut on U.S. cotton farms, biotech cotton reached24.3million hectares in2012down from the24.7million hectares grown in2011. With lower global price of cotton, the area planted to biotech cotton globally in2012was down by half a million hectares from a record24.7million hectares in2011. Based on a global hectarage of30million hectares,81%or24.3million hectares, were biotech cotton and grown in15countries. Three largest countries grew more than1.0million hectares, in descending order of hectarage, they are:India (10.8million hectares), USA with (4.4million hectares), and China (4.0million). There is a general believe and substantial evidence that many scientists claimed the transgenic cotton provides economic benefits. These benefits mostly came from the yield gain, reducing chemical spray and rising farmers income. This study combined a systematic review and meta-analysis to proof the evidence on the socio economic impacts particularly in economic indicators. Statistical analysis of systematic review for meta-analysis was done using STATA12.1. Moreover, In order to estimate the factors influence the economic benefits, multiple regressions have been analyzed by using SPSS20.0. Data were collected from the peer-reviewed and non-peer-reviewed between1998and2012as the database set which successfully take into account as129papers which at least consists of one of the economic indicators (yield, net return, seed cost, pesticide cost, management and labor cost and net return). From that we selected53papers which should be considered into the database for meta-analysis (46papers have been successfully considered into the yield meta-analysis,25papers for seed cost meta-analysis,28papers as pesticide cost meta-analysis,20papers as management and labor cost meta-analysis, and29papers as net return meta-analysis). Then for regression analysis the data were tabulated by using Microsoft excel2003and16samples (number of tabulation) of the average data which consist of all the economic indicators (yield, seed cost, pesticide cost, management and labor cost and net return) were considered for regression analysis. An expanded a systematical review was applied by using meta-analysis through the database all the countries (China, India, USA and Australia). This meta-analysis provides the evidence that the significance of the study of the treatments of GM cotton and non GM cotton indicated by the percentage of I2This presented of yield response, seed cost, pesticide cost, management and labor cost and net return which is shown by the value of I2=97.2%,78,.4%,99.8%,97.170and98.28%, respectively. This means that the differences between trials are very large because of the variability of research methodologies, sampling method, farmer’s characteristic, geographical dependent, climatic conditions, and site specific. The results of the study are statistically significant which reflected by the p value<0.05under the fixed effect method for all the economic indicators. In regression analysis, overall results show that R2value of GM cotton is0.439which indicated that the liner relationship between net return (dependent variable) and predictors (yield, seed cost, pesticide cost, management and labor cost) is not highly significant. Whilst we detected that R2value of conventional cotton is0.88which revealed that the linear relationship is highly significant. Analysis of Variance (ANOVA) shows the significance ofp value0.00(P<0.05) in conventional cotton which is highly significant, whilst in the GM cotton show p value0.032(P<0.05). This also indicated that there is no multicollinearity problem with the model of GM cotton and its counterpart. Autocorrelation problem also was detected by the Durbin Watson (DW) value which indicated both the GM cotton and non GM cotton have no problem with autocorrelation which has value of1.446and2.1, respectively. Another test of multicollinearity has been tested by the VIF (Variance Inflation Factor) indicated that overall results have the VIF value less than10both GM cotton model and non GM cotton model. That is both model have no multicollinearity problem.56%the linear relationship of GM cotton can be explained by the outlier explanatory variable, whilst12%the linear relationship of conventional cotton can be explained by the outlier variable. From the statistical inferences, this study suggested that GM cotton prone to be higher risk than its conventional. Therefore, GM farmers should consider any factors beyond the economic indicators. Meta-analysis reflected the yield differences and net revenue differences between GM cotton and conventional cotton is statistically significant.Thus, we believe that GM cotton is comparable with conventional cotton in terms of economic benefits and bring sizeable benefits for cotton farmers including environmental effect.更多还原