中文摘要：

基于中国期刊全文数据库中的30篇中文社会科学引文索引（CSSCI）的期刊论文，运用Meta方法定量分析FDI（外商直接投资）对中国环境污染的影响．结果表明：①单一或几个环境污染指标和FDI相对值的回归系数分别为-2．72和-1．83，说明采用二者会降低FDI环境效应显著性，但后者不显著；②面板数据和样本时间跨度的回归系数分别为3．19和1．52，说明采用二者会提高FDI环境效应显著性，并更有可能得到FDI加剧了环境污染的结论；③样本数据平均年限变量的回归系数为-0．90，说明采用2000年后数据会降低FDI环境效应显著性，2000年前FDI对环境污染加剧效应呈下降趋势，此后FDI改善了环境质量；④OLS法和固定效应模型的回归系数分别为2．22和0．78，说明采用二者会提高FDI环境效应显著性，但后者不显著．综上可知，在纳入部分控制变量前提下，采用环境污染综合指数和FDI水平值，基于1985--2010年面板数据，运用OLS法和固定效应模型实证发现，FDI更有可能显著加剧了环境污染，但这种效应呈下降趋势，以至于2000年后FDI改善了环境质量．

英文摘要：

The paper quantitatively analyzed the effect of FDI on environmental pollutions in China based on thirty papers collected from the Chinese Social Science Citation Index in China journal full text database by using Meta regression analysis. The analyzed results showed that the regression coefficients of the single or a few environmental pollution index and FDI relative value are - 2.72 and - 1.83. The use of the indices can significantly decrease effects of FDI to environment. However, for the latter there was not significant effects. The regression coefficients of panel data （provincial or urban panel data） and the sample data span were 3.19 and 1.52. The use of the coefficients will improve significant effect of FDI to environment, and it was more likely to get FDI aggravates the environmental pollution. The regression coefficient of the average age of sample data was - 0.90, and it can significantly reduce the effect of FDI to environment by using the data after 2000. The environmental effects of FD! decreased before 2000 and increased after 2000. The regression coefficients of OLS method and fixed effect model were 2. 22 and 0.78. The OLS method significantly improved the environmental effects of FDI, while the fixed model did not. Based on the analyzed results, it can be concluded that under the premise of including control variables, FDI is more likely to aggravate the environment pollution, but the environmental pollution effect of FDI showed a downward trend and FDI improved the quality of the environment after 2000 based on panel data of 1985-2010.