Abstract: The conventional spatial homogeneity hypothesis is unable to account for the extent to which population agglomeration affects labor productivity in the context of spatial heterogeneity in China. Using provincial panel data from 2000 to 2012，a heterogeneity test is performed with spatial panel regression and threshold panel regression to investigate the non-linear relationship between population agglomeration and labor productivity. Population agglomeration tends to improve labor productivity but with marked regional differences，showing an inverted U shape relationship which has triple nonlinear threshold characteristics.Differentiated Population policies need to be carried out for different cities according to their stage in the inverted U shape relationship. During the urbanization of Midwest China，education policies are especially important to address the loss of labor productivity due to the excessive agglomeration of the homogeneous lowgrade labor force.

Keywords: Population Agglomeration, Labor Productivity, Heterogeneity, Panel Threshold Regression

摘要： 中国区域间的空间异质性使得传统的空间同质性假设难以合理解释人口聚集对劳动生产率的影响究竟几何。根据2000 ～ 2012 年的省级面板数据，利用空间面板回归和门槛面板回归进行异质性检验，考察人口聚集度与劳动生产率之间的非线性关系。研究表明，人口聚集度的上升具有提高劳动生产率的效应，但表现出地区差异: 人口聚集对提升劳动生产率的效应呈东、中、西部递减态势; 人口聚集度与劳动生产率之间呈“倒U”形关系，并具有显著的三重非线性门槛特征，印证了“威廉姆森假说”。文章政策涵义明显: 实施分类指导的人口政策，对于处于“倒U”形不同阶段的城市制定不同的人口政策。同时中西部地区城镇化过程中尤其要注重提升教育水平和吸引人才，以改善同质低端劳动力过度聚集带来的劳动生产率损失。

关键词: 人口集聚, 劳动生产率, 异质性, 面板门槛回归