摘 要最近几十年以来，生产者服务业在国际大都市中的地位越来越突出，逐渐取代制造业成为了推动城市经济增长的新引擎。鉴于此，生产者服务业成为了学者和政策决策者们关注的焦点。生产者服务业的空间集聚与演化作为研究生产者服务业的重要视角，受到了经济学者和经济地理学者的重点关注。国外对生产者服务业的研究已经比较成体系，但总体来看，以研究西方发达地区的文献为主，对发展中国家案例关注不多；国内的相关研究起步较晚，只有少数文献涉及到北京、上海、广州等大都市，很多研究还有待深入。北京作为一个崛起中的国际化大都市，其生产者服务业的发展历程在发展中国家中具有很强的代表性，理应得到更多的关注。鉴于此，本文以北京市的企业数据为基础，结合历年的统计年鉴，综合运用 Arcgis 软件和 R 语言平台，采用核密度估计和基于距离的 Kd函数对 1996 年、2001 年和 2010 年北京市生产者服务业的空间分布格局和不同空间尺度上的集聚程度进行了测度，并通过对不同时期的纵向比较，分析了生产者服务业的空间变化；最后通过构建回归模型，对北京市生产者服务业空间分布的影响因素进行了分析。主要结论如下：（1）在生产者服务业空间分布的变化方面，从 1996 年到 2010 年，北京生产者服务业发展迅速，空间分布范围不断扩展，在空间分布上仍表现为集中在中心城区的“单中心”格局，但就五环内具体来看，有些行业出现了“双中心”，甚至“多中心”。生产者服务业整体呈现出由明显的“单中心”到“主中心”与“副中心”共存的变化特征，空间分布趋势为围绕集聚中心呈圈层递减扩展，核心区域空间集聚趋势显著增强。生产者服务业内部各行业的变化规律虽然不尽相同，但总体上来看，在时间跨度上，基本上都呈现出空间集聚程度由弱到强、空间分布结构由简单趋于复杂的特点。（2）在生产者服务业空间集聚的变化方面，从 1996 年到 2010 年，北京市生产者服务业内部各行业大体上都朝着越来越集聚的方向发展，在较小的空间距离上（0-10 千米），这种日趋集聚的势头更加明显，但不同行业的变化并不完全一致。在 1996、2001 和 2010年这三年，信息传输、计算机服务和软件业的全局 Kd 指数都显著集聚，显著集聚的范围明显扩大，而集聚程度出现了先减小后增大的趋势，这与其空间分布的变化一致。在较小的空间距离上，金融业的空间集聚指数经历了从显著分散到集聚/分散情况不显著，再到显著集聚的变化，这与金融业空间分布的变化规律基本相吻合。租赁与商务服务业在较小距离上的变化规律比较明显，经历了从显著集聚到集聚/分散情况不显著，再到显著集聚的变化过程。与其他 3 个行业相比较，科学研究技术服务和地质勘查业的空间集聚变化则比较特殊，在这 3 年中，该行业在较小距离上是显著集聚，但是 2010 年与 2001 年相比较，显著集聚的范围减小了。（3）对生产者服务业空间变化的解释方面，回归模型的结果显示，集聚因素、政策因素和区位因素对北京市生产者服务业的空间分布有着不同程度的影响。集聚因子的正向影响最大，显著性最强；政策因素、学校和科研院所对生产者服务业企业的布局有吸引作用，但政策因素的促进作用更显著；街道的交通可达性越好，街道内企业越多；在集聚因子和交通因子的共同作用下，街道地价等级越高，企业越多的趋势并不明显。以上这些因素同时存在并相互作用，共同影响了北京市生产者服务业的空间分布格局的形成。

In last decades, the producer services have gradually replaced manufacturing industriesand become a new engine of the economic development in the metropolis.In view of this,producer service industry has become the focus of attention of academics and policy makers. Asan important research perspective of producer service, spatial agglomeration and evolutionarywere continuing interested by economists and economic geographers. In western developedcountries, the study concerning producer services become mature gradually, however, in mostliteratures, western developed regions were often the key concerns. On the other hand, domesticresearch was relatively less, there were only a handful papers comes to Beijing, Guangzhou andShanghai. As a rising international metropolis, Beijing is a prefect case for study producerservice industry in developing countries.The firm-level data is the basis of this paper, which including all the producer servicesenterprises in Beijing metropolitan area in 1996, 2001 and 2010. Arcgis software and R languageare the data processing tools. In this context, kernel density estimation and K-density index areused to test the spatial agglomeration of producer service industries. After comparing the spatialpattern in different years, this paper summarizes the space evolution of the producer services.Main conclusions are as follows:The first conclusion is concerning about the evolution of the spatial distribution of producerservices. From 1996 to 2010, producer service industry in Beijing has developed rapidly, and thespace distribution was expanding significantly. The spatial patterns are distinguishing amongdifferent producer service industries，and monocentric，double center，and polycentric spatialstructure coexist in this city. In this period, the evolution characteristics of spatial distribution ofthe whole producer service industry experienced the single-nuclei to the double-nuclei, then tothe multiple-nuclei. In this time span, though the spatial patterns are distinguishing amongdifferent producer service industries, the spatial agglomeration degree become more and more strong, and the spatial distribution become complicated are present in all the industries.The second conclusion is about the evolution of the producer service agglomeration. From1996 to 2010, information transmission, computer services and software enterprise's globalK-density index were all significant agglomeration. In small distances, the concentrationsignificantly expanded obviously, distance of the concentration peak appeared a trend that thisindustry decreasing at the first, and then increasing. As far as finance is concerned, its spatialagglomeration index on the smaller distance has experienced from significant dispersed toagglomeration or dispersion not significant, and then to a present that concentration issignificantly, which is consistent with the evolution of the spatial distribution of the financialsector. Tenancy and commercial services’ changing is more noticeable, which experienced fromsignificant agglomeration to no significant spatial agglomeration or dispersion, and then presenta pattern of significant spatial agglomeration. Compared with other three industries, scientificstudies, technical services and geological prospecting is more special, in these three years, thisindustry was significant agglomeration in the smaller distances, but in 2010, comparing with2001, the scope of significantly localization was reducing.The last conclusion that involves the interpretation of the producer services’ space evolutionis important. The regression model shows that agglomeration factor, policy factor and locationfactor have varying degrees of influence in the evolution of spatial distribution of the producerservice industry in Beijing .Agglomeration factor is most statistically significant and has thebiggest positive influence on the evolution process. Policy factor, universities and scientificresearch institutes will enhance the attractiveness of blocks, and attracting more firms comearound. However, policy factor is more important than R&D institutes. Traffic accessibility alsoplay a positive role in attracting enterprises. Agglomeration factor and traffic accessibilityweaken the importance of land price in producer services’ space evolution. All in all, theinteraction of agglomeration factor, policy factor and location factor promote the evolution ofproducer service industry’s spatial pattern in Beijing metropolitan area.