Edward L. Glaeser, Hedi D. Kallal, Jose A. Scheinkman, Andrei Shleifer. Growth in Cities. 1992.

简要内容

作者讨论了三种增长的外部性理论，MAR、Jacobs和Porter的外部性。并使用美国都市区内六大产业，进行了经验研究。结果支持Jacobs外部性理论。

这三种理论在两个方面具有不同的侧重点：聚集经济来源于产业内部还是产业之间？聚集经济来源于竞争还是垄断？

作者对计量模型的处理以及稳健性检验值得学习。其理论模型来源于生产函数，亮点在于对A的处理。并在模型中加入了初始条件等。稳健性检验中，处理考虑不同的产业划分，还考虑更换指标。

三个指标：专业化指数（区位商）、竞争指数和多样化指数。

控制变量

The 1956 log of wage

The log of employment in the city-industry

A dummy variable indicating a southern city

The national employment growth in that industry

National employment changes in the industry

Some analysts have argued that firms move to low-wage areas (or workers move to high-wage areas). Including the 1956 wage controls for either of these effects, even though this control variable is not strictly consistent with the assumption of a national labor market.

High observed initial employment reduces employment growth because of either measurement error or more serious economic factors. The MAR view is somewhat incompatible with the presence of real mean reversion, but since we are not correcting for potential measurement problems. We do not use such mean reversion as evidence against the MAR externalities.

稳健性

去除mining industries

将产业分为制造业和服务业

将制造业分为ubiquitous industries (fabricated metals, nonferrous metals, nonelectrical machinery, etc.) and more specialized industries (electrical equipment, transport equipment, primary metal, pulp and paper, textiles and apparel, leather products, etc.).

没看懂的

We included in each city the largest counties that cover the SMA until their combined payroll added up to at least 80 percent of the total payroll of the SMA in both 1956 and 1987.

对1956和1987年就业缺失数据，采取了两种不同的方法。不知道用意何在？为什么没有用同样的方法。

评论

MAR在很多产业是很难的，比如汽车。汽车业内各个企业存在较为强烈的竞争关系，不太可能把知识告诉你。

MAR认为产业内聚集会带来收益，但又认为这种外部性收益会降低创新的积极性，因为需要对知识进行垄断。我认为这两点在很大地方可能是矛盾的。因为垄断往往意味着企业的规模变大，这样就不会带来多少产业内聚集的收益。

Jacobs外部性，是不能受地域限制的，尤其是现在全球一体化世界。像锤子手机，其影像处理是与日本富士通公司合作的，产品的设计是跟美国的公司合作的。等等。这是一个全球化的社会，其经济联系不能只考虑本地。但受限于数据，又没有好的处理办法。

产业增长的原因不仅仅有溢出，还有更为重要的研发。这一点作者没有考虑。

专业化城市的出现，作者使用Marshall外部性解释，认为这种外部性解释了专业化，但是没有解释增长。读起来有些牵强。

影响就业增长的因素不仅仅是供给因素，还有需求因素。 这意味着需要建立 联立方程，而不是单个方程。

简要内容
作者讨论了三种增长的外部性理论，MAR、Jacobs和Porter的外部性。并使用美国都市区内六大产业，进行了经验研究。结果支持Jacobs外部性理论。Using a new data set on the growth of large industries in 170 U.S. cities between 1956 and 1987, we find that local competition and urban variety, but not regional specialization, encourage employment growth in industries.
这三种理论在两个方面具有不同的侧重点：聚集经济来源于产业内部还是产业之间？聚集经济来源于竞争还是垄断？
作者对计量模型的处理以及稳健性检验值得学习。其理论模型来源于生产函数，亮点在于对A的处理。并在模型中加入了初始条件等。稳健性检验中，处理考虑不同的产业划分，还考虑更换指标。

三个指标
s=(industry employment in city /total employment in city )/(industry employment in U.S./total employment in U.S.)

competition=(firms in city-industry /workers in city-industry )/(firms in U.S.industry /workers in U.S.  industry)

a variety of industries in the city is the fraction of the city’s employment the largest five industries other than the industry in question account for in 1956.


控制变量
The 1956 log of wage
The log of employment in the city-industry
A dummy variable indicating a southern city
The national employment growth in that industry
National employment changes in the industry


Some analysts have argued that firms move to low-wage areas (or workers move to high-wage areas). Including the 1956 wage controls for either of these effects, even though this control variable is not strictly consistent with the assumption of a national labor market.
High observed initial employment reduces employment growth because of either measurement error or more serious economic factors. The MAR view is somewhat incompatible with the presence of real mean reversion, but since we are not correcting for potential measurement problems. We do not use such mean reversion as evidence against the MAR externalities.

稳健性
去除mining industries
将产业分为制造业和服务业
将制造业分为ubiquitous industries (fabricated metals, nonferrous metals, nonelectrical machinery, etc.) and more specialized industries (electrical equipment, transport equipment, primary metal, pulp and paper, textiles and apparel, leather products, etc.).

没看懂的
We included in each city the largest counties that cover the SMA until their combined payroll added up to at least 80 percent of the total payroll of the SMA in both 1956 and 1987.

对1956和1987年就业缺失数据，采取了两种不同的方法。不知道用意何在？为什么没有用同样的方法。


评论
MAR在很多产业是很难的，比如汽车。汽车业内各个企业存在较为强烈的竞争关系，不太可能把知识告诉你。
MAR认为产业内聚集会带来收益，但又认为这种外部性收益会降低创新的积极性，因为需要对知识进行垄断。我认为这两点在很大地方可能是矛盾的。因为垄断往往意味着企业的规模变大，这样就不会带来多少产业内聚集的收益。
Jacobs外部性，是不能受地域限制的，尤其是现在全球一体化世界。像锤子手机，其影像处理是与日本富士通公司合作的，产品的设计是跟美国的公司合作的。等等。这是一个全球化的社会，其经济联系不能只考虑本地。但受限于数据，又没有好的处理办法。
产业增长的原因不仅仅有溢出，还有更为重要的研发。这一点作者没有考虑。
专业化城市的出现，作者使用Marshall外部性解释，认为这种外部性解释了专业化，但是没有解释增长。读起来有些牵强。
影响就业增长的因素不仅仅是供给因素，还有需求因素。 这意味着需要建立 联立方程，而不是单个方程。