估计
直接方法：生产函数法
间接方法：测度增长、工资、生产率、地租

估计中的三大内生性问题

聚集经济的三大维度：产业、空间和时间


评论：
一个问题，这里讨论的聚集经济都是针对企业而言，没有考虑居民和消费者。

相关papers--产业、空间和时间维度

The Determinants of Agglomeration -空间维度（上传不上）

The Determinants of Agglomeration，2001

内容简报
We regress the Ellison-Glaeser measure of spatial concentration on industry characteristics that proxy for the presence of knowledge spillovers, labor market pooling, input sharing, product shipping costs, and natural advantage. Results indicate that proxies for labor market pooling have the most robust effect, positively influencing agglomeration at all levels of geography. Proxies for knowledge spillovers positively affect agglomeration only at the zipcode level. Reliance on manufacturing inputs or natural resources positively affects agglomeration at the state level but has little effect on agglomeration at lower levels of geography.

数据
The fourth quarter of 2000.
We utilized a more manageable and affordable version of the data set in which the data were aggregated up to the zipcode level.
459 4-digit manufacturing industries


指标
空间Gini系数        
G=∑_i▒〖(x_i-s_i)〗^2
EG 指数
γ=(G-(1-∑_i▒x_i^2 )H)/((1-∑_i▒x_i^2 )(1-H))
H=∑_i▒z_i^2  is a Herfindahl index of the J plants in the industry.
(1-∑_i▒x_i^2 ) is included in order that the index have the property that E(γ) =0 when neither agglomerative spillovers nor natural advantage are present.

自然优势：the variables Energy per shipment, Natural resources per shipment, and Water per $ shipment measure energy input cost, the cost of natural resources, and water-related costs respectively.


运输成本: 库存/交货产品


投入产出的关联: two variables

Manufactured inputs per of shipment is the ratio of the cost of inputs purchased from the manufacturing sectors to the value of shipments.

Nonmanufactured inputs per of shipment.


知识溢出: Innovations per of shipment. Innovations are defined as the number of new products advertised in trade magazines in 1982.


劳动力池: 3个指标
Net productivity equal to the value of shipments less the value of purchased inputs, all divided by the number of workers in the industry.
The ratio management workers/(management + production workers)
The percentage of workers with Doctorates, Master’s Degrees, and Bachelor’s Degrees.
            
估计方程
γ_(j,m)=βX_m+ε_(j,m)
m industry at the level of geography j.

Economic Structure and Local Growth: France, 1984–1993， 2000

内容简报
测度了经济结构与就业增长的关系。其中经济结构包含5个方面：专业化指数、多样化指数、就业密度、竞争性指数和企业平均规模。

数据
52 industrial and 42 service sectors.
341 spatial entities (employment areas)
1984-1993

指标：五大指标
y_(z,s)=log⁡(〖emp〗_(z,s,1993)/〖emp〗_(z,s,1984) )-log⁡(〖emp〗_(s,1993)/〖emp〗_(s,1984) )
Sector s at date t in ZE z.

All explanatory variables are considered at the initial data, 1984. All variables are taken in logarithms: the estimated parameters are their elasticities with respect to each variable, which makes them easily comparable.

Specialization index
〖spe〗_(z,s)=(〖emp〗_(z,s)/〖emp〗_z)/(〖emp〗_s/emp)

Diversity index
As Henderson et al. (1995), the inverse of an Herfindahl index of sectoral concentration based on the share of all sectors, expect the one considered. As previously, this variable is normalized by the same variable computed at the French level.
〖div〗_(z,s)=(1/∑_█(S=1@S≠S)^S▒〖(〖emp〗_(z,s)/(〖emp〗_z-〖emp〗_(z,s)))〗^2 )/(1/∑_█(S=1@S≠S)^S▒〖(〖emp〗_s/(emp-〖emp〗_s ))〗^2 )

规模指数
size_(z,s)=(〖emp〗_(z,s)/〖nbr〗_(z,s))/(〖emp〗_s/〖nbr〗_s )
〖nbr〗_(z,s)和〖nbr〗_s are the number of plants belonging to sector s in ZE z and France, respectively.

竞争指数
comp_(z,s)=(1/∑_(i∈z)▒〖(〖emp〗_(z,s,i)/〖emp〗_(z,s))〗^2 )/(1/∑_i▒〖(〖emp〗_(s,i)/〖emp〗_s)〗^2 )
Plant i.

密度
The total employment variable, which reflects the size of the local economy, is easy interpreted. In order to simultaneously control for the differences in ZEs’ area, it is here more relevant to consider the total employment density of the ZE,
den_z=〖emp〗_z/〖area〗_z


模型和方法：Tobit 模型
y_(z,s)=I+α_1  log⁡(〖spe〗_(z,s) )+α_2  log⁡(〖div〗_(z,s) )+α_3  log⁡(〖size〗_(z,s) )+α_4  log⁡(〖comp〗_(z,s) )+α_5 log⁡(〖den〗_z)
Sector s in zone z. I is an intercept.

Because plants that are smaller than 20 workers are not in the dataset, sectoral employment in ZEs where there are only small plants is not observed. Because of this truncation, OLS estimates are biased.
Heckman 估计

GEOGRAPHY,  INDUSTRIAL  ORGANIZATION,  AND AGGLOMERATION, 2003

文章内容
This paper addresses the geographic and organizational nature of agglomeration by examining the birth of new establishments and the employment levels that they choose.
Specifically, we estimate the determinants of the number of births per square mile and their associated employment levels as functions of the economic environment when the location decisions were made.

模型
To control for differences in ZIP code size, both births and new-establishment employment are deflated by ZIP code area and are interpreted as arrivals per square mile hereafter. Location and employment decisions are made at time t - 1, taking the existing economic environment as given, and establishments are born one period later at time t. We suppose that the local characteristics of each ZIP code, yj, are partitioned into two parts, yz,j and ymj. The elements of yz vary by ZIP code, whereas the elements of ym vary by metropolitan area. Aggregating over establishments gives the number of births per square mile (B) and total new-establishment employment (N) in ZIP code j, which we express as linear functions of yz, and ym.
B_(j,t)=b_z y_(z,j,t-1)+b_m y_(m,j,t-1)+ε_(b,t)
N_(j,t)=n_z y_(z,j,t-1)+n_m y_(m,j,t-1)+ε_(n,t)
使用固定效应处理都市区的效应.


因变量
1997年第4季度新建企业数量和他们的就业人口。

自变量
For each industry we calculate the existing level of employment both within and outside the industry in question.
第一个地理聚集的指标.
To measure the geographic extent of agglomerative externalities we create a set of concentric-ring variables for both types of employment. These variables are calculated as follows. First, employment in a given ZIP code is treated as being uniformly distributed throughout the ZIP code. Then, using mapping software, circles of radius ri, i=1,2,…15, are drawn around the geographic centroid of each ZIP code in the United States. The level of own-industry employment contained within a given circle is then calculated by constructing a proportional (weighted) summation of the own industry employment for those portions of the ZIP codes intersected by the circle.
The same procedure is used to calculate the level of other industry employment within each circle.
Differencing employment levels for adjacent cicles (by employment type) yields estimates of the levels of own- and other-industry employment within a given concentric ring. Thus, the 2 mile ring (r2) reflects employment between the 1 and 2 mile circles, and so on out to 15 miles.
第二个竞争力指数. The number of establishments per worker is calculated to proxy for local competitiveness. This variable is calculated separately for own-industry employment and employment outside of the own industry.
第三个指标多样化指数。Herfindahl index of employment by two-digit SIC industries as in Henderson et al. (1995).
Both the competitiveness variables and the Herfindahl index are calculated only at the ZIP code level.
第四个metropolitan-area fixed effects.  

产业: 6个
Software (SIC 7371, 7372, 7373, 7375)
Food processing (SIC 20)
Apparel (SIC 23)
Printing and publishing (SIC 27)
Fabricated metals (SIC 34)
Industrial and commercial machinery (SIC 35)

估计方法:Tobit模型,因为有大量0值
This raises a technical issue in that imprecise estimation of the fixed effects in nonlinear models typically leads to inconsistent estimates of the slope coefficients. In addition, tobit models are more sensitive to distributional assumptions than are linear regressions.
(1) the bias resulting from noisy estimates of fixed effects in nonlinear models goes to zero as the number of observations per fixed effect becomes arbitrarily large.
(2) our results are robust. Probit fixed effect model examines whether individual ZIP codes have positive or zero births. To facilitate review of the attenuation pattern, the partial derivatives based on the probit coefficients are reported.

评议
作者使用地理信息系统软件做分析很新颖,尚不知如何操作. 其仅仅将聚集经济变量考虑到ZIP之外,考虑外部性.其他变量都是ZIP内的数值.