城市化进程和经济增长中英文外文文献2017年

本科毕业设计（论文）

中英文对照翻译

(此文档为word格式,下载后您可任意修改编辑!)

文献出处：Davy Haw. The urbanization process and economic growth:

The so-what question[J]. Journal of Economic Growth, 2017, 1(8): 47-71.

原文 The Urbanization Process and Economic Growth: The

So-What Question

Davy Haw

Introduction

There is an enormous literature on the urbanization process that

occurs with development (see Davis and Henderson, 2003 for a review).

There are two key aspects to the process. One is urbanization itself and

the other is urban concentration, or the degree to which urban resources

are concentrated in one or two large cities, as opposed to spread over

many cities. Part of the interest in the urbanization process arises because

urbanization and growth seem so interconnected. In any year, the simple

correlation coefficient across countries between the percent urbanized in a

country and, say, GDP per capita (in logs) is about 0.85. The reason is

clear. Usually economic development involves the transformation of a

country from a rural agricultural based economy to an industrial service

based economy (as well as releasing labor from agriculture, as

labor-saving technologies are introduced). That transformation involves

urbanization, as firms and workers cluster in cities to take advantage of

Marshall's (1890) localized external economies of scale in manufacturing

and services (Henderson, 1974; Fujita and Ogawa, 1982; Helsley and

Strange, 1990; Duranton and Puga, 2001). Economists have tended to

focus on the issue of urban concentration, rather than urbanization per se.

The literature that does exist on urbanization examines rural versus urban

bias in the transformation process. Governments may favor the

urban-industrial sector with trade protection policies, infrastructure

investments, or capital market subsidies or they may discriminate against

the rural sector with agricultural price controls (Renaud, 1981; O, 1993),

both leading workers to migrate to cities. But there can be a bias towards

inhibiting urbanization. For example, former planned economies tend to

exhibit a rural bias, in the sense of discouraging rural-urban migration,

but not necessarily industrial development (Ofer, 1977; Fallenbuchl,

1977). The more extensive literature on the degree of urban

concentration and changes in that degree which occurs as urbanization

and growth proceed has a variety of strands. Countries and international

policy officials worry about whether key cities are too big or too small

(Renaud, 1981; UN, 1993; WDR, 2000) and over the years various

countries such as Egypt, Brazil, Korea, Mexico, and China have pursued

medium size city programs designed to forestall the growth of larger

cities (Henderson, 1988; Ades and Glaeser, 1995). International agencies

presume that many of the world's mega-cities are overpopulated, at

considerable cost to those economies. The UN (1993) asks how bad "the

negative factors associated with very large cities" need to get "before [it

is in the] self interest of those in control to encourage development of

alternative centers." The same report warns of "unbalanced urban

hierarchies" and the crime, congestion and social inequality in

mega-cities. The World Development Report (2000) has a chapter (7) on

the grim life of people in mega-cities in developing countries. And the

Economist in one of its special surveys has posed the question directly

(July 29, 1995): Do the splendors of large cities outweigh their dark side?

The Effects of Urban Concentration on Growth Development In

this section, I examine the effect of urban concentration on productivity

growth. I start with urban concentration, or primacy, because that

examination yields the key results. The examination also develops the

methodology that is then applied to the examination of the effect of

urbanization on growth. The first issue is how to measure urban

concentration. There are three measures that people use. First, Wheaton

and Shishido (1981) and Henderson (1988) use the standard

Hirschman-Herfindahl index of concentration which in an urban context

is the sum of squared shares of every city in a country in national urban

population. Second, Rosen and Resnick (1981) use the Pareto parameter

looking at the distribution of city sizes within a country, which measures

how quickly size declines as we move from top to bottom in the size

distribution, or the overall degree of disparity in the size distribution. In

these papers, both measures were constructed for just one year for a

limited sample of mostly larger countries in the world; they are not

available for a larger group of countries over the time span that we look at,

1960-1995.

The key question is why urban concentration affects productivity

growth. Losses from excessive or deficient primacy in static urban

models come from GDP losses from resource misallocation, where, for

example, under excessive primacy where urban development is

concentrated in just one or two primate cities, these cities are subject to

exhausted scale economies, excessive congestion, and excessive per

capita infrastructure costs, while smaller cities have unexploited scale

economies and often deficient capital investment (e.g., Tolley et al., 1979;

Fujita, 1989; Henderson and Becker, 2000; Au and Henderson, 2002). In

Black and Henderson (1999) building on Lucas (1988), in an endogenous

growth model of a system of cities, city size affects positively the degree

of local information spillovers, which interactively affects local

knowledge accumulation, promoting productivity growth. However, cities

of excessive size draw resources away from investment and innovation in

productive activity to try to maintain quality of life in a congested local

environment.

From the urban literature, there are promising micro-foundations

for these ideas in Duranton and Puga (2001). In that paper, primate cities

are urban areas of experimentation, in deriving appropriate product

designs. Relatively under-sized primate cities result in environments that

have too little experimentation, affecting productivity nationally.

Relatively over-sized primate cities have people devoting excessive

amounts of time to commuting and other "wasteful" activities, drawing

resources away from experimental activity. In principle, one could adapt

the Duranton and Puga dynamic model to a growth context where

under-concentration in an economy results in lower knowledge

accumulation due to lack of experimentation and over-concentration

siphons resources away from experimental activity, similarly inhibiting

productivity growth. Then primacy affects growth in a continuous

non-linear fashion. But in this context, given the Williamson (1965)

hypothesis, we would expect the effect of urban concentration to depend

on a country's level of development, representing national scarcity of

knowledge accumulation and economic infrastructure. All these

statements cry out for a comprehensive growth model that captures these

specific considerations, but that is simply beyond the scope of this paper.

Whatever the precise model, the empirics with cross-country data are

going to come down to asking the so-what question-to what extent does

primacy affect growth?

Basic Primacy Results

With these results in mind, I now turn to the primacy variable. The

raw data do not tell us much. There is a modest negative correlation

between either d ln（Y=N） and primacy or dd ln （Y=N）and dprimacy.

Controls and a non-linear specification to the effect of primacy are

needed to sort out what is going on.

The basic econometric results are in Table 2, columns (1)-(4)

where there is a quadratic form to primacy and it is interacted with output

per worker to allow best primacy to vary with output per worker. Before

analyzing those results, I note that a simple linear primacy term has a

negative coefficient. Second, in columns (5) and (6) of Table 2, I report

on a simple quadratic, to make the point that there is a best degree of

primacy. In columns (5) and (6) and in all other reported results in the

paper, OLS and instrumental variable results on primacy do differ. OLS

tends to give lower best primacy values with less curvature to the f(.)

function. In column (5), under OLS the best primacy value is 0.20, while

under instrumental variables estimation (GMM) the best degree primacy

has a high point estimate of 0.46, with strong and significant coefficients.

From the best level of primacy, a one-standard deviation (0.13) increase

in primacy leads productivity growth to be 0.12 less over five years, a

huge effect, albeit for a large change in primacy. However, the best

degree of primacy should vary with level of development, under the

Williamson hypothesis.

The Effect of Urbanization on Growth

Examining the effect of urbanization on productivity is difficult, in

the sense of the ability to isolate meaningful results. I start by discussing

three reasons for this difficulty. First, rapid urbanization in African

countries in particular over the last 30 years has occurred in the face of

negative and low-income growth. This in itself suggests urbanization is a

result of a variety of factors related to changes in national output

composition and social conditions, not a force promoting growth per se.

Second, urbanization is a transitory process, where with economic growth

all countries eventually "fully urbanize". At some middle-income level,

urbanization tops off or ceases when a country is in the 65-85 percent

urbanized category; and almost 50 percent of our countries fall into a

fully urbanized category by 1990. Finally, urbanization definitions vary

widely across countries, making it very difficult to quantify any best

degree of urbanization, since that would depend on how the country

counts urban.

Focusing on the definition of urbanization for a moment, fully

urbanized for Switzerland, Austria and Finland means 60-65 percent

urbanized; for the USA it is just over 70 percent (with minuscule full-time

employment in agriculture); and for countries like Argentina, Chile, and

Brazil, fully urbanized is 80-85 percent urbanized. A lot of these

differences depend on how low density non-agricultural populations are

treated in defining urban, especially around the fringes, or ex/peri-urban

areas of metropolitan areas. For example, while China is officially 30

percent urbanized, about 70 percent of its population live within

"municipal" boundaries (jurisdiction of the city). With these problems

in mind, I econometrically explore the relationship between growth and

urbanization. As with primacy, we hypothesize that for any income level,

there is a best degree of urbanization. Even if "urbanization promotes

growth", presumably no one would argue that low-income countries, with

high degrees of semi-subsistence farming and high illiteracy rates, should

switch to being fully urbanized over night.

To examine the urbanization-growth, I use an function of the form

corresponding to equation (4) and columns (2) of Table 3, although the

national scale variable is unimportant. In this formulation, or any other,

there are no significant results to the f（.） function for the whole sample.

To get any results with an optimal degree of urbanization, it is necessary

to restrict the sample to potentially urbanizing countries. Here I define

that as the set of countries period by period that are less than 70 percent

urbanized; an alternative restriction is to eliminate all countries that are

high-income in 1965. For this restricted sample, OLS and instrumental

variable (GMM) results are reported in Table 6, columns (1) and (2). The

instrumental variable results suggest (1) country size is not a factor in

determining an optimal degree of urbanization and (2) that there is

potentially an optimal degree of urbanization. But beyond that, the results

are perverse, in the sense that the effect of output per worker growth is to

reduce the ``best degree'' of urbanization. That is the best degree of

urbanization declines, as output per worker rises, a completely

implausible result.

Moreover, results deteriorate when I put urbanization and primacy in

the same estimating equation. For that estimation, I interact national scale

with primacy but not urbanization given results in column (2) of Table 6

and column (2) of Table 3. Results are in column (3) of Table 6. While the

basic best primacy patterns persist, the notion that there is a best degree

of urbanization evaporates, albeit in a much more limited sample size in

estimation (requiring for any period-country urbanization 5 0.7 and for

primacy data to be available). In those instrumental variable results in

column (3) of Table 6, in fact, urbanization would appear to have a

negative effect on growth over most output per worker ranges. In

summary, these results suggest that urbanization per se, at least as

measured across countries, does not directly affect productivity growth.

Conclusion

This paper argues that urbanization represents sectoral shifts within

an economy as development proceeds, but is not a growth stimulus per se.

However, the form that urbanization takes, or the degree of urban

concentration, strongly affects productivity growth. Urban concentration

is affected by national policies and institutions, reflecting the extent to

which a particular city (e.g., a national capital such as Bangkok or

Mexico City) is favored. For any country size and level of development,

there is a best degree of urban concentration, which balances the gains

from enhanced concentration such as local knowledge accumulation

against the losses such as resources diverted to shoring up the quality of

life in crowded mega-cities. That best degree of concentration declines

with country size and level of development.

译文城市化进程和经济增长研究

Davy Haw

引言

大量的文献对城市化进程的发生与发展作出了研究(戴维斯和亨德森,2003年)。城市化的这一过程包含两个主要方面。一是城市化本身,另一个是城市集中,或者叫做城市资源在一个或两个大城市的集中程度,而不是分布在许多城市。因为城市化和经济增长之间联系紧密，人们对城市化过程研究的兴趣就愈发浓厚。在任何一年, 各国城市化比率与人均国内生产总值约为0.85。其原因是显而易见的。通常情况下，经济发展涉及一个国家从以农村农业经济为基础向工业服务经济为主的转变(以及从农业中释放劳动力,引进节省劳力的技术)。

这一转换涉及城市化，在城市的公司与工人集群能充分利用马歇尔(1890)关于制造业与服务业的外部规模经济 (亨德森,1974; 藤田昌久和小川,1982;赫舍利和斯特兰奇,1990;丹顿与普加,2001)。经济学家倾向于关注城市集中的问题,而不是城市化本身。有关城市化的文献考察了农村对抗城市的转换过程。政府可能更倾向于城市工业部门与贸易保护政策,基础设施投资,或资本市场补贴，相反，他们会利用价格管制农村与农业 (雷诺,1981;吉奥,1993),从而引导员工迁移到城市。但这种做法可能会抑制城市化的发展。例如,前计划经济往往表现出一种对农村农业的偏见,从这个意义上讲即为令人沮丧的乡-城迁移,但不一定会阻碍工业发展(奥弗,1977;Fallenbuchl,1977)。

关于城市化的文献更广泛地集中在研究城市集中的程度和发生

变化的程度上。国家和国际政策官员担心重点城市规模是否太大或太小(雷诺,1981;联合国,1981;《世界发展报告》,2000年)多年来，不同国家如埃及、巴西、韩国、墨西哥和中国一直致力于建设中等城市项目旨在防止大城市的发展(亨德森,1988;阿德斯和格莱泽,1995)。

国际机构认为,世界上的许多大城市人口过剩,这些经济体的发展需要耗费相当大的成本。联合国(1993)“特大城市相关的负面因素”怎样恶劣，需要“对自身利益的控制并鼓励发展替代中心。”同样有报告发出警告”不平衡城市层次结构”以及大城市的犯罪,拥挤和社会不平等。《世界发展报告》(2000)的第七章描述了发展中国家大城市人们生活的残酷景象。《经济学家》在一个特殊的调查中提出了直接的问题(1995年7月29日):大城市的美好真的大于其阴暗面吗? 城市集中对生产发展的影响

在这一部分,我们主要探讨城市集中对生产率增长的影响。从城市集中程度或其主导地位入手进行调查,才能获得收益率的关键结果。这种调查方法同样能应用于考察城市化对经济增长的影响。第一个问题是如何衡量城市集中。人们通常采用三种方法来衡量。首先,威顿、宍戸 (1981)与亨德森 (1988)使用赫希曼-赫芬达尔集中度指数，计算一个城市的面积与城市在这个国家的人口总和。第二,罗森和雷斯尼克(1981)使用帕累托参数确定一个国家城市大小的分布,并衡量从上到下大小规模下降的速度,或大小分布的总体差异程度。在这些文献中,这两项指标的构建仅仅只用了一年的时间,却不可以在时间跨度较大的国家集团中应用,1960-1995。一个关键的问题是为什么城市集中

会影响生产率增长。静态城市模型中的损失主要是由过度或不足造成的，国内生产总值的损失则是资源配置不当的后果，过度主导表现为,城市发展都集中在一个或两个首要城市,这些城市都受到疲惫的规模经济,过度拥挤,人均基础设施成本过多的限制,而小城市有着尚未开发的规模经济,却常常缺乏资本投资(如，托利et al .,1979; 藤田昌久,

1989;亨德森和贝克尔,2000;奥和亨德森,2002)。布莱克和亨德森(1999)在卢卡斯(1988)

的基础上,建立了一个内生增长模型,城市规模与本地信息溢出效应的程度呈正相关,并交互式的影响当地的知识积累,促进生产率增长。

然而,城市规模过大，在生产活动中便要吸引大量资源进行投资和创新,并试图在一个拥挤的环境中维持生活质量。丹顿与普加(2001)对城市相关文献作出了贡献。在他们的研究中，首要城市的发展能获得适当的产品设计。相对规模较小的首要城市由于城市环境的限制,最终会限制国家生产力的发展。规模相对较大的首要城市，人们则要投入过多的时间来乘公车上下班与其他“浪费”活动。原则上，人们能够适应丹顿和普加经济增长背景下的动态模型,在这里，由于缺乏资源从事开发活动，滞后的城市集中会导致较低的知识积累，这同样会抑制生产力的发展。进而连续影响非线性的发展方式。但在这种情况下,考虑到威廉姆森(1965)的假说,我们希望城市集中的程度取决于一个国家的发展水平,代表国家稀缺的知识积累和经济基础设施。所有这些表述都迫切需要一个全面的增长模式来分析其特定的因素,但这已经超出了本文的研究范围。无论多么精确的模型,超越国家数据

的经验研究归根到底是要明确这样一个问题——这会在多大程度上主导生产率的增长?

基本的占主导地位的结果

为了获得这一研究结果,需要从一个主导变量着手。原始数据没有为我们提高太多有效信息。d ln（Y=N）与主导变量以及dd ln （Y=N）

与主导变量之间都为负相关的关系。需要我们弄清楚的问题是，控制和非线性规范对主导变量的影响。

基本的计量经济学结果在表2中得到了体现,第(1)-(4)列展现给我们的是一个二次型主导变量,它与人均产出相互作用，允许主导变量随人均产出而变化。在分析这些结果之前,我注意到，一个简单的线性主导术语有一个负系数。其次,在表2中的第(5)和(6)列,展示了一个简单的二次方程式，为了证明这一论点，我们引出了一个最佳的主导变量。在第(5)和(6) 列和文中其他所有报告结果中, 普通最小二乘法（OLS）和主导工具变量结果是不同的。OLS倾向将主导价值加于少曲率值f()函数之上。在第（5）列中，OLS之下的主导价值是0.20，在辅助变量估计(GMM)之下，其主导价值为0.46，并有着稳固与重要的系数。从最佳水平的主导地位来讲,一个标准偏差(0.13)的增加会带来生产力的增长, 尽管对于主导地位的变化会产生巨大影响。然而,

根据威廉姆森的假设，主导地位的最佳程度会随发展水平而有所改变。城市化对经济增长的影响

研究城市化对生产力的影响是困难的,特别是分离出有意义的结

果。首先分析这个困难的三个原因。首先, 在过去30年里，快速城市化特别是在非洲国家已经引发了低收入与经济负增长问题。这本身说明，城市化是多种因素的结果，它与国民产出构成和社会条件的变化相关,并不能促进本身经济的增长。第二,城市化是一个短暂的过程,随着经济发展，所有的国家最终都将“完全城市化”。在一些中等收入水平、城市化到一定程度已终止的国家，其城市化类别中的65 –85%;或者几乎50%的国家已经落入完全城市化这一类别。最后,各国对于城市化的定义千差万别,这就使得我们很难量化城市化的最优程度,因为这取决于每个国家如何统计其城市的数量。关注城市化的定义,完全城市化的城市如瑞士、奥地利和芬兰的城市化率已达到60 –

65%;对美国来说需城市化率达70%以上;而对阿根廷、智利和巴西这样的国家来讲,完全城市化率为80 – 85%。这些差异取决于这个城市低密度非农业人口的数量,特别是在大城市边缘地区的城市。例如,尽管中国官方表明城市化率为30%, 还有大约70%的人口生活在“市政”边界（市管辖）。

在考虑这些问题的基础上,探讨经济增长和城市化之间的关系。

首先,我们假设，对于任何收益水平的国家都有一个最优程度城市化。

即使“城市化促进经济增长”,大概没有人会认为，拥有半自给农业和高文盲率的低收入国家能很快实现城市化。为了研究城市化的发展情况，我使用了一个函数公式，它与方程式（4）和图表3中的

函数保持一致，尽管说国家规模这一变量是不重要的。在这个公式中，对整个样本来说，f(·)的值都没有显著意义的结果。为了得到城市化发展程度的一个最优结果，有必要对潜在的城市化国家的样本选择进行一定的限制。在这里，我选择了一组样本，这些国家的城镇化程度都低于70%；另一个限制是去除那些高收入的国家，不然会影响到最后的计算结果。对于选择的这些样本，使用最小二乘法和工具变量(GMM)来计算，结果见表6中,第一列和第二列的值。GMM工具变量的计算结果显示，在决定城市化的最佳程度以及潜在的最优程度的城市化时，一个国家的大小并不是一个决定性因素。但除此之外，其他结果值则不是很恰当，在某种意义上说，人均生产值的增长的影响，就是会减少城市化的“最佳程度”。,随着人均生产值的上升，城市化程度却下降了，这真是一个完全令人难以置信的结果。

另外，当我把城市化放到方程式中计算后，结果令人堪忧。对于计算过的结果值，我把国家规模列为重要考虑参数，不过城市化的计算并不是表6中显示的结果值。当城市化的最优模式能够一直持续下去，会有达到一个最好的城市化程度，尽管要很多样本选择上的限制。

表6第三列中的这些工具变量的结果，事实上，城市化似乎对经济增长产生了负面影响。总之,这些结果表明，城市化本身, 至少根据对这些样本国家的研究，并不会直接影响到经济水平的增长。结论本文认为,城市化代表的是经济发展中产业结构的不断转变,这是一个发展过程，而不是一个自身的刺激发展现象。然而,城市化进程采用的形式,或城市集中的程度,会对生产率的提高产生非常强烈的影响。城市集

中程度受国家政策和机构的影响,并反映一个特定的城市(如国家资本,如曼谷或墨西哥城)被国家重视的程度。对于国家规模和发展水平来讲,最优程度的城市集中,能平衡收益，增强当地知识积累对抗损失的能力，如资源转移到拥挤的大城市以提高生活质量。因此，城市集中程度的转变与国家规模和发展水平相关。