Productivity Gains from Agglomeration
and Migration in the People’s Republic
of China between 2002 y 2013
Pierre-Philippe Combes, Sylvie Démurger, and Shi Li∗

We evaluate the evolution of productivity gains in cities in the People’s Republic
of China between 2002 y 2013. En 2002, rural migrants exerted a strong
positive externality on the earnings of urban residents, which were also higher
on average in cities with access to foreign markets through a seaport. En
2007 y 2013, city size (measured in terms of both employment density and
land area) was the crucial determinant of productivity. Market access, si
internal or external, played no direct role. Rural migrants still enhanced urban
residents’ earnings in 2007 y 2013, though the effect was less than half that
en 2002. Urban gains and their evolution over time are very similar on a total
and a per hour earnings basis. Finalmente, skilled workers and females experienced
slightly larger gains than unskilled workers and males.

Palabras clave: agglomeration economies, migration, cities, urban development,
wage disparities
JEL codes: J31, O18, O53, R12, R23

I. Introducción

Although the empirical literature on productivity gains associated with
larger cities has matured over the past decade (Combes and Gobillon 2015),
evidence about the characteristics and role of cities in developing economies is
still largely incomplete (Chauvin et al. 2017). This stands in sharp contrast with
urbanization rates that are much lower but increasing rapidly in many developing
economies compared with developed countries where urbanization is fairly stable.
Por lo tanto, quantifying gains and costs from such a rapidly evolving spatial

∗Pierre-Philippe Combes: CNRS Research Professor at Université de Lyon, CNRS, GATE L-SE UMR 5824;
Professor at Sciences Po; and Research Fellow at the CEPR. Correo electrónico: ppcombes@gmail.com; Sylvie Démurger
(Autor correspondiente): CNRS Research Professor at Université de Lyon, CNRS, GATE L-SE UMR 5824; y
Research Fellow at IZA. Correo electrónico: demurger@gate.cnrs.fr; Shi Li: Professor, Business School, Beijing Normal
University and Research Fellow at IZA. Correo electrónico: lishi@bnu.edu.cn. This paper was part of a research program on
Amenities and Agglomeration in China (NSFC Grant No. 71503023). We would like to thank the participants at the
Asian Development Review Conference on Urban and Regional Development in Asia held in Seoul in July 2016,
the managing editor, and an anonymous referee for helpful comments and suggestions. We are also grateful to Zhu
Mengbing for excellent research assistance. en este documento, the Asian Development Bank recognizes “China” as the
People’s Republic of China. The usual disclaimer applies.

Asian Development Review, volumen. 34, No. 2, páginas. 184–200

© 2017 Asian Development Bank
and Asian Development Bank Institute

yo

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

/

mi
d
tu
a
d
mi
v
/
a
r
t
i
C
mi
–
pag
d

yo

F
/

/

/

/

3
4
2
1
8
4
1
6
4
4
3
7
2
a
d
mi
v
_
a
_
0
0
0
9
9
pag
d

/

.

F

b
y
gramo
tu
mi
s
t

t

oh
norte
0
9
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

Productivity Gains from Agglomeration and Migration in the PRC 185

concentration of economic activity is a highly topical issue. If one expects the gains
to outweigh the costs, then urbanization should be promoted, por ejemplo, por
facilitating rural-to-urban migration. If costs dominate, appropriate policy must be
implemented in order to refrain urban expansion.

The People’s Republic of China (PRC) offers the best example of a large
increase in rural–urban migration, having undergone such a process that has fed
the development of cities for more than 15 años. In an early contribution, Au
and Henderson (2006) document the productivity gains that could emerge from
larger cities in the PRC. They conclude that, apart from the largest cities, grande
productivity gains might also emerge from a larger number of medium-sized cities.
Using microdata from the mid-2000s, Combes, Démurger, and Li (2015) estimate
the specific externality that migrants exert on native urban workers—beyond the
impact of standard agglomeration variables such as total employment density and
city land area—and find evidence of a large positive effect of the city’s migrant
share on urban residents’ wages. This paper complements the analysis over time by
evaluating the evolution of productivity gains in cities in the PRC between 2002 y
2013. Measuring possible changes has important policy implications because the
success of urban policies ultimately depends on the nature of the gains, which may
have changed over time. Another contribution of this paper consists of estimating
urban gains not only in terms of total earnings but also earnings per hour, cual
is a closer measure of labor productivity, as well as for both skilled and unskilled
workers and by gender. Al hacerlo, we combine nationally representative household
income surveys conducted in 2002, 2007, y 2013 with city-level data.
We find that city characteristics have a significant

impact on labor
productivity and that the characteristics that matter have changed over time. En
2002, the migrant externality was very strong. To a lesser extent, access to foreign
markets through proximity to a seaport also played a positive role. En 2007 y, a
a greater extent, en 2013, agglomeration variables that are standard in the literature
and mostly reflect city size (p.ej., employment density and land area) had become
crucial determinants of a city’s productivity. Por el contrario, market access, si
internal or external, played no significant direct role. As for rural migrants, ellos
still enhanced urban residents’ earnings, but the effect was less than half what it had
been in 2002. These findings show that the PRC moved from urban gains shaped
by specific features (p.ej., large disparities across cities in terms of access to the sea
and migrant shares) en 2002 to a situation in 2007 y 2013 where variables more
typically affecting urban workers’ earnings are playing a more influential role. Qué
matters for cities in the PRC now is their capacity to host migrants as well as their
overall size in terms of both density and land area. Urban gains and their evolution
over time are very similar on a total and per hour earnings basis. Asombrosamente, nosotros
find little difference in earnings gains between skilled and unskilled workers, yet the
city effects are slightly larger for skilled workers. This may be due to the fact that the
substitution and externality effects from migrants are difficult to disentangle from

yo

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

/

mi
d
tu
a
d
mi
v
/
a
r
t
i
C
mi
–
pag
d

yo

F
/

/

/

/

3
4
2
1
8
4
1
6
4
4
3
7
2
a
d
mi
v
_
a
_
0
0
0
9
9
pag
d

/

.

F

b
y
gramo
tu
mi
s
t

t

oh
norte
0
9
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

186 Asian Development Review

one another with the data sets we use. Last, female workers seem to gain slightly
more from cities than male workers do.

Section II briefly recalls the theoretical background and the associated
empirical strategy, and presents the data we use. Section III is devoted to measuring
the impact of city characteristics for all workers taken together on both a total and
per hour earnings basis. Section IV evaluates agglomeration and migration gains
separately for skilled and unskilled workers, and for male and female workers.
Section V concludes.

II. Mechanisms, Empirical Specification, and Data

A.

Background and Methodology

marshall (1890) suggested long ago that gains from cities arise from various
local external economies of scale. The underlying mechanisms, which are now well
comprendido, were classified by Duranton and Puga (2004) as sharing (specialized
inputs, diversity, or risk); matching (quantity and quality of matches in local labor
markets); Y aprendiendo (creating, diffusing, and accumulating knowledge). Estos
mechanisms either directly impact household utility or increase their productivity,
which in turn leads to higher earnings. The largest part of the empirical literature
augments standard wage equations with city variables to evaluate the determinants
of the nominal gains that result from larger cities. Similar conclusions are usually
reached when one works on total factor productivity measures. Comparing nominal
gains to the higher cost of living in cities, and therefore assessing real income gains
in cities, is the focus of another strand of the literature.1

We use a standard two-step procedure for the estimation of agglomeration
effects discussed in Combes and Gobillon (2015) and extended by Combes,
Démurger, and Li (2015) to properly account for the role of migrants. El procedimiento
consists in estimating the following specification:

log wit = Xitθt + Lcstγt + δct + εit

δct = Uctβt + αt + vct

(1)

(2)

The first-step estimation of equation (1) evaluates the impact on individual
i’s wage at date t, wit, of city-time fixed effects, δct for city c where worker
i is employed at date t and of city c’s characteristics for sector s where i is
empleado, Lcst, net of the role of individual characteristics Xit. We estimate equation
(1) for each year separately, which makes all estimated parameters year specific.
Controlling for individual characteristics is crucial to remove the bias arising from a

1Ver, por ejemplo, Moretti (2013) and Combes, Duranton, and Gobillon (2016).

yo

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

/

mi
d
tu
a
d
mi
v
/
a
r
t
i
C
mi
–
pag
d

yo

F
/

/

/

/

3
4
2
1
8
4
1
6
4
4
3
7
2
a
d
mi
v
_
a
_
0
0
0
9
9
pag
d

.

/

F

b
y
gramo
tu
mi
s
t

t

oh
norte
0
9
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

Productivity Gains from Agglomeration and Migration in the PRC 187

possible nonrandom sorting of individuals across cities depending on their abilities.
The sorting of individuals has been documented to be large in some economies,
although this does not seem to be the case in the PRC as shown by Combes,
Démurger, and Li (2015). Introducing city-sector variables, typically here only the
logarithm of the share of sector s in city c employment, allows identifying the
role of the characteristics of the sector within the city beyond the role of other
nonsector-specific city characteristics captured by δct.2

The second step disentangles the characteristics of the city (vector Uct) eso
impact city productivity δct. We pool the 3 years of data together and compare
estimates when some (or all) estimated parameters are year specific. The two main
variables of interest are employment per square kilometer (density) and the share
of internal migrants. The literature has shown the former to be the main driver of
urban productivity. The role of the latter has been highlighted in the case of the PRC
by Combes, Démurger, and Li (2015). These authors discuss the functional forms
to be used in order to properly interpret the estimated coefficients. The logarithm of
the employment density of urban residents identifies the standard impact of density
that can be compared to the one obtained in the literature for other economies.
Entonces, the impact of the logarithm of the inverse of 1 minus the share of migrants in
the city’s total employment corresponds to the sum of the impact of density—for
a given number of urban residents, migrant inflows increase density—and the
separate migrant effect net of such density gains. The latter effect results from
a complementarity (if positive) or substitution (if negative) effect between urban
residents and migrants in the production function and a possible further migrant
externality (either positive or negative). Desafortunadamente, these last two effects cannot
be separately identified without appealing to extended data and a more complex
acercarse, as surveyed by Lewis and Peri (2015).

We augment the second-step specification by a number of city characteristics
that can shape productivity in cities in the PRC. Two margins shape city size: density
and land area. Density corresponds to the role of the intensive margin: eso es, a
larger number of workers per square kilometer within the city boundary. Land area
captures the role of the extensive margin, the city being made larger by extending
its boundaries at a given density: eso es, under a simultaneous proportional increase
in total employment.

Since trade takes place between locations and workers are mobile, el
literature has also emphasized the role of the access to distant markets that can
generate imported external economies in cities. We use two variables to disentangle
the roles of internal and international markets. The first variable is a within-PRC

2Lcst is centered with respect to its city mean, Lc.t, so that it does not capture the role of the overall
distribution of city-sector shares in the city, a city characteristic we want to be captured by δct. En cambio, individual
characteristics are not centered with respect to their city mean because the possible correlation between city and
individual characteristics is interpreted as a pure composition effect, which does not arise from agglomeration
externalities and thus needs to be controlled for so that δct encompasses city externalities only.

yo

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

/

mi
d
tu
a
d
mi
v
/
a
r
t
i
C
mi
–
pag
d

yo

F
/

/

/

/

3
4
2
1
8
4
1
6
4
4
3
7
2
a
d
mi
v
_
a
_
0
0
0
9
9
pag
d

/

.

F

b
y
gramo
tu
mi
s
t

t

oh
norte
0
9
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

188 Asian Development Review

market potential based on Harris (1954), which is the sum of the employment
density in other cities in the PRC divided by the distance to the city being
consideró. It assesses how far the city is from other large cities. As a large share of
the PRC’s exports is coursed through coastal ports, the second variable corresponds
to the city’s distance to the closest seaport.3

Por último, and even if this variable has never proved to be very influential, nosotros
also consider the role of the city’s industrial diversity measured by the inverse
of a Herfindhal index computed on the shares of each industry in the city’s total
employment.

Although we pool together different years, the impact of local variables is
mostly identified over the cross section of cities because time variations are much
smaller than across cities and also because few cities are sampled for the 3 years that
we consider. The literature emphasizes two types of possible biases. The larger one
is due to the spatial distribution of workers, which could depend on unobserved
skills that largely affect their productivity, and does not necessarily result from
city externalities. This can make the estimated impact of local variables twice as
large as it would be for the same individual who would locate in different cities in
different periods. Since we do not have access to an individual panel where workers
are followed across years, we can only control for observable individual skills. Para
ejemplo, a large spatial sorting according to skills has been observed in France and
the United States. In the case of France, Combes, Duranton, and Gobillon (2008)
show that controlling for observable skills is not strong enough to remove the bias.
Por el contrario, no sorting according to unobserved heterogeneity seems to be present
in the United States according to Baum-Snow and Pavan (2013). As for the PRC,
Combes, Démurger, and Li (2015) show that no spatial sorting occurs according to
observable skills. Por lo tanto, only a large sorting of unobserved skills, which should
not be correlated to observed skills, could significantly bias the estimates. We see
this as unrealistic, at least until many years of intensive migration and possibly
unequal access to higher education have further shaped the spatial distribution of
individual skills.

The second possible bias arises from reverse causality. Local characteristics
could be driven by local productivity, and not the reverse, if the location choices of
firms and workers are endogenous. The local density variable is the most suspicious
one under this perspective. This is why we use the specification proposed by
Combes, Démurger, and Li (2015), which decomposes total employment density
into the impacts of urban employees and the share of migrants. The urban residents’
density, which is shown to have the same marginal impact on local productivity
as total density, should be much less prone to reverse causality as urban residents
settled in the city prior to the review period.

3El 12 major coastal seaports in the PRC in terms of the volume of freight handled are, in decreasing order,
Shanghai, Ningbo, Guangzhou, Tianjin, Qingdao, Qinhuangdao, Dalian, Rizhao, Yingkou, Yantai, Lianyungang, y
Zhanjiang.

yo

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

/

mi
d
tu
a
d
mi
v
/
a
r
t
i
C
mi
–
pag
d

yo

F
/

/

/

/

3
4
2
1
8
4
1
6
4
4
3
7
2
a
d
mi
v
_
a
_
0
0
0
9
9
pag
d

/

.

F

b
y
gramo
tu
mi
s
t

t

oh
norte
0
9
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

Productivity Gains from Agglomeration and Migration in the PRC 189

Por el contrario, the migration variable is likely to be affected by reverse
causality. Migrants choose locations where employment conditions and earnings,
En particular, are favorable. By appealing to specifications considering different
sets of control variables and by using various instrumental variable strategies,
Combes, Démurger, and Li (2015) find evidence of biases in the PRC very similar
in magnitude to those usually documented for other economies. At around 20% en
mayoría, this is not very large. Addressing reverse causality is beyond the scope of
the present paper, though it deserves further investigation. As emphasized by Lewis
and Peri (2015), this would require more sophisticated strategies, which might be
difficult to implement with the data available.

Other local variables could be endogenous too, which is why we present
estimates with and without controlling for them. De nuevo, the literature that has
made some attempts to instrument these variables never reported any large bias.
Todavía, the problem could be more severe here for the land area variable because city
boundaries are regularly reshaped by the authorities to match local demographic
evolution. De nuevo, the causality would go the other way round. We have checked that
similar magnitudes were obtained when we use a land area definition that is constant
con el tiempo (using the 2007 valor) but this again deserves further investigation
provided that more comprehensive data are available.

B.

Datos

The data used in this paper come from various sources. Ecuación (1) es
estimated using three different nationally representative household income surveys
for the years 2002, 2007, y 2013. Ecuación (2) requires city-level data compiled
from the China City Statistical Yearbook edited by the National Bureau of Statistics
(NBS). For the city-level share of migrants, we use the 2000 population census, el
1% 2005 population survey, y el 2010 population census for the years 2002,
2007, y 2013, respectively.4

Individual databases for 2002 y 2013 are part of the China Household
Income Project (CHIP), mientras que la 2007 data set is a subsample of the NBS Urban
Household Survey.5 The CHIP survey is closely related to the NBS household
survey, especially on the income section, which makes all 3 years fairly comparable.
Todavía, because working time is documented only in the CHIP 2002 and CHIP 2013
datos, we compute hourly wages for these 2 years only. Además, the surveyed
prefecture-level cities vary substantially across all 3 años. El 2002 survey covers
53 cities in 12 provinces, el 2007 survey covers 87 cities in 16 provinces, y el
2013 survey covers 99 cities in 14 provinces. There are only 20 cities common to all

4See Combes, Démurger, and Li (2015) for a discussion of city-level data sources.
5Para 2002, a detailed description of the survey can be found in Li et al. (2008). The NBS 2007 data set is

described in Combes, Démurger, and Li (2015).

yo

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

/

mi
d
tu
a
d
mi
v
/
a
r
t
i
C
mi
–
pag
d

yo

F
/

/

/

/

3
4
2
1
8
4
1
6
4
4
3
7
2
a
d
mi
v
_
a
_
0
0
0
9
9
pag
d

/

.

F

b
y
gramo
tu
mi
s
t

t

oh
norte
0
9
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

190 Asian Development Review

three surveys, while there are 38 cities common to the CHIP 2002 and CHIP 2013
surveys.

A major feature of the three individual databases is that they cover registered
urban residents only (the urban hukou holders).6 This means that an important
segment of the urban labor market, comprising migrants not officially registered
in cities, is excluded. Representing about 17% of the PRC’s 1.3 billion people,
migrants constitute an important social grouping, yet they are clearly marginalized
(Cai, Parque, and Zhao, 2008; Démurger et al. 2009). The exclusion of migrants
whose hukou is not in the same location as their place of employment has an
important implication for our estimates: our data set refers to urban residents
(natives) only and one cannot infer that earnings determinants across the cities
under review would equally apply to migrant workers. Sin embargo, we provide
an estimate of the impact that migrants exert on local urban workers simultaneously
with agglomeration effects.

All explanatory variables are defined consistently across all 3 years for both
estimation steps. The only exception is for enterprise ownership in equation (1),
which is more detailed in 2002 y 2013 in comparison with 2007.

III. Agglomeration Gains in 2002, 2007, y 2013

A.

Total Earnings

Mesa 1 displays the results obtained for all workers’ earnings for the
estimation of equation (2). Rows with the variable’s name without vintage (p.ej.,
“Density”) correspond with the effect common to all years. This is also the total
effect for year 2013. The effects for 2007 o 2002 are obtained by adding to this
number the coefficient reported in the corresponding line (p.ej., “Density 2007”
or “Density 2002”). The significance test denotes whether the effect for 2007 (o
2002) is significantly different from the 2013 efecto. The first five columns provide
estimates based on total earnings and the last two columns give estimates based on
earnings per hour. As indicated above, the number of hours worked is not available
in the microdata for the year 2007, which implies that these per hour earnings
estimates are based on 2002 y 2013 solo.

En columna (1), density only is introduced in the specification. Obtenemos
an elasticity of total earnings with respect to density equal to 0.097 para 2013, como
reported in the first line of the table. The value estimated for 2007 is not significantly
different from the one for 2013. The elasticity for these 2 years is close to the values
reported by Combes, Démurger, and Li (2015) and is almost three times larger

6The hukou (household registration) system is a distinctive institutional feature of the PRC that divides the
population based on occupation (agricultural or nonagricultural) and place of residence. By entitling access to social
benefits to local hukou holders only, the system has limited population mobility for decades. Mayoría (rural) migrants
still hold their hometown hukou, which prevents them from permanently settling in cities.

yo

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

/

mi
d
tu
a
d
mi
v
/
a
r
t
i
C
mi
–
pag
d

yo

F
/

/

/

/

3
4
2
1
8
4
1
6
4
4
3
7
2
a
d
mi
v
_
a
_
0
0
0
9
9
pag
d

.

/

F

b
y
gramo
tu
mi
s
t

t

oh
norte
0
9
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

Productivity Gains from Agglomeration and Migration in the PRC 191

Mesa 1. City Determinants of Total and Per Hour Earnings

Total Earnings

Hourly Earnings

(1)

(2)

(3)

(4)

(5)

(6)

(7)

0.097***
(0.025)
−0.005
(0.033)
−0,069
(0.043)

Density

Density 2007

Density 2002

Migrants

Migrants 2007

Migrants 2002

Land area

Land area 2007

Land area 2002

Market potential

Market potential

2007

Market potential

2002

Distance to seaport

Distance to seaport

2007

Distance to seaport

2002
Diversity

Diversity 2007

Diversity 2002

0.088***
0.059***
(0.022)
(0.022)
−0.030
−0,017
(0.028)
(0.026)
−0.067* −0.053
(0.033)
(0.035)
0.210***
0.267***
(0.060)
(0.061)
0.096
0.106
(0.071)
(0.075)
0.390***
0.491***
(0.114)
(0.120)
0.080***
(0.017)

0.086***
(0.027)

0.412***
(0.145)
0.097***
(0.030)

−0.085*
(0.047)
0.185***
(0.067)

0.070***
(0.025)

0.393***
(0.127)
0.072***
(0.023)

0.096***
0.092***
(0.025)
(0.024)
−0.030
−0,028
(0.030)
(0.034)
−0.082* −0.091** −0.044
(0.036)
(0.043)
(0.041)
0.194***
0.207***
0.205***
(0.066)
(0.062)
(0.062)
0.086
0.092
(0.077)
(0.076)
0.412***
0.466***
(0.133)
(0.131)
0.092***
0.089***
(0.028)
(0.027)
0.011
0.008
(0.039)
(0.041)
−0.059
−0.046
(0.044)
(0.043)
0.099
0.088
(0.084)
(0.055)
0.048
(0.125)
−0.137
(0.148)
−0.020** −0.019** −0.009
(0.014)
(0.009)
(0.009)
0.002
(0.019)
−0.052**
−0.056**
(0.026)
(0.023)
−0.109** −0.115** −0.111** −0.097* −0.095
(0.056)
(0.061)
(0.048)
0.006
(0.145)
−0.104
(0.146)

−0.070
(0.162)
−0.029** −0.010
(0.015)
(0.012)

−0,058
(0.048)
0.070
(0.091)

−0.112
(0.160)

0.093*
(0.055)

0.057
(0.075)

(0.056)

(0.048)

Observaciones
R2

239
0.67

239
0.80

239
0.82

239
0.83

239
0.83

152
0.84

152
0.85

Notas: The panel of cities is not balanced (99 cities for 2013, 87 cities for 2007, y 53 cities for 2002.) Time
dummies are included in all specifications. Standard errors in brackets. *** = p < 0.01, ** = p < 0.05, * = p < 0.10. Source: Authors’ calculations. than usual estimates for developed economies. Large density economies seem to prevail in the PRC, at least from the mid-2000s onward. By contrast, the effect appears much lower in 2002, even though the estimated year-specific effect is not significantly different from zero. We will return to this issue once other variables have been introduced. l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . / e d u a d e v / a r t i c e - p d l f / / / / 3 4 2 1 8 4 1 6 4 4 3 7 2 a d e v _ a _ 0 0 0 9 9 p d / . f b y g u e s t t o n 0 9 S e p e m b e r 2 0 2 3 192 Asian Development Review Column (2) adds the role of migrants, which is found to be largely positive albeit to a decreasing degree between 2002 and 2013. In this augmented specification, the effect of density is lower. This can be due to the fact that if denser cities host more migrants, then estimates in column (1) capture both effects. The correlation of density with other city characteristics is also high; in particular, it is positive with market access and negative with distance to seaports and land area. Meaningful estimates are therefore obtained in column (3), where all variables are introduced together, and even more in columns (4) and (5), our preferred specification, which progressively considers more year-specific effects. Most control effects compensate each other with regard to density, and its impact, at 0.096 in column (5), is very close to that of column (1). While the density impact is 33% lower in 2007, though not significantly because of the imprecision of the estimation, it is significantly lower in 2002 to the point that it is fully canceled out. Hence, while there were no density gains present in the PRC in 2002, by 2013 these gains were three times larger than in developed economies. Moreover, while the extensive margin of city size is mostly found to have no impact in developed economies, it generates further gains in the PRC from 2007 onward. As with density, the impact of land area is significantly different from zero in 2007, while it is not significant though slightly positive in 2002.7 By contrast, the impact of migrants was three times larger in 2002 than what it was in 2013, with 2007 presenting an intermediate value that is somewhat closer to the 2013 estimate. The 2002 impact is very large. Typically, moving from the first to the last quartile of the migrant share variable increases natives’ earnings by 17.5% in 2002, but by only 8.5% in 2013.8 Distance to a seaport also presents a significant negative effect in 2002, which vanishes in the latter years. This finding suggests that proximity to the sea (and thus access to international markets) was a factor of productivity gains in 2002 that did not seem to hold in subsequent years. This could be related to the changing role of international trade in the PRC’s economic growth over the past decade. Finally, a city’s industrial diversity was found to be only marginally significant in all 3 years under review, as usually found in the literature. Overall, we find that the two standard city-size variables, density and land area, as well as the presence of migrants, generate higher earnings in cities in 2007 and 2013, with other characteristics playing no significant direct role.9 The impact of migrants in 2007 and 2013 is, however, less than half of what it was in 2002. At that time, density and land area had no effect, access to the sea being the only other city characteristic significantly impacting earnings. These findings highlight 7These comments are based on additional regressions where a dummy normalization is done so that significance tests for the total effects for 2002 and 2007 are obtained. They are available upon request. 8Note that the effect is divided by two despite the fact that the interquartile gap in the migrant share variable increased from 2 to 2.6 between 2002 and 2013. 9Some variables (market potential, for instance) could drive density or the presence of migrants and thus have an indirect effect. Yet, the strategy developed here does not allow this to be assessed. l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . / e d u a d e v / a r t i c e - p d l f / / / / 3 4 2 1 8 4 1 6 4 4 3 7 2 a d e v _ a _ 0 0 0 9 9 p d / . f b y g u e s t t o n 0 9 S e p e m b e r 2 0 2 3 Productivity Gains from Agglomeration and Migration in the PRC 193 the PRC’s move from urban gains shaped by some of its specific features in the early 2000s—large disparities across cities in terms of access to the sea and the presence of significant numbers of rural migrants—to a situation where variables that typically affect urban earnings are now playing an influential role. What matters for cities in the PRC now is their capacity to host migrants as well as their overall size, in terms of both density and land area. First-step estimates for equation (1) from which the city fixed effects used in the second step are obtained are provided in Appendix Table A.1. We do not further comment on the impact of individual characteristics, which are consistent with usual findings for the PRC (Démurger, Li, and Yang 2012). The first-step specification also includes the role of a city–industry-specific variable, specialization, which is measured by the logarithm of the share in the local economy of employment in the firm’s industry. Being specialized in an industry increases earnings in a city, which is an additional agglomeration effect typically found in the literature. The effect seems to have reinforced over time: it is significant on earnings per hour only in 2002, while the magnitude of the impact in 2013 is larger than that usually obtained. As mentioned above, the cities surveyed for each wave differ across time. The regressions provided in Table 1 are performed on the resulting nonbalanced panel. We checked that the distribution of city characteristics does not differ too much between waves. As a further robustness check, we run the same regressions on the subsample of cities that are common in all three waves of the survey. Results are presented in Table A.2 in the Appendix. Our main conclusions remain unchanged, even though some estimates are less precise because of the much smaller sample size. B. Earnings per Hour The agglomeration literature mostly focuses on productivity externalities, which are better tested empirically by earnings per hour than by total earnings. The 2002 and 2013 surveys allow us to compute for each worker a measure of earnings per hour, which should thus be closer to labor productivity. If the number of hours worked varies significantly across cities, as documented by Rosenthal and Strange (2008) for the United States, the city determinants of total and per hour earnings can differ. As shown by the last two columns in Table 1, this appears not to be the case in the PRC, whether only density and migrants are interacted with year (column [6]) or all city variables (column [7]).10 The estimated impact of density is only very slightly lower on earnings per hour than it is on total earnings. Importantly, both 10This is supported by the fact that the impact of city characteristics on hours worked is not significant. Indeed, when using the same specification for hours as for earnings, market potential is the only local characteristic to have a significant (negative) impact (in 2002 only). Results are available upon request. l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . / e d u a d e v / a r t i c e - p d l f / / / / 3 4 2 1 8 4 1 6 4 4 3 7 2 a d e v _ a _ 0 0 0 9 9 p d / . f b y g u e s t t o n 0 9 S e p e m b e r 2 0 2 3 194 Asian Development Review density and land area effects still totally offset in 2002. Neither the intensive nor the extensive margins of city size increase labor productivity at that date. By contrast, the larger presence of migrants does significantly increase earnings per hour, with an impact equal to the one estimated on total earnings. Again, the effect is less than half in 2013, but is still significant. Overall, both total and per hour earnings are influenced by the same city characteristics and to the same extent. IV. Gender and Skills Heterogeneity The literature shows that both the returns to individual characteristics and the returns to city characteristics, typically gains from larger city size, should and do differ across skills. High-skilled workers tend to benefit more from agglomeration, in particular when local externalities result from technological spillovers (Bacolod, Blum, and Strange 2009). The comparison of agglomeration effects between genders is less frequent; Phimister (2005) is one of the rare exceptions. The data we use allow us to test whether city characteristics have different impacts on skilled and unskilled workers, and on male and female workers. The results are reported in Table 2. Consistent with what is often observed for developed economies, skilled workers gain slightly more from larger city size (both higher density and larger land area) than unskilled workers. This finding holds for both total and per hour earnings, suggesting that at least part of city-size gains arise from technological spillovers rather than from market effects. Yet, the skills gap is not large since point estimates for high-skilled workers are only around 10% higher than for their unskilled counterparts. Because of the lack of the estimates’ precision, the gap is not significantly different from zero most of the time. A similar conclusion is reached for the impact of the city’s migrant share for the year 2013. The estimated effects are very similar for skilled and unskilled workers, and even slightly larger for unskilled workers on total earnings. When the migrant effect is twice as large in 2002, the gap in favor of skilled workers is also larger, with the marginal impact being around 25% larger, though it is not significant. A larger positive impact of migrants on skilled urban residents was expected as migrants are more likely to substitute for unskilled workers in production functions. To the best of our knowledge, a gender gap for urban returns has never been documented in the PRC. We do find small differences in favor of females, even though the size of the sample makes it difficult to obtain significant gaps. Gaps are also slightly larger for earnings per hour than for total earnings. Female workers are found to gain slightly more from city size (both higher density and larger land area). They also gain slightly more from the presence of migrants. A possible explanation for the larger benefits for females in cities may relate to household-level choices of which neighborhood to live in. If the location decision is made by men, l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . / e d u a d e v / a r t i c e - p d l f / / / / 3 4 2 1 8 4 1 6 4 4 3 7 2 a d e v _ a _ 0 0 0 9 9 p d . / f b y g u e s t t o n 0 9 S e p e m b e r 2 0 2 3 Productivity Gains from Agglomeration and Migration in the PRC 195 y l r u o H l a t o T y l r u o H l a t o T s g n i n r a E r u o H r e P d n a l a t o T f o s t n a n i m r e t e D y t i C . 2 e l b a T e l a m e F ) 8 ( * * * 8 9 0 0 . ) 3 3 0 0 ( . e l a M ) 7 ( * * 4 7 0 . 0 ) 9 2 0 . 0 ( * * 5 1 1 0 − . * * 8 8 1 0 . ) 7 5 0 0 ( . ) 1 8 0 0 ( . * * 0 8 1 . 0 ) 9 4 0 . 0 ( ) 1 7 0 . 0 ( 0 6 0 . 0 − * * 5 4 4 0 . ) 6 7 1 0 ( . * * * 5 1 1 0 . ) 6 3 0 0 ( . * * 9 9 3 . 0 * * 0 8 0 . 0 ) 3 5 1 . 0 ( ) 2 3 0 . 0 ( ) 8 5 0 0 ( . 7 7 0 0 . ) 1 1 1 0 ( . 7 7 0 0 − . ) 0 5 0 . 0 ( 8 5 0 . 0 ) 6 9 0 . 0 ( 6 3 0 . 0 − . d e u n i t n o C ) 6 9 1 0 ( . 1 9 0 0 − . ) 0 7 1 . 0 ( 5 3 0 . 0 − e l a m e F ) 6 ( e l a M ) 5 ( * * * 4 0 1 . 0 * * * 8 8 0 . 0 ) 9 2 0 . 0 ( 4 4 0 . 0 − ) 1 4 0 . 0 ( * * 1 2 1 . 0 − * * * 0 4 2 . 0 ) 1 5 0 . 0 ( ) 3 7 0 . 0 ( 4 8 0 . 0 ) 1 9 0 . 0 ( * * * 3 1 4 . 0 * * * 1 0 1 . 0 ) 7 5 1 . 0 ( ) 3 3 0 . 0 ( 6 0 0 . 0 ) 8 4 0 . 0 ( 7 6 0 . 0 − ) 2 5 0 . 0 ( 5 6 0 . 0 ) 9 9 0 . 0 ( 6 5 0 . 0 ) 8 4 1 . 0 ( 1 1 1 . 0 − ) 5 7 1 . 0 ( ) 6 2 0 . 0 ( 3 1 0 . 0 − ) 6 3 0 . 0 ( 6 6 0 . 0 − ) 5 4 0 . 0 ( * * * 7 8 1 . 0 ) 5 6 0 . 0 ( 2 7 0 . 0 ) 1 8 0 . 0 ( * * * 2 1 4 . 0 * * * 1 8 0 . 0 ) 1 4 1 . 0 ( ) 9 2 0 . 0 ( 6 2 0 . 0 ) 3 4 0 . 0 ( 5 4 0 . 0 − ) 6 4 0 . 0 ( 6 1 1 . 0 ) 8 8 0 . 0 ( 8 4 0 . 0 ) 2 3 1 . 0 ( 3 3 1 . 0 − ) 7 5 1 . 0 ( ) 4 ( * * 8 7 0 . 0 ) 3 3 0 . 0 ( * * * 8 8 0 . 0 ) 3 3 0 . 0 ( d e l l i k s n U d e l l i k S d e l l i k s n U d e l l i k S ) 3 ( ) 2 ( ) 1 ( * * 6 9 1 . 0 ) 7 5 0 . 0 ( ) 3 8 0 . 0 ( 9 7 0 . 0 − * * 2 0 2 . 0 ) 7 5 0 . 0 ( ) 2 8 0 . 0 ( 2 9 0 . 0 − * * * 4 2 3 . 0 * * 6 8 0 . 0 ) 8 7 1 . 0 ( ) 7 3 0 . 0 ( * * 5 1 4 . 0 ) 8 7 1 . 0 ( * * * 9 2 1 . 0 ) 7 3 0 . 0 ( ) 9 5 0 . 0 ( 1 6 1 . 0 ) 2 1 1 . 0 ( 1 5 0 . 0 − ) 9 5 0 . 0 ( 2 0 0 . 0 ) 2 1 1 . 0 ( 2 9 0 . 0 − ) 9 9 1 . 0 ( 3 4 0 . 0 − ) 8 9 1 . 0 ( 8 1 0 . 0 − * * * 9 8 0 . 0 * * * 8 9 0 . 0 y t i s n e D ) 1 3 0 . 0 ( 4 2 0 . 0 − ) 3 4 0 . 0 ( * 6 9 0 . 0 − ) 3 5 0 . 0 ( ) 9 2 0 . 0 ( 3 3 0 . 0 − ) 0 4 0 . 0 ( * 5 9 0 . 0 − ) 0 5 0 . 0 ( 7 0 0 2 y t i s n e D 2 0 0 2 y t i s n e D * * * 2 1 2 . 0 * * * 3 9 1 . 0 s t n a r g i M ) 7 7 0 . 0 ( 2 7 0 . 0 * * 5 6 3 . 0 ) 6 9 0 . 0 ( ) 6 6 1 . 0 ( * * 2 7 0 . 0 ) 4 3 0 . 0 ( 7 2 0 . 0 ) 1 5 0 . 0 ( 4 5 0 . 0 − ) 5 5 0 . 0 ( 6 5 1 . 0 ) 4 0 1 . 0 ( 8 6 0 . 0 ) 6 5 1 . 0 ( 4 2 1 . 0 − ) 4 8 1 . 0 ( ) 2 7 0 . 0 ( 1 1 1 . 0 ) 0 9 0 . 0 ( * * * 3 4 4 . 0 * * * 9 2 1 . 0 ) 6 5 1 . 0 ( ) 2 3 0 . 0 ( 4 1 0 . 0 − ) 8 4 0 . 0 ( * 5 9 0 . 0 − ) 2 5 0 . 0 ( 3 4 0 . 0 ) 8 9 0 . 0 ( 2 8 0 . 0 ) 7 4 1 . 0 ( 0 8 0 . 0 − ) 4 7 1 . 0 ( 7 0 0 2 s t n a r g i M 2 0 0 2 s t n a r g i M a e r a d n a L 7 0 0 2 a e r a d n a L 2 0 0 2 a e r a d n a L l a i t n e t o p t e k r a M 7 0 0 2 l a i t n e t o p t e k r a M 2 0 0 2 l a i t n e t o p t e k r a M l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . / e d u a d e v / a r t i c e - p d l f / / / / 3 4 2 1 8 4 1 6 4 4 3 7 2 a d e v _ a _ 0 0 0 9 9 p d . / f b y g u e s t t o n 0 9 S e p e m b e r 2 0 2 3 196 Asian Development Review y l r u o H l a t o T y l r u o H l a t o T . d e u n i t n o C . 2 e l b a T e l a m e F ) 8 ( ) 8 1 0 0 ( . 1 2 0 0 − . e l a M ) 7 ( ) 6 1 0 . 0 ( 2 0 0 . 0 − 0 5 0 0 − . * * 9 5 0 . 0 − ) 1 3 0 0 ( . 8 9 0 0 − . ) 4 7 0 0 ( . ) 7 2 0 . 0 ( 2 8 0 . 0 − ) 4 6 0 . 0 ( ) 3 9 1 0 ( . 5 1 2 0 − . ) 8 6 1 . 0 ( 5 4 0 . 0 − 2 5 1 2 8 0 . 2 5 1 4 8 . 0 e l a m e F ) 6 ( 7 1 0 . 0 − ) 6 1 0 . 0 ( 2 2 0 . 0 ) 3 2 0 . 0 ( * 8 4 0 . 0 − ) 8 2 0 . 0 ( * 0 1 1 . 0 − ) 6 6 0 . 0 ( 3 3 0 . 0 − ) 0 7 1 . 0 ( 2 1 2 . 0 − ) 3 7 1 . 0 ( 9 3 2 0 8 . 0 e l a M ) 5 ( 4 0 0 . 0 − ) 5 1 0 . 0 ( 5 1 0 . 0 − ) 0 2 0 . 0 ( * * 5 5 0 . 0 − ) 5 2 0 . 0 ( * 9 9 0 . 0 − ) 9 5 0 . 0 ( 9 2 0 . 0 ) 3 5 1 . 0 ( 0 3 0 . 0 − ) 5 5 1 . 0 ( 9 3 2 3 8 . 0 d e l l i k s n U d e l l i k S d e l l i k s n U d e l l i k S ) 4 ( 1 0 0 . 0 ) 9 1 0 . 0 ( ) 3 ( ) 8 1 0 . 0 ( 5 2 0 . 0 − ) 1 3 0 . 0 ( * * 5 5 1 . 0 − ) 5 7 0 . 0 ( 9 4 0 . 0 − ) 1 3 0 . 0 ( * * 1 7 1 . 0 − ) 5 7 0 . 0 ( 7 4 0 . 0 − ) 6 9 1 . 0 ( 9 5 1 . 0 − ) 6 9 1 . 0 ( 0 1 0 . 0 − ) 2 ( 1 0 0 . 0 − ) 7 1 0 . 0 ( 5 0 0 . 0 ) 4 2 0 . 0 ( * 9 4 0 . 0 − ) 9 2 0 . 0 ( * * * 6 9 1 . 0 − ) 0 7 0 . 0 ( 7 2 0 . 0 ) 0 8 1 . 0 ( 3 8 0 . 0 − ) 2 8 1 . 0 ( 2 5 1 6 8 . 0 2 5 1 8 6 . 0 9 3 2 3 8 . 0 ) 1 ( ) 6 1 0 . 0 ( 8 0 0 . 0 2 2 0 . 0 − ) 3 2 0 . 0 ( 3 4 0 . 0 − ) 7 2 0 . 0 ( * * 3 3 1 . 0 − ) 6 6 0 . 0 ( 4 0 0 . 0 − ) 0 7 1 . 0 ( 4 6 0 . 0 − ) 2 7 1 . 0 ( 9 3 2 3 7 . 0 7 0 0 2 t r o p a e s o t e c n a t s i D 2 0 0 2 t r o p a e s o t e c n a t s i D t r o p a e s o t e c n a t s i D 7 0 0 2 y t i s r e v i D 2 0 0 2 y t i s r e v i D y t i s r e v i D s n o i t a v r e s b O 2 R l l a n i d e d u l c n i e r a s e i m m u d e m T i . ) 2 0 0 2 r o f s e i t i c 3 5 d n a , 7 0 0 2 r o f s e i t i c 7 8 , 3 1 0 2 r o f s e i t i c 9 9 ( d e c n a l a b t o n s i s e i t i c f o l e n a p e h T : s e t o N . 0 1 . 0 < p = * , 5 0 . 0 < p = * * , 1 0 . 0 < p = * * * . s t e k c a r b n i s r o r r e d r a d n a t S . s n o i t a c fi i c e p s . s n o i t a l u c l a c ’ s r o h t u A : e c r u o S l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . / e d u a d e v / a r t i c e - p d l f / / / / 3 4 2 1 8 4 1 6 4 4 3 7 2 a d e v _ a _ 0 0 0 9 9 p d / . f b y g u e s t t o n 0 9 S e p e m b e r 2 0 2 3 Productivity Gains from Agglomeration and Migration in the PRC 197 either because of cultural and social norms or because male earnings are higher, females have to adapt to the nonfreely chosen location with respect to their own job opportunities. Better job matches are more easily found in dense and large areas, and externalities are more important there than in less dense and smaller locations where job opportunities are less abundant. As a result, the gender gap should be lower in cities benefiting from characteristics such as larger size and a larger pool of migrants. V. Conclusion Gains from urbanization are evident in the PRC, where they seem to be larger than in developed economies. The nature of these gains has evolved rapidly and the city characteristics that shaped such gains in 2002 are different from those in effect in 2013. The presence of rural migrants and access to a seaport were the largest productivity drivers in cities in the PRC in 2002. City size, which is measured by both the intensive margin (employment density) and the extensive margin (land area), became the most important determinant over the next decade even as the migrant externality remained strong. In 2013, workers in cities benefited from more migrants, increased density, and a larger land area. This contrasts with developed economies, where density has the largest effect (and with much lower elasticity). Internal market access, which is almost the only other important local characteristic in developed countries, does not seem to play a major role in the PRC. In this paper, we do not assess possible biases due to missing city variables or reverse causality. In earlier studies on developed economies, reverse causality was found to affect estimates by no more than 20%. Yet, this should be carefully evaluated in future studies on the PRC because high migration rates could make such biases larger. Furthermore, the land area of cities is also regularly updated by authorities in the PRC based on past development, which could affect the estimated impact. Similarly, we do not find evidence of any strong spatial sorting according to observed individual skills, but this issue could emerge progressively through the endogenous location choices of an increasingly heterogeneous population with regard to skills. Lastly, the migrant externality we document is largely a black box that needs to be better understood. The impact of migrants is a mix of externality and substitution effects. Identifying these effects separately is needed to design consistent local policies. Identifying these channels separately requires working on larger data sets because some differences (for instance, between skilled and unskilled workers) seem to emerge from our results. Investigating in more detail the role of workers’ allocation between industries and occupations, which sharply differs between local residents and migrants, is necessary too. Finally, as migrants themselves become a larger share of a city’s population, measuring the impact of city characteristics on their own earnings would help to complete the analysis of l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . / e d u a d e v / a r t i c e - p d l f / / / / 3 4 2 1 8 4 1 6 4 4 3 7 2 a d e v _ a _ 0 0 0 9 9 p d / . f b y g u e s t t o n 0 9 S e p e m b e r 2 0 2 3 198 Asian Development Review urbanization gains in the PRC. Data limitations in the surveys used in this study did not allow us to investigate these additional issues. References Au, Chun-Chung, and J. Vernon Henderson. 2006. “Are Chinese Cities Too Small?” Review of Economic Studies 73 (3): 549–76. Bacolod, Marigee, Bernardo S. Blum, and William C. Strange. 2009. “Skills in the City.” Journal of Urban Economics 65 (2): 136–53. Baum-Snow, Nathaniel, and Ronni Pavan. 2013. “Inequality and City Size.” Review of Economics and Statistics 93 (5): 1535–48. Cai, Fang, Albert Park, and Yaohui Zhao. 2008. “The Chinese Labor Market in the Reform Era.” In China’s Great Economic Transformation, edited by Loren Brandt and Thomas Rawski, 167–214. Cambridge: Cambridge University Press. Chauvin, Juan Pablo, Edward Glaeser, Yueran Ma, and Kristina Tobio. 2017. “What is Different about Urbanization in Rich and Poor Countries? Cities in Brazil, China, India and the United States.” Journal of Urban Economics 98:17–49. Combes, Pierre-Philippe, Sylvie Démurger, and Shi Li. 2015. “Migration Externalities in China.” European Economic Review 76: 152–67. Combes, Pierre-Philippe, Gilles Duranton, and Laurent Gobillon. 2008. “Spatial Wage Disparities: Sorting Matters!” Journal of Urban Economics 63 (2): 723–42. _____. 2016. “The Costs of Agglomeration: Land and House Prices in French Cities.” Mimeograph. Combes, Pierre-Philippe, and Laurent Gobillon. 2015. “The Empirics of Agglomeration Economies.” In Handbook of Urban and Regional Economics, Volume 5A, edited by Gilles Duranton, J. Vernon Henderson, and Will Strange. Amsterdam: North-Holland. Démurger, Sylvie, Marc Gurgand, Shi Li, and Ximing Yue. 2009. “Migrants as Second-Class Workers in Urban China? A Decomposition Analysis.” Journal of Comparative Economics 37 (4): 610–28. Démurger, Sylvie, Shi Li, and Juan Yang. 2012. “Earnings Differentials between the Public and Private Sectors in China: Exploring Changes for Urban Local Residents in the 2000s.” China Economic Review 23 (1): 138–53. Duranton, Gilles, and Diego Puga. 2004. “Micro-Foundations of Urban Agglomeration Economies.” In Handbook of Regional and Urban Economics, Volume 4, edited by J. Vernon Henderson and Jacques-François Thisse, 2063–117. Amsterdam: North-Holland. Harris, Chauncy. 1954. “The Market as a Factor in the Localization of Industry in the United States.” Annals of the Association of American Geographers 44 (4): 315–48. Lewis, Ethan, and Giovanni Peri. 2015. “Immigration and the Economy of Cities and Regions.” In Handbook of Urban and Regional Economics, Volume 5A, edited by Gilles Duranton, J. Vernon Henderson, and Will Strange. Amsterdam: North-Holland. Li, Shi, Chuliang Luo, Zhong Wei, and Ximing Yue. 2008. “The 1995 and 2002 Household Surveys: Sampling Methods and Data Description.” In Inequality and Public Policy in China, edited by Bjorn Gustafsson, Shi Li, and Terry Sicular, 337–53. Cambridge: Cambridge University Press. Marshall, Alfred. 1890. Principles of Economics. London: Macmillan. Moretti, Enrico. 2013. “Real Wage Inequality.” American Economic Journal: Applied Economics 5 (1): 65–103. l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . / e d u a d e v / a r t i c e - p d l f / / / / 3 4 2 1 8 4 1 6 4 4 3 7 2 a d e v _ a _ 0 0 0 9 9 p d . / f b y g u e s t t o n 0 9 S e p e m b e r 2 0 2 3 Productivity Gains from Agglomeration and Migration in the PRC 199 Phimister, Euan. 2005. “Urban Effects on Participation and Wages: Are There Gender Differences?” Journal of Urban Economics 58 (3): 513–36. Rosenthal, Stuart S., and William C. Strange. 2008. “Agglomeration and Hours Worked.” Review of Economics and Statistics 90 (1): 105–18. Appendix Male Years of education Experience Experience2 Private enterprises Foreign enterprises Government administration Other ownership Specialization Occupation dummies Industry dummies City dummies Observations R2 Table A.1. Individual Wage Disparities— Ordinary Least Squares Estimates for the First Step 2013 Total (1) Hourly (2) 2007 Total (3) 2002 Total (4) Hourly (5) 0.236*** (0.015) 0.065*** (0.004) 0.044*** (0.002) 0.246*** 0.272*** (0.010) (0.015) 0.065*** 0.059*** (0.002) (0.003) 0.040*** 0.054*** (0.002) (0.002) −0.001*** −0.001*** −0.001*** −0.001*** −0.001*** (0.000) (0.000) 0.100*** (0.014) 0.052*** (0.003) 0.039*** (0.003) 0.138*** (0.013) 0.049*** (0.003) 0.043*** (0.003) (0.000) (0.000) (0.000) (0.037) (0.026) (0.021) (0.036) −0.190*** −0.207*** −0.218*** −0.136*** −0.203*** (0.023) (0.018) 0.047 (0.042) 0.038 (0.029) (0.021) 0.126*** (0.043) 0.094*** (0.024) (0.020) 0.184*** (0.041) 0.098*** (0.023) (0.023) 0.038 (0.044) 0.029 (0.030) (0.027) 0.085*** (0.013) Yes Yes Yes 7,181 0.39 0.075*** (0.014) Yes Yes Yes 7,181 0.39 −0.220*** (0.013) 0.060*** (0.011) Yes Yes Yes 14,028 0.44 0.015 (0.014) Yes Yes Yes 7,659 0.41 0.032** (0.014) Yes Yes Yes 7,591 0.41 Urban collective enterprises −0.211*** −0.188*** −0.272*** −0.195*** −0.224*** Notes: Estimations for individuals aged 16–70 years old who declared working at least part of the year and earning (positive) wages. Reference category for ownership is state-owned enterprises. Standard errors in brackets. *** = p < 0.01, ** = p < 0.05, * = p < 0.10. Source: Authors’ calculations. l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . / e d u a d e v / a r t i c e - p d l f / / / / 3 4 2 1 8 4 1 6 4 4 3 7 2 a d e v _ a _ 0 0 0 9 9 p d / . f b y g u e s t t o n 0 9 S e p e m b e r 2 0 2 3 200 Asian Development Review Table A.2. City Determinants of Total and Per Hour Earnings— Cities Common to 2002, 2007, and 2013 Surveys Total Earnings Hourly Earnings (1) (2) (3) (4) (5) (6) (7) 0.130*** (0.034) 0.137*** (0.035) (0.043) 0.119 (0.088) 0.155*** (0.046) −0.134* (0.075) 0.134*** (0.042) −0.101 (0.061) (0.069) 0.326*** (0.112) 0.366** (0.170) 0.390** (0.165) 0.140*** 0.166** 0.112** (0.044) (0.049) (0.069) −0.106 −0.066 −0.099 (0.103) (0.065) (0.073) −0.194* −0.141** −0.158*** −0.172*** −0.210*** −0.118*** −0.147*** (0.062) (0.101) 0.173* (0.100) 0.296* (0.156) 0.411*** (0.147) 0.111*** (0.039) −0.025 (0.055) −0.045 (0.055) 0.145* (0.079) (0.058) 0.187* (0.097) 0.281* (0.142) 0.374*** (0.137) 0.088*** (0.024) 0.414** (0.159) 0.099*** (0.034) 0.443*** (0.133) 0.077*** (0.026) −0.050 (0.051) 0.236* (0.129) (0.052) 0.137 (0.091) 0.151* (0.077) 0.140 (0.093) −0.018 (0.011) −0.019 (0.012) −0.022 (0.016) −0.242** −0.258** −0.354** −0.022 (0.088) (0.098) (0.101) −0.215 (0.184) 0.005 (0.022) −0.061* (0.032) 0.045 (0.105) −0.269 (0.195) 76 0.91 Observations R2 60 0.70 60 0.88 60 0.92 60 0.92 60 0.93 76 0.90 Notes: Balanced panel of cities (20 cities common to 2002, 2007, and 2013; 38 cities common to 2002 and 2013). Time dummies are included in all specifications. Standard errors in brackets. *** = p < 0.01, ** = p < 0.05, * = p < 0.10. Source: Authors’ calculations. Density Density 2007 Density 2002 Migrants Migrants 2007 Migrants 2002 Land area Land area 2007 Land area 2002 Market potential Market potential 2007 Market potential 2002 Distance to seaport Distance to seaport 2007 Distance to seaport 2002 Diversity Diversity 2007 Diversity 2002 (0.069) 0.206* (0.107) 0.267 (0.179) 0.332** (0.162) 0.121*** (0.041) −0.034 (0.063) −0.069 (0.059) 0.077 (0.145) 0.189 (0.202) 0.015 (0.197) −0.005 (0.020) −0.004 (0.029) −0.040 (0.029) (0.140) 0.202 (0.275) 0.222 (0.268) l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . / e d u a d e v / a r t i c e - p d l f / / / / 3 4 2 1 8 4 1 6 4 4 3 7 2 a d e v _ a _ 0 0 0 9 9 p d . / f b y g u e s t t o n 0 9 S e p e m b e r 2 0 2 3
Descargar PDF