UNIVERSITAT
ROVIRA I VIRGILI
DEPARTAMENT D’ECONOMIA

WORKING PAPERS
Col·lecció “DOCUMENTS DE TREBALL DEL
DEPARTAMENT D’ECONOMIA - CREIP”

Urban Employment in Small Businesses and
the Level of Economic Development: Evidence
from Chinese Cities
Ana I. Moreno Monroy
Shu Yu
Victoria Euse

Document de treball n.23- 2015

DEPARTAMENT D’ECONOMIA – CREIP
Facultat d’Economia i Empresa

UNIVERSITAT
ROVIRA I VIRGILI
DEPARTAMENT D’ECONOMIA

Edita:
Departament d’Economia
www.fcee.urv.es/departaments/economia/publi
c_html/index.html
Universitat Rovira i Virgili
Facultat d’Economia i Empresa
Avgda. de la Universitat, 1
43204 Reus
Tel.: +34 977 759 811
Fax: +34 977 300 661
Email: sde@urv.cat

CREIP
www.urv.cat/creip
Universitat Rovira i Virgili
Departament d’Economia
Avgda. de la Universitat, 1
43204 Reus
Tel.: +34 977 558 936
Email: creip@urv.cat

Adreçar comentaris al Departament d’Economia / CREIP

Dipòsit Legal: T - 1089 - 2015
ISSN edició en paper: 1576 - 3382
ISSN edició electrònica: 1988 - 0820

DEPARTAMENT D’ECONOMIA – CREIP
Facultat d’Economia i Empresa

Urban Employment in Small Businesses and the Level of Economic
Development: Evidence from Chinese Cities

Ana I. Moreno-Monroy*
Shu Yu**
Victoria Euse***

Based on a panel of Chinese cities over the period 2004-2009, we analyze the relationship between the
level of economic development and the share of urban employment in small businesses. We find that
this relationship can be described by an inverted U-shape. In cities with lower levels of economic
development, the restructure of the state sector along with a booming service industry is associated
with a higher share of employment in small businesses. On the other hand, in cities with higher levels
of economic development, a more vibrant manufacturing sector is related to a lower share of
employment in small businesses.

1. Introduction
The pattern and nature of urban employment in small businesses is expected to change during
economic development. In particular, the amount and quality of urban labor employed in small
business is different during the three stages of economic development identified by Porter et al.
(2002). During the factor-driven stage, if the rural-urban migration rate outpaces the creation of
urban employment in the manufacturing sector, unemployed migrant workers who wish to stay
in cities resort to (self)employment in small-scale, low productivity activities (Harris and Todaro,
1970). If large-scale industrialization takes off during the investment-driven stage, a first turning
point takes place, after which the presence of large enterprises with scale economies increases,
*

Corresponding author. Marie Curie Research Fellow at the Economics Department and the Research Centre on
Industrial and Public Economics (CREIP), Rovira i Virgili University. E-mail address:
ana.i.moreno.monroy@gmail.com.
**
Postdoctoral fellow, Skalny Center, University of Rochester, Rochester.
***
International operations business graduate, Novo Nordisk A/S, Zürich.

1

leading to more demand for wage workers, an increase in the opportunity cost of selfemployment and a consequent decrease in the share of urban employment in small businesses
(Acs and Naude, 2011).
When opportunities in large-scale resource-intensive activities are exhausted, there is a shift
towards activities with higher knowledge content during the innovation-driven stage. The
predictions on the share of urban employment at this stage are ambiguous. On the one hand,
specialization in innovative activities opens up the opportunity for business ownership, new
specialized market niches for small businesses and a higher demand for entrepreneurship as an
occupational choice. When these forces are sufficiently strong, a second turning point may take
place, leading to a rising share of urban employment in small businesses (Gries and Naude,
2010; Wennekers et al., 2005). On the other hand, the share of employment in small businesses
may keep decreasing or remain stagnant at higher levels of development because of
technological progress (Gollin, 2007).
The purpose of this paper is to empirically study the relationship between urban employment
in small businesses and economic development. This relationship is of interest because structural
trends in employment in small businesses matter for the timing and choice of policies. Because
of its effects on employment, innovation and welfare, entrepreneurship has gained an increasing
role in economic development (Acs et al, 1994). Consequently, considerable resources are spent
on small business and microenterprise development programs. A relevant question from a longterm policy perspective is: how is the share of urban employment in small businesses expected to
change over the course of development? If it is the case that this share does not show a linear
pattern over the course of development, relevant questions are when and how to promote
employment creation in urban small businesses. For instance, a policy targeting brute urban
employment creation in small businesses may not be appropriate in cities close to the first
turning point. On the other hand, a policy that focuses on the quality of employment may be
more appropriate for cities that are beyond the second turning point. Furthermore, given the
benchmark of different turning points, a large discrepancy between the share of urban
employment in small businesses and the level of economic development in a city may be a
symptom of underlying institutional weaknesses and market failures, which policies can aim to
correct (Gollin, 2007).
2

Related studies have analyzed the relationship between entrepreneurship and economic
development using cross-sectional data and country-level panel data (Acs et al., 1994; Carree et
al., 2002; Pietrobelli et al., 2004; Wennekers et al., 2005; Li and Zhao, 2011). These studies have
established, theoretically and empirically, the existence of a U-shape relationship between
entrepreneurship or self-employment levels and growth and the level of economic development.
However, the empirical strategy used in these studies does not include a control for unobserved
heterogeneity, which can represent a significant source of estimation bias.
In our empirical specification, we study the nature of the relationship between the share of
urban employment in small businesses and economic development under the assumption of a
non-linear association. To do so, we use a comprehensive panel of over 280 prefecture-cities in
China for the period 2004-2009. China offers an exceptional testing ground for our purposes
because it has not only experienced very rapid and sustained growth rates during the last two
decades, but it has also shown large geographical disparities in the pace of this process. Given
that empirically this association can display a monotonic, (inverted) U-shape or N-shape pattern,
we specify an equation containing a third degree polynomial and relevant control variables and
fixed effects. We start, however, by performing an OLS cross-section regression in order to
assess the magnitude of the bias introduced by not controlling for unobserved heterogeneity, and
compare the results with regressions using a fixed-effects (FE) panel data estimator. In line with
previous studies, in the OLS cross-section results we find a U-shape relationship between the
share of urban employment in small businesses and the level of economic development.
However, when using a panel data FE (or an Instrumental Variables (IV) estimator), we find
an inverted U-shape pattern. We argue that this result is reasonable not only econometrically, but
also in the context of China. The reasons are that, first, we focus on urban as opposed to
aggregate employment, so an increasing share of urban employment in small businesses at low
levels of development can be expected, and second, that an increasing share at higher levels of
development is not to be expected in China, given the strong presence of large-scale activities
and the rate of technological progress on the studied period.
Based on the results, we estimate that the first turning point, or the point where the share of
urban employment in small businesses is at its maximum, occurs at an urban disposable income
level between 8,476 and 9,765 CNY per capita (p.c.) (about $1,238 -$1,425 p.c. using the
3

average USD/CNY conversion rate for 2009). Additionally, as the effect of the control variables
is expected to change over economic development, we establish a separate moderating effect of
each control variable by creating an individual interaction with our proxy for the level of
economic development. The separate estimations for different levels of economic development
reveal that some cities may be still at the factor-driven stage, where the restructuring of the state
sector along with more opportunities in the service sector have determined an upward trend of
urban employment in small businesses. On the other hand, the development of a more vibrant
manufacture industry has crowded out employment in small businesses in cities experiencing
large-scale industrialization, or the investment-driven stage. We find no evidence of a higher
share of employment in small businesses in cities at the top tier of the income distribution.
The article is organized as follows. The next section provides theoretical predictions on the
nature of employment in small businesses during different stages of economic development. The
third section introduces the case of China. We present styled facts on small businesses, industrial
structure and spatial distribution. The fourth section describes the empirical strategy and results.
The fifth section presents the econometric specification for different stages of economic
development. The last section provides a discussion of the results and concludes.

2. Review of theoretical predictions and empirical findings
According to Porter et al. (2002), the process of economic development can be divided into three
broad stages: factor-driven, efficiency-driven and innovation-driven. The pattern and nature of
employment in urban small businesses are expected to be different in each of these stages
(Figure 1). At the beginning of the factor-driven stage, most employment is absorbed by smallscale agricultural activities. The expansion of the modern sector causes a relocation of surplus
labor in rural areas into the industrial sector of urban areas. However, as the migration rate of
workers into cities outpaces the creation of wage employment, unemployed workers wishing to
stay in cities become (self)employed in small-scale, low productivity (informal) activities with
little capital requirements (Harris and Todaro, 1970). Thus, employment in urban small
businesses as an occupational choice during the factor-driven stage reflects mostly the inability
of individuals to access wage employment.

4

Most empirical studies do not make a distinction between urban and rural employment in
small businesses. Consequently, the starting point at low levels of development is at a high share
of employment in small businesses. Rural and urban small businesses are different in nature,
however, given differences in the organization of production, production technology and labor
productivity, among others. The distinction between rural and urban employment is important
because during the first stages of development, the pace of urbanization varies greatly across
countries and regions. An aggregate measure of employment is not able to capture the
substitution between rural and urban small scale activities at low levels of development.
When large-scale industrialization takes off during the efficiency-driven stage, employment in
small businesses increasingly becomes a less attractive occupational choice for marginal
entrepreneurs given higher opportunity costs of managing compared to the alternative of wage
employment in larger businesses (Lucas, 1978; Mesnard and Ravallion, 2001). Increasingly
larger urban markets attract the presence of enterprises and industries with scale economies,
which locate closer to their demand market and suppliers and engage in vertical integration,
raising entry barriers for new small businesses (Glaeser et al., 2010; Haltiwanger et al., 2010). At
the first turning point, these forces bring the once increasing share of employment in urban small

% urban employment
in small businesses

businesses to decrease rapidly (Acs and Naude, 2011; Pietrobelli et al., 2004).

Factor-driven stage

Investmentdriven stage

Innovationdriven stage

Path 1

Path 2

Level of Economic Development

FIGURE 1. Share of Urban Employment In Small Businesses During Economic Development
5

At some point, due to increasing costs and competition, the profitability of investmentintensive activities comes to a halt and there is an increasing role of knowledge in generating
value. This stage is accompanied by a sharp reduction in the share of manufacturing activities
and an increase in the share of the tertiary sector, and in particular of modern industries such as
financial and technology services (Acs et at, 1994). A second turning point may take place and
the share of urban employment in small businesses increases again (Path 1 in Figure 1). This
alternative has been proposed based on empirical studies finding a U-shape relationship between
self-employment or entrepreneurial rates and the level of economic development (Wennekers et
al., 2005; Acs et al., 1994; Li and Zhao, 2011). Several possible explanations have been offered.
For starters, modern industries offer more opportunities for business ownership than large-scale
manufacturing activities, which drives up urban self-employment rates. Additionally, there can
be increasing demand for subcontracting with small units as large-scale businesses vertically
disintegrate (Doi and Cowling, 1998). Other factors contributing to the increase of the share of
urban employment in small businesses during this stage of development is increasing demand for
non-standardized, specialized goods and services in wealthier urban areas, which small-scale
entrepreneurs can satisfy (Glaeser et al., 2010), and a higher demand for self-employment as a
superior occupational choice to wage employment in terms of autonomy and self-satisfaction
(Wennekers et al., 2005).
Albeit these elements can indeed cause a resurge in small businesses, they may not be strong
enough to drive a second turning point in the share of small businesses in urban employment.
Available empirical evidence for development and developing countries gives good reasons to
believe that the share of urban employment in small businesses can continue to decrease or
stagnate at higher levels of economic development (Path 2 in Figure 1). For the U.S., Poschke
(2014) reports that the share of small firms has steadily fallen over the period 1977-2009. Gollin
(2007) finds that the self-employment rate falls with income per capita. In particular, for Japan
he reports a steady fall in the share of self-employed in manufacturing and stagnant or decreasing
shares for richer countries such as Denmark and Italy. According to his model, technical change
alone can account for these trends, so the decrease in the share of self-employed occurs even in

6

the absence of country specific distortions, such as regulatory barriers and credit market
inefficiencies.

3. Urban small businesses in China: history and trends
Before 1979, China had a specific policy that guaranteed employment to all workers. Jobs were
directly allocated, wages were controlled and migration was severely restricted through the
system of local registration, hu kou.1 In urban areas, workers were assigned to work in State
Owned Enterprises (SOEs) or collectively owned enterprises. In that context, private ownership
and free allocation of labor were not allowed. This situation changed during the first period of
the transition, with friendlier experimental policies towards Foreign Direct Investment (FDI), the
de-collectivization of agriculture, and the legitimation of township and village enterprises
(TVEs). This process led to a rapid growth of light industries and the re-emergence of private
entrepreneurship (Gregory et al., 2000; Huang, 2008).
In the 1990s, further reforms were carried out to reactivate the economy, and the private
sector experienced the largest expansion since the beginning of the transition. Between 1991 and
1997, the annual growth of private economic activity was nearly 71 percent, and urban
employment in small businesses experienced its highest peak in 1998 (Figure A1 in Appendix
2).2 While private enterprises flourished, SOEs faced a fierce struggle. Since 1995, a fall of total
industrial output led to large scale privatizations and massive layoffs in SOEs (Eesley, 2009;
Tan, 2007; Ghose, 2005). In this period, employment in small units was encouraged by local and
national governments to deal with the temporary economic slowdown and high unemployment
rates in urban areas (Li and Zhao, 2011; Wehrfritz and Seno, 2003). The steep decrease in SOEs
activity together with lower growth rates of employment in small businesses contributed to an
increasing employment share in larger private businesses (Figure A2 in Appendix 2).
During the following decade, some local governments further relaxed their restriction on
private enterprises and new funds were created to encourage the creation of technologically
intensive small businesses. Simultaneously, rural individuals were actively seeking off-farm
jobs, especially in the migrant wage earning sector (Wang et al., 2011). This process was
facilitated by important changes in the migration regulations (Chan and Buckingham, 2008;
Chen and Coulson, 2002; Song and Zhang, 2002).
7

Since the early 2000s, and especially after China’s accession to the WTO in 2002, policies
have been formulated to emphasize the protection of private ownership in order to increase value
added in industrial sectors and further expand employment opportunities through
entrepreneurship. The new regulations of the late 1990s and 2000s gave rise to the proliferation
of science parks, 3 technology business incubators, associated tax incentives for research and
development, and an increasing emphasis on tertiary education (Cai et al., 2007). After a
recovery in economic performance during the early 2000s, the national growth rate of larger
private businesses surpassed that of employment in small businesses (Figure A2 in Appendix 2).
According to the 2009 Global Entrepreneurship Monitor (GEM) Global Report, in the period
2004-2009 China had one of the highest prevalence of growth-oriented entrepreneurs (i.e., those
who expect to create 20 jobs or more) in the world. China’s Early Stage Entrepreneurial Activity
(TEA) rate in 2009 was 18.8, above the average of 11.2 percent for efficiency-driven economies.
The optimism among entrepreneurs with respect to their growth potential evidenced in the period
is in accordance with the institutional reforms benefiting entrepreneurship and the prevalent
high-rates of economic growth. As Zhang (2013) notes, however, the GEM Global report uses
data from large cities only, and is not representative of the situation across the country.
Indeed, small businesses have developed in a very uneven way across China. Early exposure
to foreign markets, openness to FDI and industrialization have fueled rapid development in the
Eastern coastline. These developments increased the opportunities for innovation-driven
entrepreneurs, who also likely benefitted from the increased pool of educated workers in the
period 2004-2009 (Zhang, 2013). Meanwhile, provinces in the West and North have remained
mostly agricultural and reliant on natural resources (Ge, 2009). In these provinces, small
businesses are likely to be driven by necessity rather than by opportunity. Figure A3 in Appendix
2 shows the distribution of average urban disposable income per capita and employment in small
businesses for the urban areas of the prefectures in our sample (data sources and measurement
are detailed in Appendix 1). Remarkably, the richest prefecture city has an average urban
disposable income 4.5 times larger than that of the poorest prefecture city. The variation in the
average share of urban employment is also significant, ranging from 4 to 51%. Some of the
richest prefectures display low shares of employment in small businesses, whereas in some poor

8

prefectures concentrated in the northeast and south, this type of employment is an important
component of the labor market.4
The nature of entrepreneurship is likely to be highly heterogeneous, with necessity-based and
opportunity-based entrepreneurial activities coexisting across cities. The process of economic
development in China has been so rapid and unevenly distributed that even a relatively short
time period such as 2004-2009 evidences considerable structural transformation, making it an
attractive case for empirical analysis.

4. Empirical analysis
Econometric specification.

Given that theoretically the relationship between the share of urban

employment in small businesses and the level of development can take different forms, we
specify a third degree polynomial. Our empirical specification is:

1)

where i is an index for prefecture, t is an index for year, se_share is the share of urban
employment in small businesses over total urban employment, ln(inca) is a measure of economic
development, proxied by the natural logarithm of real urban household disposable income per
capita and X is a matrix of control variables.5 ci represents prefecture-specific fixed effects that
capture unobserved time-invariant effects on the share of urban employment in small businesses.
These include geographical conditions such as ruggedness, access to coastline and natural
resource endowments. yt is a series of time dummies which capture unobserved time-specific
effects common to all the prefectures-cities, such as changes in laws and (macro)economic
shocks at the national level. Finally, ε represents a stochastic error term.
For the main controls, matrix X includes: 1) the ratio of urban employment in larger private
businesses over the sum of this variable and SOEs employment, or “private share”; 2) the share
of employment in secondary sector in total urban employment; and 3) the share of employment
in tertiary sector in total urban employment.
The proposed flexible parametric model, which is widely used in the environmental Kuznets
curve literature, captures three different possible relationships between the share of urban
9

employment in small businesses and the level of income (Shafik and Bandyopadhyay, 1992;
Lieb, 2003): 1) N shape, if α1, α2 and α3 are all statistically significant in the cubic estimation; 2)
(inverted) U-shape, if α1, α2 and/or α3 are statistically insignificant in the cubic estimation, but α1
and α2 are statistically significant in the quadratic estimation (and α2 <0) and; 3) monotonic, if α2
is statistically insignificant in the quadratic estimation.
To further check for the robustness of the results we include additional control variables that
have been identified in the literature as possible determinants of the share of employment in
small businesses. These include: a variable measuring competing farm labor opportunities
(Groom et al., 2006); the level of education of the workforce (Glaeser, 2007; Bartik, 2005; Wu,
2002), and population density (Bartik, 2005; Mueller, 2005; Glaeser, 2007). We use the share of
cultivated area of total area as a proxy for job opportunities in the agricultural sector; the share of
secondary school graduates per capita as a proxy for the level of educated workforce; and the
number of inhabitants per square kilometer as a proxy for population density. Descriptive
statistics of all the variables used in the empirical analysis are available in Table A1 in Appendix
2.
In order to estimate equation (1), we use a fixed effects (FE) panel estimator (or ‘within’
estimator). These FE estimates may be biased and inconsistent if the errors are not uncorrelated
with the dependent variables, which may be the case if there is a two-way causal relationship
between the level of income and the share of urban employment in small businesses, if there is
measurement error or if there are omitted variables. To deal with these concerns, we implement a
Two Stage Least Squares (TSLS) Instrumental Variables (IV) estimation of equation (1). As an
instrument for urban disposable income per capita, we propose the use of a measure of external
market potential for each prefecture j and year t (MPjt), defined as sum of the real GDP per capita
of all other prefectures in the sample, excluding own-prefecture GDP per capita, weighted by the
inverse of the bilateral distance between each pair of cities (dij). This distance is calculated using
the great-circle distance formula. In formal terms, the external market potential measure is given
by:
∑

(2)

For the case of China, it has been empirically shown that market potential significantly
explains income differences across prefecture-cities (Moreno-Monroy, 2011; Bosker et al.,
10

2012). We argue that external market potential does not directly affect the share of urban
employment in small businesses, and that can be expected to influence this variable only through
its effect on income levels. The rationale is that small business performance is determined
mainly by local economic factors. This is likely if small businesses produce non-tradable goods
and services, if they are oriented towards local urban markets given a lower export capacity, or if
the products and services offered by small businesses can be easily replicated across locations (a
likely scenario for the case of standardized, low-value added products and services). We will
establish the appropriateness of this instrument through a battery of tests.
Besides the fixed-effects (IV) estimations, we perform the following OLS cross-section
regression in order to compare our results with previous studies and measure the extent of the
bias introduced by not eliminating unobserved heterogeneity:

Estimation results.

Table A2 in Appendix 2 reports the estimation results for equation (1)

using the TSLS estimator. As evidenced by the reported first stage F-values, the external market
potential variable and its square are highly significant in explaining real urban disposable income
and its square. The F-value of the Cragg-Donald Wald test indicates that the instruments are not
weak.6 The Kleibergen-Paap LM statistic indicates that the equation is not underidentified. From
this we can conclude that our instrument passes the relevance and exogeneity tests. However,
based on the results of the Hausman endogeneity test, we cannot reject the null hypothesis that
the specified endogenous regressors can actually be treated as exogenous, in which case the
Fixed Effects (FE) panel estimator results are more efficient and are consequently preferred over
the IV results. 7 We also perform Hausman tests for random vs. fixed effects. As can be seen in
Table 1, all tests are significant at 1 percent level, which suggests that the Random-Effects (RE)
estimator will produce biased estimates. Thus, FE estimator is the preferred estimator in our
study.
Thus, we consider the FE results, reported in columns 4-6 of Table 1, as our baseline results.
Columns (1) to (2) contain the results of estimating equation (3) using OLS. In line with previous
studies using a similar methodology, the point estimate of urban household disposable income
11

per capita is significant and positive at the 99% level of confidence, giving support to the
existence of a U-shape relationship between the share of urban employment in small businesses
and the level of economic development.
Strikingly, this result is entirely reversed once we eliminate unobserved heterogeneity. As
columns (4) to (6) show, both α1 and α2 are significant at the 95 percent level of confidence in the
quadratic estimation. Given that α2 has a negative sign, this result points to the existence of an
inverse U-shape relationship between the share of urban employment in small businesses and the
level of development. The share of employment in the tertiary sector and the private share have a
positive significant effect on the share of urban employment in small businesses. The share of
employment in the secondary sector has a negative albeit statistically insignificant effect.
Regarding additional variables, cultivated land, education and population density are not
statistically insignificant in explaining the share of urban employment in small businesses at the
95 percent level of confidence.8
Based on the estimates in columns (4) to (6), we can establish that the first turning point, or
the point where the share of urban employment in small businesses is at its maximum, occurs at
an urban disposable income level between 8,476 and 9,765 CNY p.c. (about $1,238 -$1,425 p.c.
using the average USD/CNY conversion rate for 2009). Cities to the left of this turning point had
over the period on average secondary, tertiary and private shares of 19, 28 and 22 percent
respectively, while the values for those to the right of this first turning point were 27, 32 and 34
percent.
Finally, as column (3) shows, α1, α2 and α3 are statistically insignificant in the cubic
estimation, so we find no support for the occurrence of a second turning point at higher levels of
economic development. 9
In order to rule out possible non-stationarity problems, we performed a panel Fisher-type unit
root test, where ADF unit-root tests are performed on each panel. The test results (not shown, but
available upon request) reject the null hypothesis that all panels contain unit roots at the
significant level of 1 percent and suggest the alternative that at least one panel is stationary.
Furthermore, to rule out the possibility of omitted variables, we included a lagged dependent
variable to Column (4) in Table-1 using FE estimator and IV estimator. Then, to deal with the
potential over-significant t-statistics, we use a cross-sectional time-series FGLS estimator on the
12

model specification of Column (4) in Table 1. This estimator allows estimation in the presence of
AR(1) autocorrelation within panels and cross-sectional correlation and heteroskedasticity across
panels. The results (not shown here) do not invalidate our previous findings. They all have the
expected signs and the FGLS estimator also gives significant results. We still consider the results
on Table 1 as our preferred estimates because including lagged dependent variables forces us to
drop 16 percent of our sample, largely reducing the efficiency of the estimators.

5. Econometric specification for different stages of economic development
As the control variables included in equation (1) may have a different effect at different stages of
economic development, we interact each control variable with urban household income per
capita. The specification is:

(4)

where Z is the analyzed control variable. The point at which marginal effect of Z changes sign is
.10
We estimate equation (4) also using a FE panel estimator, and calculate and plot the marginal
effects for each variable of interest. The regression results can be found in Table A3 in Appendix
2. Figure 2 plots the marginal effect of the private share, the share of employment in the
secondary sector and the share of employment in the tertiary sector.
As panel panels a) and c) show, the private share and the share of employment in the tertiary
sector have a positive effect on the share of urban employment in small businesses for low and
middle levels of urban household income per capita. This effect is statistically significant up to
approximately 13,000 CNY p.c. for the private share and 10,600 CNY p.c. for the tertiary share.
For higher levels of income, this effect becomes insignificant.

13

14

0.646
0.106***
0.035

0.600
0.116***
0.032

0.048***
0.017
-0.160***
0.026
-0.068**
0.029
-0.120***
0.016
-1.191**
0.468
0.016***
0.003

OLS (2)
-2.046***

OLS (1)
-2.261***
11.130
-0.661
1.194
0.021
0.043

FE (3)
6.698
0.446
-0.067***
0.025

FE (4)
1.215***
0.473
-0.075***
0.026

FE (5)
1.364***

0.453
-0.079***
0.025

FE (6)
1.430***

0.102**
0.101**
0.048
0.043
Secondary share
-0.173
0.109
Tertiary share
0.220***
0.245***
0.075
0.078
Cultivated area
0.020
0.094
Education
0.544*
0.330
Population density
-0.003
0.008
Hausman test RE vs. FE p-value
0.000
0.007
0.000
0.000
Prefecture fixed-effects
No
No
Yes
Yes
Yes
Yes
Year fixed effects
No
No
Yes
Yes
Yes
Yes
Observations
2,016
1,944
1,843
1,843
1,568
1,745
R-squared
0.112
0.169
0.138
0.138
0.179
0.17
Note: Standard errors clustered (robust) by prefecture. A constant term is included but not reported. t values are reported
below the point estimates; *** p<0.01, ** p<0.05, * p<0.1

Private share

Household Income^3

Household Income^2

Household Income

TABLE 1. Regression Results
Dependent variable: Share of urban employment in small businesses

Regarding the private share, it is important to note that there are two means by which this
share can increase: through an increase in private employment in larger businesses or through a
decrease of employment in SOEs. As can be seen in Figure A4 in Appendix 2, for low and lowmiddle levels of urban disposable household income per capita, the decrease in employment in
SOEs outpaces growth in private employment in larger businesses. For higher levels of
disposable income per capita, the opposite is true. Thus, for low and middle levels of income, the
positive sign of the private share’s marginal effect is related to the increased availability of laidoff SOEs workers who are not absorbed by the private larger businesses. A crowding out effect
where employment in small businesses is reduced because of a relative increase in employment
in larger businesses happens only at the highest levels of income (23,535 CNY p.c.) but this
effect is statistically insignificant.

FIGURE 2. Marginal Effect of the Different Determinants of Employment in Small Businesses
For Different Levels of Household Income

-.2

0

.2

.4

.6

a) Marginal effect of private employment share

8

8.5

9
9.5
10
log of urban disposable income per capita

10.5

15

-1

-.5

0

.5

b) Marginal effect of employment in the secondary sector

8

8.5

9
9.5
10
log of urban disposable income per capita

10.5

-.5

0

.5

1

c) Marginal effect of employment in the tertiary sector

8

8.5

9
9.5
10
log of urban disposable income per capita

10.5

According to the estimates in Table A3 in Appendix 2, the point at which the crowding-out
effect of the secondary sectors starts operating (i.e., the point at which secondary employment
share’s marginal effect changes sign from positive to negative) occurs at 4,861 CNY p.c.
However, as Panel b) in Figure 2 shows, the marginal (and negative) effect on the share of urban
employment in small businesses is significant only for medium and high-levels of income
(starting at around 10,938 CNY p.c.).

6. Discussion and conclusions
In this article, we have analyzed the nature of the relationship between employment in small
businesses and the level of economic development based on a sample of over 280 Chinese
prefecture-cities during the period 2004-2009. We find that this relationship can be represented
by an inverted U-shape pattern. We estimate that the first turning point, or the point where the
16

share of urban employment in small businesses is at its maximum, occurs at an urban disposable
income level between 8,476 and 9,765 CNY p.c. ($1,238-$1,425 p.c.). During the studied period,
some cities in China have moved beyond this turning point and are experiencing a reduction in
the share of urban employment in small businesses, while others are still experiencing increasing
rates of this variable. Our results do not give support to the existence of a N-shape pattern in the
proposed relationship for the case of China, where cities at the top of the income distribution
would be experiencing again increasing shares of urban employment in small businesses.
Furthermore, they suggest that previous findings of a U-shape do not longer hold once
unobserved heterogeneity is eliminated.
In the light of the three stages of economic development, our results suggest that during the
factor-driven stage when urban areas experience a takeoff in terms of income, there is a pool of
urban workers not absorbed by larger businesses that find employment in small businesses as an
alternative occupational choice. Matching the results, these workers most probably undertake
activities in the service sector. The negative effect of the share of secondary employment is
related to the investment-driven stage of economic development when, as the urban
manufacturing sector expands, more people opt for wage employment in detriment of selfemployment or employment in small businesses. Nevertheless, in reality the decrease of
employment in small businesses in the midst of a more dynamic secondary sector may be
overstated given that our estimations do not include employment in the informal economy.
On the other hand, the share of urban employment in small businesses may be still decreasing
in the most developed cities in China (e.g. Beijing and Shanghai), as long as these cities continue
to host a considerably large secondary sector and an underdeveloped modern service sector.11 In
this case, the technological progress model proposed by Gollin (2007) seems to better suit the
case of China than alternative explanations related to extensive subcontracting or small business
conglomerates.

Notes
1

Under the hu kou system, each person was registered in a particular locality and the change of registration status
was not possible because it granted people access to food rations, education and health services and social security
(Ghose, 2005).
2
See Appendix 1 for a definition of small businesses.

17

3

As of 2007 there were nearly six thousand industrial parks and 58 national-level science parks.
One clear example of heterogeneity within the province level is Guangdong. Out of 21 prefectures in this province,
12 are located in the low and low-middle income quartiles and 9 in the middle-high and high income quartiles. In
terms of employment in small businesses, 19 have average shares above 20.5 percent (the only two exceptions are
the cities of Guangzhou and Foshan).
5
These variables have been winsorized to reduce the influence of extreme outliers.
6
Given that our model is just identified, we report the results of the Cragg-Donald Wald test instead of the Hansen J.
7
Alternatively, we used a two-step system-GMM estimator, where household income and household income square
are treated as endogenous while the rest explanatory variables were treated as exogenous variables. For endogenous
variables, two periods lags are used as instruments in the first-difference equations and their once lagged firstdifferences were used in the levels equation. The results are not significantly different from the ones presented
below so they are not shown here, but they are available upon request.
8
The same result is obtained when we proxy education by the share of tertiary school graduates per capita, or
expenditure on education per capita.
9
These results hold after including control variables.
10
This point is given by the first order condition (δ(se_share it))/(δZit )=δ3 ln(inca)it+ δ4 =0
11
Data from the City Statistical Yearbook reveals that the share of urban employment in modern services (i.e.,
banking and insurance, real state, ITC services, rent and business services and scientific research) was 4.23 percent
in 2004 and 4.25 percent in 2007.
4

18

References
Acs, Z; Audretsch, D; Evans, D (1994) Why does self-employment rate vary across countries
over time? Discussion Paper No. 871. CEPR. London.
Acs, Z; Naude, W (2011) Entrepreneurship, Stages of Development and Industrialization.
Working paper N. 2011/80. United Nations University: Maastricht
Bartik, J (2005) Solving the Problems of Economic Development Incentives. Growth and
Change. Vol. 36, N 2, pp 139-166.
Bosker, M., Brakman, S., Garretsen, H. and Schramm, M. (2012) Relaxing Hukou. Increased
labor mobility and China's Economic geography. Journal of Urban Economics, 72: 252266.
Cai, H; Todo, Y; Zhou, L. (2007) Do multinationals' R&D activities stimulate indigenous
entrepreneurship? evidence from china's "Silicon valley". Working Paper, 13618. National
Bureau of Economic Research: Cambridge, MA.
Carree, M; Van Stel, A; Thurik, R; Wennekers, S (2002) Economic Development and business
ownership: An analysis using data of 23 OECD countries in the period 1976 – 1996. Small
Business Economics. Vol. 19, N 3, pp 271-290.
Chan, K (2007) Misconceptions and Complexities in the Study of China’s Cities: Definitions,
Statistics, and Implications. Eurasian Geography and Economics. Vol. 48, N 4, pp 383412.
Chan, K.; Buckingham, W (2008) Is China abolishing the Hukou System? The China Quarterly.
Vol. 160, pp 818-855.
Chen, A; Coulson, E (2002) Determinants of Urban Migration: Evidence from Chinese Cities.
Urban Studies. Vol. 39, N 12, pp 2189-2197.
Doi, N and Cowling, M (1998) The Evolution Of Firm Size and Employment Share Distribution
in Japanese and UK Manufacturing: A Study of Small Business Presence. Small Business
Economics, 10: 283-292
Eesley, C (2009) Entrepreneurship and China: history of policy reforms and institutional
development. Working Paper. Stanford University: Stanford.
Ge, Y (2009) Globalization and Industry Agglomeration in China. World Development. Vol. 37,
N 3, pp. 550 – 559.
19

Ghose, K (2005) Employment in China: recent trends and future challenges. Employment
Strategy Papers. International Labour Organization: Geneva.
Glaeser, E (2007) Entrepreneurship and the city. Working paper 13551. National Bureau of
Economic Research: Cambridge, MA
Glaeser, E; Rosenthal, S; Strange, W (2010) Urban Economics and Entrepreneurship. Journal of
Urban Economics. Vol. 67, pp. 1- 14.
Gollin, D (2007) Nobody's business but my own: Self-employment and small enterprise in
economic development. Journal of Monetary Economics 55, Vol. 2, pp. 219-233
Gregory, N; Tenev, S; Wagle, D (2000) China’s emerging private enterprises: Prospects for the
new century. World Bank: Washington DC.
Gries, T; Naude, W (2010) Entrepreneurship and structural economic transformation. Small
Business Economics. Vol. 34, pp. 13-29.
Groom, B; Grosjean, P; Kontolean, A; Swanson, T; Zhang, S (2006) Relaxing constraints with
compensation: evaluating off-farm labour responses to reforestation policy in China.
Working Paper. University of London: London.
Haltiwanger, J; Jarmin, R; Miranda, J (2010) Who creates jobs? Small vs. Large vs. Young.
Working Paper 16300. National Bureau of Economic Research: Cambridge, MA
Harris, J; Todaro, M (1970) Migration, unemployment and development: A two sector analysis.
The American Economic Review. Vol. 60, N 1, pp. 126 – 142.
Huang, P (2009) China’s neglected informal economy: reality and theory. Modern China. Vol.
35, pp. 405-440.
Huang, Y (2008) Capitalism with Chinese characteristics: Entrepreneurship and state during the
reform era. Cambridge University Press: New York.
Lucas, R (1978) On the size distribution of business firms. The Bell Journal of Economics. Vol.
9, pp. 508-523.
Li, K; Zhao, C (2011) Determinants of Self-employment in China: Evidence from Cross-regional
Data. China and World Economy. Vol. 19, N 3, pp. 49 -67.
Lieb, CM (2003) “Environmental Kuznets Curve: A Survey of the Empirical Evidence and of
Possible Causes”, Discussion Paper Series 391, University of Heidelberg

20

Mesnard, A; Ravallion, M (2001) Wealth Distribution and Self-Employment in a Developing
Economy. Discussion Paper No. 3026. Centre of Economic Policy Research – CEPR -. 32
Moreno-Monroy, AI (2011) Market access and the heterogeneous effect of shocks on wages:
evidence from Chinese cities. Papers in Regional Science. Vol 90, N 1, pp. 9-25
Mueller, P (2005) Entrepreneurship in the Region: Breeding Ground for Nascent Entrepreneurs?.
Small Business Economics. Vol. 27, N 1, pp. 41-58.
Pietrobelli, C; Rabetolli, R; Aquilina, M (2004) An empirical study of the determinants of selfemployment in developing countries. Journal of International Development. Vol. 16, N 6,
pp. 803-820.
Porter, M; Sachs, J; McArthur, J (2002) Executive Summary: competitiveness and stages of
economic development. In Michael E. Porter, Jeffrey D. Sachs, Peter K. Cornelius and
John W. McArthur, eds, The Global Competitiveness Report 2001-2002. Oxford University
Press: New York, pp. 16-25.
Poschke, M. (2014) The firm size distribution across countries and skill-biased change in
entrepreneurial Technology. IZA Discussion Paper 7991, Bonn: Institute for the Study of
Labor
Shafik, N; Bandyopadhyay, S (1992) “Economic Growth and Environmental Quality: Time
Series and Cross-Country Evidence”, World Bank Policy Research Working Paper, WPS
904, Washington DC: World Bank.
Song, S; Zhang, K (2002) Urbanization and City Size Distribution in China. Urban Studies. Vol.
39, N 12, pp. 2317-2327.
Tan, J (2007) Phase transition and emergence of entrepreneurship: The transformation of
Chinese SOEs over time. Journal of Business Venturing. Vol. 22, pp. 77-96.
Wang, X; Huang, J; Zhang, L; Rozelle, S (2011) The rise of migration and the fall of selfemployment in rural China’s labor market. China Economic Review. Vol. 22, pp. 573-584.
Wehrfritz, G; Seno, A (2003) Chugging along; no one saw China as one of the big ‘engines’ that
could drive the world out of a slump until now. Newsweek (International Edition), June
23rd, p42.
Wennekers, S; Van Stel, A; Thurik, R; Reynolds, P (2005) Nascent entrepreneurship and level of
economic development. Small Business Economics. Vol. 24, pp. 293–309.
21

Wu, X (2002) Embracing the Market: Entry to Self-Employment in Transitional China, 1978 1996. Research Report No. 02-526. Populations Study Center, University of Michigan:
Ann Arbor.
Zhang, Y. (2013) Entrepreneurship development in China: A multilevel approach. PhD
dissertation, Eindhoven University of Technology.

Appendix 1

Data sources and measurement

We have gathered annual data on 280 cities at the prefecture and prefecture-city level for the
period 2004-2009 from two available sources, the City Statistical Yearbook (CSY) and the
Regional Statistical Yearbook (RSY). Unfortunately, there is no data available prior 2004 on
private employment from these or other sources. Original data are collected from local and
provincial statistical bureaus and reported by the National Bureau of Statistics. Even though both
sources report values on some of the same variables, in some cases the numbers diverge. As we
do not have information on the quality of each source, in order to test the quality of the two
sources for comparable variables we aggregate the prefecture data to the province level and
compare the resulting values with those statistics available at the National Statistical Yearbook
(NSY). The RSY was chosen over the CSY as our source as it showed to be more consistent with
aggregates from the NSY. This source also provides more relevant variables and is more
comprehensive in terms of the number of prefectures included.
Within the administrative division of China, prefectures are at the second level after
provinces. There are 333 divisions composed by 283 prefecture-cities, 17 prefectures, 30
autonomous prefectures and 3 leagues. Prefectures are subdivided into counties, autonomous
counties and cities. In China, cities (shì) refer to the continuous core urbanized administrative
area. Total urban area (市 区, shìqū) further includes city adjacent districts and sub-districts
containing both residential and industrial suburbs. The main difference with the definition of the
city as the metropolitan area is that the latter includes the commuting area to the city core. In
this article we follow the definition of shìqū when we refer to the urban area of the city. This
definition describes the city core with high-density built up area where the largest agglomeration
of population within the prefecture is located (Chan, 2007).

22

The national, provincial and prefecture-city statistics make a distinction between
employment in private enterprises with less than eight workers (getihu) and employment in
private enterprises with more than eight workers (siying qiye). Allegedly this cut-off of eight
employees was attributed to Marxist theories on the maximum size a private business could have
before it constituted exploitation of labor (Eesley, 2009). In any case, it served as a cut-off point
for the important reforms approved in 1988 on The First Plenary of the Seventh People’s
Congress, which allowed the existence of private enterprises with more than eight employees.
Given our interest in urban small businesses, we use data on urban employment in private
businesses with less than eight workers. Although we cannot distinguish in the data the sectoral
and quality composition of small businesses, our dependent variable in principle covers all
economic activities, ranging from low-hierarchy services (such as retail trade) and artisanal
manufacturing to high-hierarchy services (such as professional services).
Our variable of interest also includes one-person business or the self-employed. However, it
does not include persons not officially registered as self-employed. The bulk of migrants
(nongmingong) coming to the cities from rural areas do not have local registration and are
missing from the official data (Huang, 2009). It is likely then that our data underestimates the
real number of self-employed, possibly causing a bias in the estimations.
To construct our dependent variable, the share of urban employment in small business, we
divide the number of urban employees in businesses with less than eight workers by the total
number of urban employed persons. We use urban household disposable income per capita as
our proxy for economic development. This and all other nominal variables introduced later on
have been deflated using China’s Provincial Annual CPI with 2000 as the base.

23

Appendix 2

Tables and Figures

FIGURE A1. Employment Share in SOES, Urban Private Larger Private and Small Private
Businesses, 1994-2009
70
60
50

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

40
30
20
10
0

Rate of urban employment in small businesses (%)
Rate of urban employment in SOEs (%)
Rate of urban employment in private enterprises (%)

Source: China Statistical Yearbook. National Bureau of Statistics of China (1994-2010)

FIGURE A2. GDP per Capita and Rate of Urban Private Employment and Employment In Small
Businesses, 1994-2009
30000

20

25000

15

20000
15000

10

10000

5

5000
2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

0
1994

0

GDP per capita
Rate of urban employment in private enterprises (%)
Rate of urban employment in small businesses (%)

Source: China Statistical Yearbook. National Bureau of Statistics of China (1994-2010)

24

FIGURE A3. Geographical distribution of variables of interest
a) Average household income per capita, 2004-2009

b) Average urban employment in small businesses, 2004-2009

25

TABLE A1. Descriptive Statistics
Variable

Obs

Mean

Std.

Min

Max

Dev.
Share urban employment in small

1913

0.22

0.11

0.00

0.90

1929

10284

3319

3634

29467

Private share

1896

0.30

0.17

0.00

0.93

Share of employment in

1873

0.24

0.11

0.00

0.46

1873

0.31

0.08

0.06

0.49

Cultivated area

1769

0.22

0.16

0.00

0.61

Education

1981

0.02

0.01

0.00

0.07

Population density

1964

381.46

640.11

0.78

24880

businesses
Real urban disposable income per
capita

secondary sector
Share of employment in tertiary
sector

TABLE A2. IV Regression Results
Dependent Variable: Share of urban employment in small businesses
(1)

Private share

6.878***

6.281**

6.675***

2.491

2.444

-0.417***

-0.383***

-0.406***

0.139

Household Income Square

(3)

2.375

Household Income

(2)

0.147

0.143

0.068
0.064

Secondary share

-0.177
0.124

26

Tertiary share

0.206**

0.277**

0.100

0.110

Cultivated area

0.135
0.131

Education

0.521
0.451

Population density

-0.021
0.016

Observations

1,818

1,551

1,747

Kleibergen-Paap LM statistic P-value

0.000

0.001

0.018

Cragg-Donald Wald F-value

13.74

8.24

11.98

First stage F-value (Household Income)

23.20

12.12

22.23

First stage F-value (Household Income

19.23

9.75

18.25

Square)
Note: Clustered (robust) standard errors by prefecture. t values are reported below the point
estimates; *** p<0.01, ** p<0.05, * p<0.1

TABLE A3. Interaction Models Regression Results
Dependent variable: Share of urban employment in small businesses
(1)

Private share

0.437

0.607

0.851

0.559

0.534

-0.022

-0.031

-0.042

0.035

Household Income Square

(3)

0.615

Household Income

(2)

0.032

0.031

1.671**
0.747

Private Share*Household Income

-0.166**
0.080

Secondary employment

1.927
27

1.270
Secondary employment*Household Income

-0.227
0.141

Tertiary employment

2.815*
1.527

Tertiary employment*Household income

-0.287*
0.168

Constant

-2.008

-2.727

-4.105*

2.741

2.454

2.327

Prefecture Fixed Effects

Yes

Yes

Yes

Year Fixed Effects

Yes

Yes

Yes

Observations

1,817

1,771

1,771

R-squared

0.161

0.152

0.161

Note: Clustered (robust) standard errors by prefecture. t values are reported below the point
estimates; *** p<0.01, ** p<0.05, * p<0.1

FIGURE A4. Urban Employment in Private Enterprises and SOEs for Different Levels of
Disposable Household Income per Capita
0.6
0.5
0.4
0.3
0.2
0.1
0
Low

Low-middle

Middle

Private employment

Middle-high

High

SOEs employment

Source: China Regional Statistical Yearbook. National Bureau of Statistics of China (20042009)
28