International Review of Economics and Finance 58 (2018) 246–258

Contents lists available at ScienceDirect

International Review of Economics and Finance

journal homepage: www.elsevier.com/locate/iref

The relationship between credit constraints and household

entrepreneurship in China
Dongliang Cai a, Quanyun Song a, *, Shuang Ma b, Yang Dong c, Qiuhua Xu a
a
School of Finance, Southwestern University of Finance and Economics, Chengdu 611130, China
b
School of Economics and Statistics, Guangzhou University, Guangzhou 510006, China
c
Industrial and Commercial Bank of China, Sichuan Branch, Chengdu 610074, China

A R T I C L E I N F O A B S T R A C T

JEL classification: This paper evaluates the impact of credit constraint on entrepreneurship by utilizing 2011 China
E51 Household Finance and Survey micro-data. We measure household credit constraint directly and
L26 use the LPM model to estimate the impact of credit constraint on entrepreneurship. We employ
R29 predetermined credit constraint, quasi-experiment, and instrumental variable to eliminate the
possible endogeneity problem. Empirical estimates show that credit constraint significantly de-
Keywords: creases the probabilities of households to start businesses by around 3 percentage points in gen-
Credit constraint eral. Further analysis shows that credit constraint significantly decreases the propensity of
Entrepreneurship households to entry into enterprise operation, while it has no significant impact on small handi-
Household finance craft operation.

1. Introduction

Entrepreneurship has been embraced by many policy makers as an important development priority because entrepreneurship raises
social welfare via its positive effect on economic growth and job creation. According to the Global Entrepreneurship Monitor (GEM)
China Report published by the Entrepreneurship Research Center of Tsinghua University in 2012, China ranks second among about 60
participating countries and areas. This indicates that China is becoming one of the most active countries in entrepreneurship. Entre-
preneurship creates nearly 60% of the GDP and absorbs about 50% of urban employment, as is shown in the statistics released by the
National Bureau of Statistics of the People's Republic of China. Despite this huge importance for job creation, the employment of the
working population in China is still a serious issue due to the global economic depression. The number of college graduates continues to
expand, and more and more migrant workers return to their hometowns. In this context, the central and local governments of China
implemented the development strategy of promoting employment through entrepreneurship, and have introduced a series of policies to
encourage entrepreneurship of college graduates, returned migrant workers, and unemployed workers. Promoting the development of
the financial market and alleviating credit constraint for would-be entrepreneurs are high on the agenda.
In reality, how to set up enterprises is indeed an issue worthy of discussion. The vast majority of studies show that the lack of initial
capital is a significant barrier to entrepreneurship (Cagetti & De Nardi, 2006; Evans & Jovanovic, 1989). Under the condition that the
technical level achieves minimum requirements, if the initial capital accumulation does not meet the minimum requirements for
entrepreneurial activities, then the entrepreneur can borrow to compensate for the lack of it. However, entrepreneurs under credit

* Corresponding author.
E-mail addresses: [email protected] (D. Cai), [email protected] (Q. Song), [email protected] (S. Ma), oceanfl[email protected] (Y. Dong), xuqh@
swufe.edu.cn (Q. Xu).

http://dx.doi.org/10.1016/j.iref.2018.03.024
Received 16 January 2017; Received in revised form 11 March 2018; Accepted 29 March 2018
Available online 29 March 2018
1059-0560/© 2018 Elsevier Inc. All rights reserved.
D. Cai et al. International Review of Economics and Finance 58 (2018) 246–258

constraint cannot be called “real entrepreneurs” because they do not have enough initial capital, and thus cannot carry out innovative
activities (Buera, 2009). In addition, financing channels are relatively limited due to the under-developed capital market in China, so
entrepreneurs and small businesses owners are more likely to be credit-rationed, and thus may be more vulnerable to credit constraint.
Some studies have examined the relationship between wealth and entrepreneurship in China. Wang (2012) found that the housing
reform of the 1990s in China mitigated the credit constraint of household in China and encouraged entrepreneurship. Li and Wu (2014)
found that surging housing price discouraged entrepreneurship in general though the appreciation of housing value had a positive
wealth effect for potential entrepreneurs. Few studies examined the effect of credit constraint on entrepreneurship in China directly due
to the difficulty of measuring credit constraint and endogeneity of credit constraint.
Using data from the 2011 China Household Finance Survey (CHFS), we measure credit constraint directly. CHFS questionnaire
includes a series of questions on households' liabilities, including business liability, housing liability, automobile liability, financial
liability, education liability, credit card liability, and other liabilities. The CHFS first inquires whether households have bank loans, and
households which have no bank loans are further asked why they have no bank loans. Then, the CHFS askes households whether they
have informal borrowing from friends, relatives, or other institutions. Considering the difficulty of measuring credit constraint with
household survey data, we try to measure credit constraint as accurately as possible. We define two measurements of credit constraint.
Firstly, we define the first measurement of credit constraint (called ‘Measure 1’): Households which have no bank loans and the reason
for having no bank loans is “Needed but did not apply” or “Applied but were rejected” are regarded as suffering from formal credit
constraint. This measurement may under-estimate households' credit constraint since households that have bank loans might not get
enough credit and thus also suffer from credit constraint. The CHFS does not ask the question of whether bank loans obtained are enough
to meet credit demands. As a supplement, we believe that households that have bank loans but did not get enough credit might borrow
from informal channels to satisfy their credit demand. Thus households that have both bank loans and informal borrowing are also
regarded as suffering from credit constraint. We define this as the second measurement of credit constraint (called ‘Measure 2’) on the
basis of the first one.
To address the endogeneity concern, we firstly utilize the predetermined credit constraint to identify the causal effect of credit
constraint on the new entry of entrepreneurship. CHFS conducted the second round of survey across China in 2013, and 6846 out of
8438 households surveyed in 2011 were followed up. This makes it possible for us to study how the credit constraint households
suffered in 2011 affected their subsequent entrepreneurship decisions during 2011 and 2013. Secondly, we employ the housing de-
molition as a natural experiment. Households that experience demolition always get compensation from the government and may lead
to the relaxation of credit constraint. Thirdly, we use households' distance from downtown or the county center as an instrumental
variable for credit constraint. Households far away from the downtown or the county center have less knowledge of bank loans and are
more likely to be credit rationed by formal banks and thus more likely to be credit constrained.
The empirical results confirm our hypothesis that suffering from credit constraint significantly decreases the probability of
a household starting their own businesses in China. Credit constraint mainly depresses households' propensity of running enterprise
operation, while it has no significant effect on households' propensity of running small handicraft operation. Heterogeneity analysis
shows that the inhibiting effect of credit constraint on entrepreneurship is much larger for households in the urban and western areas.
The remained of this paper is organized as follows: Section 2 reviews the relevant literature. Section 3 introduces the data and
empirical approach. Section 4 presents the baseline empirical results. Section 5 discusses some concerns of relying on the estimates of
Section 4. Section 6 concludes.

2. Literature review

The interaction between entrepreneurship and credit constraint has been discussed for a long time. Knight (1921) argued that capital
markets cannot provide sufficient funds to entrepreneurs due to adverse selection and moral hazard. Entrepreneurs had to raise funds
and take the risks of failures by themselves. Evans and Jovanovic (1989) first suggested the importance of wealth for entrepreneurship
through an entrepreneurial choice model under credit constraint. Follow-up literature expanded their work from many perspectives and
our paper mainly relates to two strands of literature.
The first strand of literature discusses the impact of financial development on entrepreneurship. The well-developed financial market
can make sure that would-be entrepreneurs have good access to the credit market and can obtain the required credit. However, the
underdeveloped financial market is widespread around the world, which attracted lots of studies on the financial frictions that may lead
to financial constraint and thus depress entrepreneurship activities. By using household survey in Thailand, Paulson and Townsend
(2004) demonstrated that credit constraint is an important factor influencing household entrepreneurship. They pointed out that
households with higher capital accumulation are more likely to start their own businesses or reinvest in the existing industries because
such families face fewer credit constraint. They also found that the propensity of households who were commercial bank customers to
run their own businesses is 4 percentage points higher than those who were not commercial bank customers. Guiso, Sapienza, and
Zingales (2009) studied the impact of financial depth on entrepreneurship with Italy data and found that the entrance rate of new firms
is much higher in places where access to bank loans was better. Kerr and Nanda (2011) argued that deregulation of banking branches in
U.S. significantly decrease the competition between financial intermediaries and thus relaxed the financial constraint of potential en-
trepreneurs. Dutta and Sobel (2018) discussed the intermediate role of financial development played on the impact of human capital on
entrepreneurship.
The second strand of literature links household or individual wealth to the propensity of entrepreneurship. Hurst and Lusardi (2004)
showed that the positive correlation between wealth and propensity to start their own businesses was only true for the extremely
wealthy households and that borrowing constraint might not impact the entrepreneurship activities of less wealthy households. Fairlie

247
D. Cai et al. International Review of Economics and Finance 58 (2018) 246–258

and Krashinsky (2012) found that the probability of entry into self-employment increased with wealth for both currently unemployed
and employed. In recent years, lots of literature was interested in the importance of housing wealth for entrepreneurship. Schmalz,
Sraer, and Thesmar (2017) argued that housing appreciation as a positive shock to the collateral value significantly increased the
probability of becoming an entrepreneur. Wang (2012) suggested that the housing reform of the 1990s in China mitigated the credit
constraint of household in China and thus encouraged entrepreneurship. Li and Wu (2014) found that surging housing prices discourage
entrepreneurship in general though the appreciation of housing value had a positive wealth effect for potential entrepreneurs.
In this paper, we plan to measure the financial constraint of a household directly and estimate its impact on the propensity of
households to enter self-employment. One difficult issue with determining the causal effect of financial constraint on entrepreneurship is
the endogeneity problem, and lots of econometric methods are used in literature to separate the effect of financial constraint. Holtz-
Eakin et al. (1993) employed heritage wealth as the instrumental variable for the financial constraint. Hurst and Lusardi (2004)
employed heritage wealth and local housing price as instrumental variables to separate the effect of wealth on entrepreneurship.
Different-in-difference approach is also popular in identifying the effect of financial constraint on entrepreneurship. Using the French
Loan Guarantee Program from 1988 to 1999 as a quasi-experiment, Lelarge, Sraer, and Thesmar (2010) found that the program sig-
nificantly increased the scale of entrepreneurial applications and promoted the development of new enterprises though it had no effect
on the number of new enterprises. Buera, Jaef, and Shin (2015) and Greenstone, Mas, and Nguyen (2014) argued that great recessions or
financial crises led to sudden tightening collateral constraint and would deter entrepreneurship activities. Herkenhoff, Phillips, and
Cohen-Cole (2016) suggested that individual with bankrupt flag removal had better access to consumer credit and thus were more likely
to enter self-employment.

3. Data description and model specification

3.1. Data description

The data used in this paper is the China Household Finance Survey (CHFS) data from 2011 collected by the Survey and Research
Center for China Household Finance, Southwestern University of Finance and Economics. The CHFS project is a biannual survey and it
collects detailed information on Chinese households, including basic demographic characteristics, financial and nonfinancial assets,
liabilities and assets, household expenditure and income, social security and insurance, etc. To ensure the randomness and repre-
sentativeness of the dataset, CHFS adopts a stratified, three-stage, and Probability Proportionate to Size (PPS) sampling method. The
first round survey was conducted in 2011 on 8438 households distributed across 25 provinces (excluding Tibet, Xinjiang, Inner
Mongolia, Hong Kong, and Macao), 80 counties, and 320 villages.1 The second round survey conducted in 2013 covered 29 provinces,
262 counties (districts and county-level city), and 1048 communities (villages), and collected more than 28000 households' detailed
information. More importantly, 6846 households surveyed in 2011 were followed up in 2013, which makes it possible for us to identify
the causal effect of predetermined credit constraint on subsequent entrepreneurship decisions of households.

3.2. Variable definitions

We define a household to be self-employed if they engage in business operations (including small handicraft operations and en-
terprise operations). In consideration of the great difference in initial capital investment and credit demand of different types of business
operations, we classify household entrepreneurship into two types according to the business income. We classify entrepreneurship as
small handicraft entrepreneurship if the total business income through the year is less than 50000 RMB; otherwise, it is classified as
enterprise operation. It is worth noting that we focus on the non-agricultural entrepreneurship of households, and peasant households'
agriculture production and management projects (including agriculture, forestry, animal husbandry and fishing) are not classified as
entrepreneurship since peasant householders themselves are a kind of self-employed groups with agricultural production and man-
agement as the mainstay. We only regard peasant households who engaged in the industrial and commercial production and man-
agement as entrepreneurship.
When utilizing the followed-up sample to study the impact of predetermined credit constraint on the entrance of entrepreneurship,
we define a household to be new entrepreneurs if a household did not run business when surveyed in 2011 and was already running
business when surveyed in 2013.
The biggest difficulty of this paper is how to measure the credit constraint of households. CHFS designed an extensive list of
questions aimed at measuring and differentiating among different levels of credit demand and credit constraint. Several modules on
different types of households' liabilities including business liability, housing liability, automobile liability, financial liability, education
liability, credit card liability, and other liabilities are designed respectively. These questions make it possible for us to measure the credit
constraint of households directly. The sets of questions in each module are a little bit different. The modules to measure and evaluate
credit demand and credit constraint due to business operation, housing, and automobile purchasing contain three important parts. Each
module firstly askes households whether they have bank loans and households that have no bank loans then answer the question of why
they have no bank loans. Lastly, each module asks households whether they have informal borrowing from friends, relatives, or other
institutions. The modules to measure and evaluate households' credit demand and credit constraint due to financial investment,

1
The volume of samples may vary slightly during the regressions. This is mainly due to the fact that some variable values may be missing as the variables used by the
different regressions are different.

248
D. Cai et al. International Review of Economics and Finance 58 (2018) 246–258

education, and others only ask whether a household has bank loans. And for the module to measure credit card constraint, it first asks
whether a household has any credit card, and households have no credit card are further asked why they have no credit cards.
Given the difficulty of measuring credit constraint with household survey data, we try to measure credit constraint as accurate as
possible according to the information mentioned above. We define two measurements of credit constraint. For the first measurement of
credit constraint (Measure 1), households have no bank loans for the purpose of housing and automobile purchasing and their reasons
for having no bank loans are “Needed but did not apply” or “Applied but were rejected” are regarded as suffering from housing credit
constraint and automobile credit constraint. Households have no credit cards and their reasons for having no credit card are “do not
know”, “inability to pay”, or “Applied but were rejected” are regarded as suffering from credit card constraint. Households that suffer
from at least one type of the three credit constraint defined above are classified as being credit constrained. One potential problem of
relying on the first measurement of credit constraint is that households might be credit-rationed by formal banks though they obtained
loans from banks. As a result, the first measurement of credit constraint might under-estimate households' credit constraint. Considering
of this problem, we define the second measurement of credit constraint (Measure 2) on the basis of the first one. CHFS does not ask the
question about whether bank loans households obtained are enough to meet their credit demands among each module. As a supplement,
we believe that households that have bank loans but did not get enough credit might borrow from informal channels to satisfy their
credit demand. Thus households that have both bank loans and informal borrowing are also regarded as suffering from credit constraint.

3.3. Model specification

Our empirical strategy is to identify the causal effect of credit constraint on household entrepreneurship decisions. Literature on such
issues usually uses the Probit model or Logit model in empirical analysis. In addition, LPM models are also used frequently. Moreover, in
the case of large samples, the regression results of the LPM model are consistent with the Probit model and Logit model. For the
simplicity of model transformation, this paper mainly uses the following LPM model for analysis2:

Enterpreij ¼ α0 þ α1 Creditijs þ Xi β þ τj þ uij (1)

where, Enterpreij is a dummy variable indicating whether household i starts its own business in region j. CreditCijs is the dummy
variables indicating whether household i is subject to credit constraint, where s has three options: mortgage constraint, automobile loan
constraint, and credit card constraint. Xi is the control variables including characteristics of household head and household. τj is the
regional dummy variable used to eliminate the region fixed effect. At the same time, to eliminate the heteroscedasticity caused by binary
dependent variables, the LPM model requires that all regression results adopt the White robust standard error in estimation.
Our empirical specification recognizes that there are many determinants of entrepreneurship decisions, and we include a wide set of
variables that are available in the CHFS data as control variables. We consider demographic characteristics of the household head such
as age, education, marital status, gender, and risk attitude. We include the logarithm of transfer income and number of siblings to control
for the effect of social networks on household entrepreneurship decisions. The number of labor populations and family size are
employed to control for the effect of demography. Wealth is important for entrepreneurship since starting businesses requires initial
investment; thus, we add the logarithm of non-business net worth and three dummy variables for housing ownership, automobile
ownership, and credit card ownership to account for the wealth factors. Additionally, we add regional dummies and a dummy for the
rural area to account for the fact that entrepreneurship activities differentiate across different regions and areas. Definition and summary
statistics of these will be introduced in detail below.

3.4. Summary statistics

Considering the possible impact of schooling and retirement on entrepreneurship decisions, we restrict the estimation sample among
households with household headers aged between 18 and 60. And we also drop the observations with non-business assets below 1st or
above 99th percentiles of non-business assets. Observations with main control variables being missing values are also dropped, and 6423
households run into our final estimation. Table 1 describes the definition and summary statistics of main variables.
Table 1 shows that 15.4% of households are engaged in business operation. According to our definition of small handicraft entre-
preneurship and enterprise entrepreneurship based on a 50,000 RMB threshold, we can find that 8.83% of households are engaged in
small handicraft operation, and only 3.83% of households are engaged in enterprise operation. While the level of credit constraint is
high in China, about 27% of households in our sample are subject to certain types of credit constraint.
For the characteristics of household head, the average schooling year is 9.88, which roughly equivalent to the high-school education
level in China. The risk appetite is determined by the respondents' investment risk attitude. A respondent is regarded as a risk-lover if he
is willing to invest in high risk and high return projects or slightly high risk and slightly high return projects. In our sample, 15.7% of
households prefer to take risks. In estimations, the logarithm of household total asset excluding business asset (i.e., non-business asset) is
included to account for the fact that wealthier households are more likely to run businesses. We also add the dummy variables of housing
ownership, automobile ownership, and credit card ownership to further control for their impact on entrepreneurship.

2
The Probit regression results are available from the authors upon request.

249
D. Cai et al. International Review of Economics and Finance 58 (2018) 246–258

Table 1
Variable definition and summary statistics.

Variable The Meaning of Variable Mean Standard deviation Min Max

Entrepreneurship Household Engaged in business operation ¼ 1, Otherwise ¼ 0 0.1541 0.3611 0 1

Small handicraft entrepreneurship Household Engaged in small handicraft operation ¼ 1, Otherwise ¼ 0 0.0883 0.2837 0 1
Enterprise entrepreneurship Household Engaged in enterprise operation ¼ 1, Otherwise ¼ 0 0.0383 0.1919 0 1
Credit constraint (Measure 1) Subject to credit constraint ¼ 1, Otherwise ¼ 0 0.2173 0.4125 0 1
Credit constraint (Measure 2) Subject to credit constraint ¼ 1, Otherwise ¼ 0 0.2698 0.4439 0 1
Age Age of Household head 44.1865 9.7878 18 60
Schooling years Schooling year of household head 9.8828 3.7793 0 19
Married Married ¼ 1, Other ¼ 0 0.8921 0.3103 0 1
Male Male ¼ 1, Other ¼ 0 0.7263 0.4459 0 1
Risk lover Risk Appetite ¼ 1, Other ¼ 0 0.1569 0.3638 0 1
Housing ownership Housing Ownership ¼ 1, Other ¼ 0 0.9078 0.2893 0 1
Have more than 2 houses Household owns more than 2 houses ¼ 1, Other ¼ 0 0.1521 0.3592 0 1
Automobile ownership Car Ownership ¼ 1, Other ¼ 0 0.1727 0.3780 0 1
Credit card ownership Credit Card Ownership ¼ 1, Other ¼ 0 0.0662 0.2486 0 1
Log(transfer income) Log(1 þ transfer income) 5.2159 4.0758 0 15.8950
Number of siblings Number of siblings 2.8941 2.0870 0 16
Number of labor populations Number of families aged between 16 and 64 2.3894 1.0986 0 8
Family size Family size 3.6015 1.4283 1 18
Log(Non-business asset) Log(total asset minus business asset) 12.2221 1.6710 0 20.7280
Rural Rural area ¼ 1, otherwise 0 0.3649 0.4815 0 1

Note: We classify small handicraft entrepreneurship and enterprise entrepreneurship according to the business income, and there is some missing values with this
variable. Thus, the total mean of Small handicraft entrepreneurship and Enterprise entrepreneurship does not equal to the mean of Entrepreneurship.

4. Empirical results

4.1. Reasonability of credit constraint measurement

In this paper, we analyze the impact of other types of credit constraint (such as housing, automobiles, credit cards, etc.) on household
entrepreneurship. In the CHFS questionnaire, only households running businesses were asked about their liabilities for the purpose of
running businesses. Due to the limitations of the data, the definition of credit constraint in this paper might be incomplete. However, it is
reasonable to analyze credit constraint of households from other aspects (such as housing, automobiles, credit cards, etc.). First of all,
housing, automobile, and credit card loans are reasonable needs of residents, and we use them to identify credit constraint and dis-
tinguish credit constraint caused by unreasonable demands from credit constraint caused by a poor credit history, where the latter are
real credit constraint. Second, credit screening is a major operation carried out by the majority of China's commercial banks. In general,
basic operations of the bank credit screening process include acceptance, investigation, evaluation, approval, issuance and post-loan
management. Households are deemed to be under credit constraint if a bank terminates the process at any stage before issuing the
loan. According to the credit manual of state-owned commercial banks, the principles of different credit products adopted by the banks
in the process of acceptance, investigation, evaluation, and approval are roughly the same. All the guiding principles are derived from
the “General Principles of Loans” promulgated by the People's Bank of China. Especially for individual customers, when a bank approves
applications for different types of loans, many of the same factors are used as references by the bank. For example, the borrower must
have a legal identity, stable employment and economic income, or assets that are easily liquidated and sufficient to repay the principal
and interest of the loan on schedule. The borrower should also provide some relevant proofs of good credit records. Therefore, there is
a reason to believe that if households are subject to the bank's credit constraint when they purchase real estate or a car or apply for
a credit card, then it is also likely they are subject to similar credit constraint when they apply for a bank loan for the purpose of the
business operation.
In Table 2, we certify the reasonability of employing other types of credit constraint as a proxy for entrepreneurial credit constraint.
The sample used in Table 2 is households who have already run businesses. We run a regression of entrepreneurial credit constraint on
three other types of credit constraint with linear probability model (LPM) and the model is set up as below:

Entre Contraintij ¼ ∝0 þ β*Other CreditConstraintij þ Xi γ þ τj þ uij (2)

where, Entre Contraintij is a dummy variable indicating whether household i in region j is subject to entrepreneurial credit constraint.
Other CreditConstraintij is the dummy variables indicating whether household is subject to other types of credit constraint, including
housing credit constraint, automobile credit constraint, and credit card constraint. Xi are the control variables including characteristics
of household head and household characteristics. τj is the regional dummy variable used to eliminate the region fixed effect. The LPM
model requires that all regression results adopt the White robust standard error in estimation to eliminate the heteroscedasticity caused
by binary dependent variables.
Empirical results in Table 2 confirm that housing credit constraint, automobile credit constraint and credit card constraint are indeed
highly correlated with entrepreneurial credit constraint. Column (1) only includes the housing credit constraint, automobile credit
constraint, and credit card constraint. We find that the coefficient of these three types of credit constraint are all significantly positive at
the 1% significance level, indicating that households that suffering from housing credit constraint, automobile credit constraint, or

250
D. Cai et al. International Review of Economics and Finance 58 (2018) 246–258

Table 2
Correlation of credit constraint.

(1) (2) (3)

Entrepreneurial credit constraint

Housing credit constraint 0.2738*** 0.2625*** 0.2698***

(0.0326) (0.0334) (0.0339)
Automobile credit constraint 0.3738*** 0.3846*** 0.3716***
(0.0701) (0.0697) (0.0728)
Credit card constraint 0.1095*** 0.0782* 0.0791*
(0.0411) (0.0420) (0.0422)
Age of household head 0.0070 0.0080
(0.0101) (0.0102)
Age sq. of household head 0.0001 0.0001
(0.0001) (0.0001)
Schooling year of household head 0.0039 0.0032
(0.0039) (0.0040)
Household head is married 0.0092 0.0092
(0.0428) (0.0429)
Gender of household head 0.0272 0.0269
(0.0261) (0.0263)
Risk appetite 0.0565** 0.0546*
(0.0280) (0.0280)
Housing ownership 0.0278
(0.0470)
Own more than 2 houses 0.0061
(0.0280)
Automobile ownership 0.0208
(0.0274)
Credit card ownership 0.0459
(0.0372)
Log(transfer income) 0.0052* 0.0050*
(0.0027) (0.0028)
Number of brothers and sisters 0.0077 0.0075
(0.0058) (0.0058)
Number of labor pop 0.0068 0.0056
(0.0155) (0.0155)
Family size 0.0064 0.0074
(0.0114) (0.0114)
Log(Non-business asset) 0.0277*** 0.0280***
(0.0079) (0.0103)
Rural 0.0621** 0.0644**
(0.0265) (0.0267)
Middle area 0.0376 0.0421
(0.0264) (0.0267)
Western area 0.0665** 0.0630**
(0.0315) (0.0319)
Constant 0.1068*** 0.4846** 0.5224**
(0.0123) (0.2233) (0.2333)
Observations 997 990 990
R-squared 0.1124 0.1459 0.1484

Note: Robust standard errors are reported in brackets; ***, **, * indicate significance levels of 1%, 5%, and 10% respectively.

credit card constraint are more likely to be experiencing entrepreneurial credit constraint. In Column (2), we further include charac-
teristics of the household head and demographic characteristics of household, and the results are similar to Column (1). Results show
that the three types of credit constraint are still significantly positive at the 1% significance level. In order to eliminate the impact of
households' ownership of housing, automobile, and credit card on the regression results, we further control for the ownership of
housing, automobile and credit card in Column (3), and the regression result is still similar to Column (1) and Column (2).

4.2. Credit constraint and entrepreneurship

In Table 3 we report regression results for the impact of credit constraint on the propensity of a household engaging in business
operation. In Columns (1) to (3), households that subject to any type of credit constraint (including housing credit constraint, auto-
mobile credit constraint, and credit card constraint) are regarded as subject to credit constraint. Baseline results in Column (1) show that
households subject to credit constraint are less likely to enroll in business operation. Suffering from credit constraint decreases the
probability of a household to enroll in business operation by about 3.36 percentage points. In Columns (2) and (3), we decompose
entrepreneurship into small handicraft entrepreneurship and enterprise entrepreneurship and then estimate the impact of credit con-
straint on each type of entrepreneurship, respectively. Results show that suffering from credit constraint significantly decreases the
propensity of a household to engage in enterprise entrepreneurship, while it has no significant impact on the small handicraft

251
D. Cai et al. International Review of Economics and Finance 58 (2018) 246–258

Table 3
The impact of credit constraint on household entrepreneurship: OLS results.

(1) (2) (3)

Entrepreneurship Enterprise Small handicraft

Credit Constraint (Measure 1) 0.0336*** 0.0242*** 0.0041

(0.0104) (0.0044) (0.0089)
Age of household head 0.0032 0.0021 0.0041
(0.0037) (0.0023) (0.0029)
Age sq. of household head 0.0001* 0.0000 0.0001*
(0.0000) (0.0000) (0.0000)
Schooling year of household head 0.0080*** 0.0026*** 0.0034***
(0.0013) (0.0007) (0.0011)
Household head is married 0.0084 0.0156* 0.0004
(0.0147) (0.0083) (0.0115)
Gender of household head 0.0238** 0.0110** 0.0163**
(0.0100) (0.0054) (0.0079)
Risk appetite 0.0320** 0.0227*** 0.0050
(0.0135) (0.0084) (0.0103)
Housing ownership 0.0595*** 0.0246** 0.0173
(0.0193) (0.0104) (0.0147)
Own more than 2 houses 0.0546*** 0.0289*** 0.0282**
(0.0149) (0.0092) (0.0120)
Car ownership 0.1654*** 0.0981*** 0.0500***
(0.0158) (0.0106) (0.0123)
Credit card ownership 0.0464** 0.0406*** 0.0105
(0.0206) (0.0146) (0.0147)
Log(transfer income) 0.0035*** 0.0001 0.0020**
(0.0012) (0.0007) (0.0009)
Number of brothers and sisters 0.0051** 0.0012 0.0033*
(0.0023) (0.0012) (0.0018)
Number of labor pop 0.0123** 0.0046 0.0089*
(0.0057) (0.0029) (0.0046)
Family size 0.0106** 0.0003 0.0058
(0.0045) (0.0022) (0.0036)
Log(Non-business asset) 0.0180*** 0.0109*** 0.0033
(0.0038) (0.0018) (0.0029)
Rural 0.0567*** 0.0005 0.0401***
(0.0115) (0.0056) (0.0097)
Middle region 0.0423*** 0.0031 0.0363***
(0.0112) (0.0056) (0.0094)
Western region 0.0197* 0.0064 0.0198**
(0.0115) (0.0053) (0.0095)
Constant 0.0276 0.0019 0.0229
(0.0838) (0.0505) (0.0641)
Observations 6423 6423 6423
R-squared 0.0761 0.0870 0.0190

Note: Robust standard errors are reported in brackets; ***, **, * indicate significance levels of 1%, 5%, and 10% respectively.

entrepreneurship. One possible explanation is that small handicraft operation requires less initial investment compared to enterprise-
type entrepreneurship.
In Table 3 households suffering from any type of credit constraint including housing credit constraint, automobile credit constraint,
or credit card constraint are regarded as subject to credit constraint. As a robustness check, in Table 4 we examine the impact of each of
these three types of credit constraint on households' entrepreneurship decisions, respectively. Column (1) of Table 4 examines the
impact of housing credit constraint on entrepreneurship and households with home ownership are used in this estimation. Estimates in
Column (1) show that the impact of housing credit constraint on the propensity of a household to enroll in business operation is sig-
nificantly negative at the 5% significance level. Column (2) of Table 4 uses the sample of households with automobile ownership to
estimate the impact of automobile credit constraint on households' entrepreneurship decisions. We can find that the estimate of
automobile credit constraint is also significantly negative at the 5% significance level. In Column (3) of Table 4 we examine the impact of
credit card constraint on households' entrepreneurship decisions with full sample and results show that suffering from credit card
constraint also significantly decreases the propensity of a household to engage in business operation.

5. Discussions

5.1. Measurement of credit constraint

There is one potential problem in relying on the estimates of Table 3. Credit constraint utilized in these estimations may be measured
with substantial errors since households having bank loans might not get enough credit from banks and still suffer from credit constraint.

252
D. Cai et al. International Review of Economics and Finance 58 (2018) 246–258

Table 4
The impact of certain type of credit constraint on entrepreneurship: OLS results.

(1) (2) (3)

Housing credit constraint 0.0289**

(0.0113)
Automobile credit constraint 0.1053**
(0.0501)
Credit card constraint 0.0367***
(0.0123)
Age of household head 0.0035 0.0194 0.0036
(0.0042) (0.0124) (0.0038)
Age sq. of household head 0.0001* 0.0003** 0.0001**
(0.0000) (0.0001) (0.0000)
Schooling year of household head 0.0070*** 0.0287*** 0.0056***
(0.0014) (0.0047) (0.0014)
Household head is married 0.0027 0.0964 0.0085
(0.0164) (0.0593) (0.0150)
Gender of household head 0.0301*** 0.0380 0.0194*
(0.0108) (0.0299) (0.0102)
Risk appetite 0.0381*** 0.0036 0.0452***
(0.0143) (0.0336) (0.0136)
Log(transfer income) 0.0031** 0.0063* 0.0029**
(0.0013) (0.0034) (0.0012)
Number of brothers and sisters 0.0034 0.0104 0.0046**
(0.0024) (0.0075) (0.0023)
Number of labor pop 0.0113* 0.0095 0.0134**
(0.0059) (0.0198) (0.0058)
Family size 0.0111** 0.0149 0.0117***
(0.0047) (0.0153) (0.0045)
Log(Non-business asset) 0.0358*** 0.0454*** 0.0253***
(0.0042) (0.0143) (0.0031)
Own more than 2 houses 0.0610***
(0.0152)
Credit card ownership 0.0629***
(0.0213)
Middle region 0.0298** 0.1048*** 0.0236**
(0.0119) (0.0399) (0.0112)
Western region 0.0031 0.0038 0.0005
(0.0121) (0.0477) (0.0115)
Rural 0.0534*** 0.0213 0.0685***
(0.0122) (0.0391) (0.0115)
Constant 0.2850*** 0.1709 0.1519*
(0.1025) (0.2923) (0.0829)
Observations 5831 1109 6423
R-squared 0.0525 0.0765 0.0448

Note: Robust standard errors are reported in brackets; ***, **, * indicate significance levels of 1%, 5%, and 10% respectively.

Thus, LPM estimates may be biased upward. In terms of this problem, we could to some extent believe that households that did not get
enough credit from formal banks might resort to informal borrowing. Thus, based on the first measurement of credit constraint utilized
in Table 3, we further regard households that have both bank loans and informal borrowing as suffering from credit constraint, and then
replicate the estimations in Table 3 with the second measurement of credit constraint (Measure 2). Results are shown in Table 5. Es-
timates of credit constraint in Column (1) of Table 5 indicates that suffering from credit constraint decrease the probability of
a household to run business by 2.84 percentage points. As we mentioned before, the second measurement of credit constraint used in
Table 5 may over-estimate credit constraint and the estimates in Table 5 may be biased downward. Thus, combining the estimates in
Tables 3 and 5, we can conclude that suffering from credit constraint significantly decreasing the propensity of households running
businesses by 2.84–3.36 percentage points. Estimations in Column (2) and (3) of Table 5 also indicate that credit constraint only matters
for enterprise-type entrepreneurship and has no significant impact on small handicraft entrepreneurship.

5.2. Endogeneity problem

One big challenge in this paper is to identify the causal effect of credit constraint on entrepreneurship. First, the measurement of
credit constraint may be measured with error. On the other hand, households may suffer from credit constraint because they are engaged
in business operation. Thus, in this sector, we employ three methods to ease these concerns.
Firstly, we identify the impact of credit constraint on entrepreneurship by estimating the predetermined credit constraint suffered by
households on their subsequent entrepreneurship decisions, and the sample used in this estimation are households that did not enroll in
business operation in 2011 and are followed-up in the 2013 CHFS survey. Followed-up households that did not enroll in business
operation in 2011 but were already enrolled in business operation when surveyed in 2013 are classified as a new entrepreneur.

253
D. Cai et al. International Review of Economics and Finance 58 (2018) 246–258

Table 5
The impact of credit constraint on household entrepreneurship: Robustness check.

(1) (2) (3)

Entrepreneurship Enterprise Small handicraft

Credit Constraint (Measure 2) 0.0284*** 0.0189*** 0.0072

(0.0097) (0.0039) (0.0083)
Age of household head 0.0033 0.0020 0.0041
(0.0037) (0.0023) (0.0029)
Age sq. of household head 0.0001* 0.0000 0.0001*
(0.0000) (0.0000) (0.0000)
Schooling year of household head 0.0081*** 0.0026*** 0.0034***
(0.0014) (0.0007) (0.0011)
Household head is married 0.0093 0.0162* 0.0006
(0.0147) (0.0083) (0.0115)
Gender of household head 0.0239** 0.0111** 0.0164**
(0.0100) (0.0054) (0.0079)
Risk appetite 0.0324** 0.0230*** 0.0050
(0.0135) (0.0084) (0.0103)
Housing ownership 0.0624*** 0.0268*** 0.0173
(0.0192) (0.0104) (0.0146)
Own more than 2 houses 0.0539*** 0.0284*** 0.0280**
(0.0149) (0.0093) (0.0119)
Car ownership 0.1656*** 0.0982*** 0.0501***
(0.0158) (0.0106) (0.0123)
Credit card ownership 0.0443** 0.0393*** 0.0111
(0.0206) (0.0147) (0.0147)
Log(transfer income) 0.0036*** 0.0001 0.0020**
(0.0012) (0.0007) (0.0009)
Number of brothers and sisters 0.0051** 0.0012 0.0033*
(0.0023) (0.0012) (0.0018)
Number of labor pop 0.0121** 0.0044 0.0089*
(0.0057) (0.0029) (0.0046)
Family size 0.0106** 0.0003 0.0058
(0.0045) (0.0022) (0.0036)
Log(Non-business asset) 0.0171*** 0.0103*** 0.0031
(0.0038) (0.0018) (0.0030)
Rural 0.0587*** 0.0020 0.0404***
(0.0115) (0.0056) (0.0097)
Middle region 0.0417*** 0.0037 0.0366***
(0.0112) (0.0056) (0.0094)
Western region 0.0192* 0.0069 0.0200**
(0.0115) (0.0053) (0.0095)
Constant 0.0133 0.0081 0.0205
(0.0838) (0.0503) (0.0641)
Observations 6423 6423 6423
R-squared 0.0759 0.0862 0.0191

Note: Robust standard errors are reported in brackets; ***, **, * indicate significance levels of 1%, 5%, and 10% respectively.

Predetermined credit constraint used in this estimation is the first measurement of credit constraint we mentioned before. As we can see
from the result reported in Column (1) of Table 6, households suffering from credit constraint in 2011 are less likely to start their
businesses during 2011 and 2013. This proves that alleviating the credit constraint of would-be entrepreneurs is important.
Secondly, we further certify the importance of credit constraint for entrepreneurship through the impact of wealth shock on the
propensity of a household to run businesses (Blanchflower and Oswald, 1998; Hurst & Lusardi, 2004; Lelarge et al. 2010; Fairlie &
Krashinsky, 2012). We employ housing demolition before entrepreneurship as a positive shock to the relaxation of credit constraint. In
China, housing demolition has been carried out for a long time due to urban renewal and construction in the process of urbanization in
recent years. According to the regulations released by the Chinese government, demolition of housing on state-owned land shall be
compensated fairly to the owner of the housing. That is to say, households facing demolition receive compensation from the govern-
ment. Therefore, demolition of housing can be regarded as a positive wealth shock to the household wealth and will lead to the
relaxation of credit constraint on a household. We regard housing demolition as a natural experiment and households that going through
housing demolition before entrepreneurship experiencing a positive wealth shock, then we examine its heterogeneity influence on the
entrepreneurship decisions among households that are subject to credit constraint and households not subject to credit constraint.
We define a dummy variable demo to indicate whether a household experiencing housing demolition before entrepreneurship, and
add demo and its interaction with credit constraint into Model (2). The regression result is showed in Column (2) of Table 6. We can find
that the coefficient of demo is significantly positive; this indicates that households experiencing housing demolition before entrepre-
neurship are more likely to start their businesses. We are interested in the coefficient of the interaction term of demo and credit con-
straint. As we can see, the coefficient of this interaction term is significantly positive at the 10% significance level. That is to say, among
households subject to credit constraint, experiencing demolition before entrepreneurship will increase the propensity of a household to

254
D. Cai et al. International Review of Economics and Finance 58 (2018) 246–258

Table 6
Endogenous estimations.

(1) (2) (3) (4)

New entry Demolition First stage 2SLS

Credit Constraint(Measure 1) 0.0181** 0.0328*** 1.2057**

(0.0082) (0.0099) (0.5242)
Distance to downtown or county center 0.0005***
(0.0002)
demo 0.7632***
(0.0221)
Credit Constraint *demo 0.0558*
(0.0335)
Age of household head 0.0005 0.0029 0.0179*** 0.0245**
(0.0038) (0.0041) (0.0047) (0.0115)
Age sq. of household head 0.0000 0.0001 0.0002*** 0.0003**
(0.0000) (0.0000) (0.0001) (0.0001)
Schooling year of household head 0.0039*** 0.0077*** 0.0082*** 0.0183***
(0.0013) (0.0014) (0.0018) (0.0052)
Household head is married 0.0033 0.0141 0.0370* 0.0564*
(0.0145) (0.0160) (0.0195) (0.0338)
Gender of household head 0.0014 0.0308*** 0.0155 0.0422**
(0.0093) (0.0104) (0.0123) (0.0195)
Risk appetite 0.0111 0.0324** 0.0139 0.0110
(0.0118) (0.0141) (0.0148) (0.0230)
Housing ownership 0.0036 0.0737*** 0.0249
(0.0182) (0.0215) (0.0503)
Have more than 2 houses 0.0189* 0.0463*** 0.0094 0.0440*
(0.0113) (0.0147) (0.0153) (0.0240)
Car ownership 0.0524*** 0.1568*** 0.0006 0.1671***
(0.0135) (0.0161) (0.0149) (0.0237)
Credit card ownership 0.0138 0.0386* 0.0925*** 0.0656
(0.0177) (0.0214) (0.0172) (0.0559)
Ln(transfer income) 0.0012 0.0035*** 0.0006 0.0041**
(0.0010) (0.0012) (0.0014) (0.0021)
Number of siblings 0.0008 0.0037 0.0010 0.0066
(0.0019) (0.0023) (0.0028) (0.0041)
Number of labor pop 0.0017 0.0094 0.0176** 0.0361**
(0.0049) (0.0058) (0.0077) (0.0144)
Family size 0.0078* 0.0104** 0.0036 0.0135
(0.0040) (0.0045) (0.0058) (0.0085)
Log(Non-business asset) 0.0003 0.0195*** 0.0275*** 0.0154
(0.0035) (0.0041) (0.0048) (0.0159)
Rural area 0.0297*** 0.0505*** 0.0173 0.0616***
(0.0098) (0.0118) (0.0147) (0.0205)
Middle region 0.0067 0.0406*** 0.1047*** 0.1617***
(0.0099) (0.0115) (0.0140) (0.0573)
Western region 0.0147 0.0183 0.0783*** 0.1087**
(0.0100) (0.0119) (0.0152) (0.0461)
Constant 0.1349 0.0933 0.1570 0.1939
(0.0882) (0.0998) (0.1036) (0.1740)
Observations 5133 5831 6241 6241
R-squared 0.0169 0.1094 0.0632
1st-stage F statistics/t-statistics of instrument 26.02(2.67)
Endogeneity test 22.62(0.00)

Note: Robust standard errors are reported in brackets; ***, **, * indicate significance levels of 1%, 5%, and 10% respectively.

start businesses by 5.58 percentage points. This indicates that the wealth plays an important role in entrepreneurial behavior. When
a household faces credit constraint, their initial wealth level can compensate for the lack of credit capital, thereby promoting household
entrepreneurial activities.
Thirdly, we try to find an instrumental variable for credit constraint. Although we employ predetermined credit constraint and
housing demolition to verify the reliability of our estimation results, we may still suffer from an omitted variables problem. Thus, we
further use instrumental variable estimation to alleviate the endogeneity problem. The instrumental variable we choose for credit
constraint is the distance of a household from the downtown or county center. Households farther away from the downtown or county
center have worse access to financial services and thus are more likely to suffer from credit constraint. The first-stage regression is
reported in Column (3) of Table 6. The result shows that our instrument is statistically significant: the F-statistics are high and above the
value recommended to avoid the weak instruments problem (Stock & Yogo, 2005). The estimates in the second stage reported in Column
(4) of Table 6 show that the relationship between credit constraint and entrepreneurship remains negative and statistically significant in
the 2SLS regression. Moreover, the endogeneity test result reported in the bottom of Column (4) reject the null hypothesis that credit

255
D. Cai et al. International Review of Economics and Finance 58 (2018) 246–258

constraint is exogenous in Model (2). Thus, the OLS estimates reported in Table 3 differ significantly from the 2SLS estimates. And the
coefficient of credit constraint is larger than the OLS estimates. Overall, our results indicate that credit constraint is an important barrier
to entrepreneurship in China. Households subject to credit constraint are less likely to start their businesses.

5.3. Heterogeneity analysis

The extent to which household entrepreneurial behavior is affected by credit constraint varies from region to region and may differ
between different areas. Thus, we divide our sample into different subsamples to analyze the heterogeneous effect of credit constraint on
entrepreneurship decisions of households.
Firstly, we investigate the different effects of credit constraint on entrepreneurship across different regions. We divide our sample
into eastern, middle, and western subsamples according to the geographical distribution of China and estimations with Model (2) are
replicated respectively. Results are reported in Column (1) to Column (3) of Table 7. We can find that credit constraint has no statis-
tically significant impact on the entrepreneurship decisions of households in eastern areas, while it significantly decreases the pro-
pensity of households in middle and eastern areas to enroll in business operation. And the decreasing effects of credit constraint on the
propensity of households resided in the middle and eastern areas to enroll in business operation are 3.28 and 4.22 percentage points,
respectively. From the magnitude of the coefficient of credit constraint in these three sub-region estimations, it is the smallest in the
estimation of the eastern subsample, and biggest in the estimation of the western region subsample. One possible explanation is that
formal financial services are less developed in middle and western areas and thus the problem of credit constraint is more serious in
these areas.

Table 7
Heterogeneity analysis.

(1) (2) (3) (4) (5)

Eastern Middle Western Urban Rural

Credit Constraint(Measure 1) 0.0121 0.0328** 0.0422** 0.0359*** 0.0224*

(0.0164) (0.0164) (0.0171) (0.0138) (0.0129)
Age of household head 0.0033 0.0030 0.0117 0.0023 0.0059
(0.0052) (0.0075) (0.0079) (0.0047) (0.0063)
Age sq. of household head 0.0001 0.0000 0.0001 0.0001 0.0001
(0.0001) (0.0001) (0.0001) (0.0001) (0.0001)
Schooling year of household head 0.0126*** 0.0056** 0.0030 0.0130*** 0.0023
(0.0021) (0.0024) (0.0026) (0.0019) (0.0020)
Household head is married 0.0191 0.0020 0.0074 0.0102 0.0400
(0.0213) (0.0288) (0.0290) (0.0180) (0.0254)
Gender of household head 0.0340** 0.0299 0.0040 0.0265** 0.0139
(0.0139) (0.0203) (0.0201) (0.0121) (0.0171)
Risk appetite 0.0365* 0.0127 0.0443* 0.0392** 0.0263
(0.0205) (0.0254) (0.0256) (0.0170) (0.0216)
Housing ownership 0.0557** 0.0680 0.1052*** 0.0460** 0.1091**
(0.0255) (0.0439) (0.0392) (0.0229) (0.0452)
Own more than two houses 0.0742*** 0.0285 0.0116 0.0459** 0.0664**
(0.0201) (0.0290) (0.0347) (0.0179) (0.0270)
Car ownership 0.1534*** 0.2200*** 0.1439*** 0.1640*** 0.1848***
(0.0195) (0.0345) (0.0435) (0.0176) (0.0345)
Credit card ownership 0.0426* 0.0150 0.0969* 0.0436* 0.0434
(0.0254) (0.0463) (0.0525) (0.0226) (0.0469)
Ln(transfer income) 0.0011 0.0050** 0.0067*** 0.0057*** 0.0020
(0.0017) (0.0023) (0.0023) (0.0015) (0.0018)
Number of siblings 0.0099*** 0.0064 0.0061 0.0075** 0.0004
(0.0035) (0.0040) (0.0042) (0.0031) (0.0031)
Number of labor pop 0.0160* 0.0113 0.0094 0.0129 0.0075
(0.0090) (0.0102) (0.0105) (0.0088) (0.0073)
Family size 0.0102 0.0140* 0.0072 0.0212*** 0.0014
(0.0071) (0.0080) (0.0084) (0.0070) (0.0054)
Log(Non-business asset) 0.0195*** 0.0196** 0.0188*** 0.0129** 0.0258***
(0.0053) (0.0087) (0.0070) (0.0053) (0.0052)
Rural 0.0006 0.1291*** 0.0398*
(0.0178) (0.0227) (0.0208)
Middle region 0.0852*** 0.0323**
(0.0150) (0.0164)
Western region 0.0297* 0.0134
(0.0159) (0.0171)
Constant 0.0237 0.1348 0.1833 0.0755 0.2085
(0.1130) (0.1865) (0.1782) (0.1044) (0.1512)
Observations 3003 1916 1504 4079 2344
R-squared 0.0855 0.1055 0.0613 0.0853 0.0944

Note: Robust standard errors are reported in brackets; ***, **, * indicate significance levels of 1%, 5%, and 10% respectively.

256
D. Cai et al. International Review of Economics and Finance 58 (2018) 246–258

Then, we study the heterogeneous effect of credit constraint on entrepreneurship in urban and rural areas. Results shown in Columns
(4) and (5) of Table 7 show that suffering from credit constraint significantly decreases the propensity for starting businesses of both
urban households and rural households, and the effect is larger for urban households.

6. Conclusions

In this paper, we show that suffering from credit constraint is a significant deterrence to entrepreneurship in China by utilizing the
China Household Finance Survey data from 2011. The different measurements of credit constraint and different estimations we conduct
all show that households subject to credit constraint are less likely to start their own businesses, and this effect is larger for households in
western and urban areas. In the LPM regression model, we control for household housing ownership, car ownership, and other dif-
ferences in household characteristics. Results show that the decreasing effect of formal financial credit constraint on the probability of
household starting their own businesses ranges between 2.84 and 3.36 percentage points. In order to eliminate the regression bias
caused by endogenous variables, we tried three types of estimations to eliminate this concern. Firstly, we employ the followed-up
households in the 2013 CHFS survey to estimate the predetermined credit constraint on their subsequent entrepreneurship decisions
during 2011 and 2013, so as to eliminate the reverse causal effect of the entrepreneurial behavior on the credit constraint. We find that
households suffering from credit constraint in 2011 are less likely to start their own businesses during 2011 and 2013. Secondly, we
study the impact of wealth shock on the entrepreneurship decisions of households under credit constraint, and house demolition
subsidies is regarded as a positive shock to the relaxing of credit constraint. The results show that credit constrained households who
obtain demolition subsidies have a greater propensity of starting their businesses. Thirdly, we use the distance of a household from the
downtown or county center as the instrumental variable for credit constraint, considering that households farther away from downtown
or county center have worse access to financial services. The instrument estimation is consistent with the OLS estimates, with the
coefficient of credit constraint increasing a bit.
Our paper has important policy implications, and the government can encourage families to carry out entrepreneurial activities by
loosening credit constraint via providing better financial services. An increasing number of small and micro-enterprises take part in the
market, so that the market could become more energetic, and the economic development momentum could be more robust. China is at
the stage of rapid urbanization, and a large number of rural workers migrating to urban areas. As a result, how to absorb these labor
forces is a serious issue in China. Stabled employment is a basis for urbanization. Entrepreneurial activities strongly support urbani-
zation construction, and can also promote the development of the individual and private economy and speed up the adjustment of the
industrial structure. In order to make urbanization a new engine of economic growth, the government should create a favorable
environment for entrepreneurship, reduce the barriers, expand the field of entrepreneurship, and encourage urban residents and the
rural labor force to start entrepreneurial activities to promote employment. Besides, through financial system innovations, the gov-
ernment could help to solve entrepreneurial funding issues. In order to achieve this, we need to improve the requirements of banking
institutions, encourage institutions to explore innovative financial products and services, diversify venture capital channels, and
optimize the loan approval process to enhance the efficiency. In addition, the government should also strengthen entrepreneurship
education, and increase the proportion and quality of entrepreneurship through multiple channels.
Recently, a State Council executive meeting was organized to discuss the reform of the registration system of enterprise in China.
This meeting abolished the minimum capital requirement for several types of companies. Our findings undoubtedly provide empirical
evidence for the necessity of these policies. However, although the government eliminates the initial capital requirement for starting
a firm, small and medium-sized enterprises may still face the problems of credit constraint during daily operation. This will also damage
the living conditions of the most dynamic small and medium-sized enterprises, and we plan to expand our work in this direction in the
future.

Appendix A. Supplementary data

Supplementary data related to this article can be found at https://doi.org/10.1016/j.iref.2018.03.024.

References

Buera, F. J. (2009). A dynamic model of entrepreneurship with borrowing constraints: Theory and evidence. Annals of Finance, 5(3), 443–464.
Buera, F. J., Jaef, R. N. F., & Shin, Y. (2015). Anatomy of a credit crunch: From capital to labor markets. Review of Economic Dynamics, 18(1), 101–117.
Cagetti, M., & De Nardi, M. (2006). Entrepreneurship, frictions and wealth. Journal of Political Economy, 114(5), 835–870.
Dutta, N., & Sobel, R. S. (2018). Entrepreneurship and human capital: The role of financial development. International Review of Economics & Finance. http://dx.doi.org/
10.1016/j.iref.2018.01.020.
Evans, D. S., & Jovanovic, B. (1989). An estimated model of entrepreneurial choice under liquidity constraints. Journal of Political Economy, 97(4), 808–827.
Fairlie, R. W., & Krashinsky, H. A. (2012). Liquidity constraints, household wealth, and entrepreneurship revisited. Review of Income and Wealth, 58(2), 279–306.
Greenstone, M., Mas, A., & Nguyen, H. L. (2014). Do credit market shocks affect the real economy? Quasi-experimental evidence from the Great Recession and ‘normal’economic
times. National Bureau of Economic Research working paper No. w20704.
Guiso, L., Sapienza, P., & Zingales, L. (2009). Does local financial development matter?. In The banks and the Italian economy. Physica-Verlag HD.
Herkenhoff, K., Phillips, G., & Cohen-Cole, E. (2016). The impact of consumer credit access on employment, earnings and entrepreneurship. National Bureau of Economic
Research working paper No. w22846.
Hurst, E., & Lusardi, A. (2004). Liquidity constraints, household wealth, and entrepreneurship. Journal of Political Economy, 112(2), 319–347.
Kerr, W. R., & Nanda, R. (2011). Financing constraints and entrepreneurship. In D. Audretsch, O. Falck, & S. Heblich (Eds.), Handbook on Research on innovation and
entrepreneurship (pp. 88–103). Cheltenham: Edward Elgar Publishing.
Knight, F. H. (1921). Risk, uncertainty and profit. New York: Houghton, Mifflin.

257
D. Cai et al. International Review of Economics and Finance 58 (2018) 246–258

Lelarge, C., Sraer, D., & Thesmar, D. (2010). Entrepreneurship and credit constraint: Evidence from a French loan guarantee program. In International differences in
entrepreneurship. University of Chicago Press.
Li, L., & Wu, X. (2014). Housing price and entrepreneurship in China. Journal of Comparative Economics, 42(2), 436–449.
Paulson, A. L., & Townsend, R. (2004). Entrepreneurship and financial constraints in Thailand. Journal of Corporate Finance, 10(2), 229–262.
Schmalz, M. C., Sraer, D. A., & Thesmar, D. (2017). Housing collateral and entrepreneurship. The Journal of Finance, 72(1), 99–132.
Stock, J. H., & Yogo, M. (2005). Testing for weak instruments in linear IV regression. In D. W. Andrews, & J. H. Stock (Eds.), Identification and inference for econometric
Models: Essay in honor of Thomas Rothenberg (pp. 80–118). Cambridge University Press.
Wang, S. Y. (2012). Credit constraint, job mobility, and entrepreneurship: Evidence from a property reform in China. The Review of Economics and Statistics, 94(2),
532–551.

258