Upload
others
View
20
Download
0
Embed Size (px)
Citation preview
1
E-commerce Expands the Bandwidth of Entrepreneurship
Ruochen Dai, Peking University
Xiaobo Zhang, Peking University, and IFPRI
Abstract:
Social networks are cornerstones of personal trade in developing countries. However, it is
costly to build and maintain social networks. The high cost imposes an entry barrier for
people who lack social networks, such as those from outside local communities. E-
commerce offers a way for them to circumvent this problem. The timely payment of e-
commerce reduces their reliance on network-based trade credit that is prevalent in
traditional trades, easing their need for social networks. Using the administrative universe
firm registry dataset in China, we show that the spread of e-commerce has enabled more
people to become entrepreneurs.
Keywords: social network; e-commerce; entrepreneurship; China
2
1. Introduction
In the early stage of development, impersonal trade is less common than in more developed
stages due to the lack of formal institutions to enforce trade contracts. Instead, people rely
more on relational contracts to conduct trades (Grief, 1993; Fafchamps, 2004). Social
networks based on hometown, kinship, religion, caste, or tribes have been widely used in
personal trade in developing countries because there is inherently greater trust within
networks than between (Munshi, 2014). Although the trust-based networks lower
transaction costs among people within networks, it is costly to build and maintain them. In
addition, the club nature of networks prohibits those outside the network from entering the
business, resulting in a misallocation of entrepreneurial talent (Banerjee and Munshi, 2004).
Therefore, network-based personal trade is the second-best response to imperfect
institutions. In the long run, building sound institutions has been regarded as a necessary
condition for transforming personal trade to impersonal trade (North, 1990; Nunn 2007).
Without denying the importance of institutions, in this paper we intend to show that
technology, such as e-commerce, can also facilitate the transition from personal to
impersonal trade. In the past decade, e-commerce has taken off in many countries, especially
in China. China has overtaken the US as the largest e-commerce market in the world in
2013, with online sales accounting for more than 10% of retail sales, compared to 7% in the
US (Economist, 2013). Alibaba, the largest online platform in the world, has become the
biggest IPO in history and been listed on NASDAQ since 2014.
E-commerce has transformed personal trade to impersonal trade and devalued social
networks in at least two ways. First, it has considerably reduced the number of
intermediaries between producers and consumers. Traditional trade has to undergo multiple
intermediaries from producers to consumers, involving a significant amount of working
capital. Traders often use trade credit to ease working capital constraints. However, it takes
a long time for traders to establish trust before extending trade credit to each other. Second,
e-commerce has mitigated delayed payment problems that plague traditional trade,
especially in long-distance trade. For example, Alibaba has set up Ali-pay as an escrow
account to certify trade and process payments. Thanks to that, sellers can quickly receive
payment from Ali-pay after it certifies that a transaction is successful. In a word, with
3
disintermediation of e-commerce and the speedy payment, trade credit is much less needed
than before, thereby diluting the value of social networks underlying trade credit.
E-commerce also reduces the entry barriers of physical capital. E-commerce enables
retailers to open a virtual shop online at a considerably lower cost than a physical store. Due
to less need for social networks and lower entry barriers, e-commerce opens up space for
many potential entrepreneurs. It particularly benefits those from outside the production
centers or marketplaces, who would otherwise be impossible to enter the trading business.
Using the administrative universe firm registry dataset from the State Administration of
Industry and Commerce (SAIC) in China, our empirical analysis shows that the spread of
e-commerce has enabled more people, in particular outsiders, who were previously
constrained by financial and social capital, to become entrepreneurs and start a business. In
short, e-commerce has expanded the bandwidth of entrepreneurship.
The remainder of the paper is organized as follows. Section 2 briefly reviews the
history of e-commerce, in particular, Taobo.com, the largest consumer to consumer (C2C)
platform. By describing the inner workings of Baigou, one of the top “Taobao Towns” in
China, it also illustrates how e-commerce results in disintermediation and reduces the
dependence on social capital. Section 3 presents empirical results using an administrative
dataset from SAIC to demonstrate that e-commerce boosts entrepreneurship in China.
Section 4 further tests the heterogeneity effects of e-commerce and explores underlying
mechanisms. Section 5 concludes.
2. Background of China’s E-commerce
2.1. Brief Review of E-commerce in China
China’s brick and mortar businesses have been much less developed than those in the United
States. However, the rudimentary nature of China’s retails provides an opportunity for e-
commerce to flourish. In the past decade, e-commerce has rapidly exploded in China.
Alibaba, the largest e-commerce company in the world, has set up two major platforms for
e-commerce. Taobao.com, a Consumer to Consumer (C2C) platform, was established in
2003. Tmall.com was found in 2008 to promote Business to Consumer (B2C) trade. JD.com,
4
established in 2004, is another major e-commerce platform. Although it has grown more
rapidly than Alibaba in the past several years, its market share is still dwarfed by Alibaba.
According to a report of Ali Research, the total sales of Taobao and Tmall reached 500
billion dollars in 2015, accounting for 75% of total online sales in China.1
In addition, the spread of smartphones has made online shopping much easier than
before. The express delivery industry, one of the most competitive industries in China, has
penetrated to almost all corners of China. These factors have also contributed to the rapid
diffusion of e-commerce in China.
E-commerce is not evenly developed across space, showing large regional variation.
As shown by Alibaba’s e-commerce index in 2013 in Figure 1, online sales were mainly
concentrated along the coastal areas, overlapping with the manufacturing production centers.
2.2. A Case Study of “Taobao Town”: Baigou
To understand the inner workings of e-commerce, we first provide an anatomy of Baigou,
a top “Taobao Town” based on our field work and primary surveys. Alibaba published a list
of “Taobao Towns” in China in 2013 according to Taobao online sales. Baigou, one of the
largest suitcase and bags clusters in China, ranks among the top.
Baigou is located near Lake Baiyangdian, one of the largest lakes in Hebei Province,
and is about two hours away from Beijing. It used to be a major port by the lake and an
important market town. In the past three decades, it has evolved into one of the largest
suitcase/bag production and market centers in China. It houses more than 3,000 luggage
enterprises and ten thousand family workshops. More than half a million local people work
in this sector (Xing, 2009). By 2013, Baigou housed at least 3,000 online shops that
primarily sell bags and suitcases (Alibaba Research, 2013b).
1 The sales of Taobao and Tmall are available from “Internet Plus: New Engine for Chinese Economy” (page 23) see http://www.useit.com.cn/thread-8616-1-1.html. The number of total online sales comes from “Worldwide Retail Ecommerce Sales: Emarketer’s Updated Estimates and Forecast through 2019” (page 12) available on https://www.emarketer.com/Report/Worldwide-Retail-Ecommerce-Sales-eMarketers-Updated-Estimates-Forecast-Through-2019/2001716.
5
To better understand how e-commerce has shaped the real economy, we conducted
a field survey in Baigou in 2014. We focused on three major groups: traditional stores in
the suitcase/bag wholesale market, suppliers to online shops (online suppliers), and online
shops. Because of their physical presence, wholesale stores and suppliers to online shops
can be easily spotted and interviewed. We randomly sampled stores in the wholesale market
based on the store list provided by the wholesale market administration. The suppliers to
online shops are concentrated in two major streets. We first counted all of them and mapped
them out before drawing a random sample for an interview.
Most online shops operate in ordinary apartment buildings. A significant number of
them, in particular those under Taobo.com, have not registered with the government.
Therefore, it is extremely difficult to obtain a complete list of online shops. Even the major
platforms, such as Taobao.com and JD.com, have only the IP addresses of the online shops
but not their physical addresses. The suppliers have a rough idea about the locations of
online shops. Most of them locate nearby the apartment buildings where online shops are
concentrated. After talking to several suppliers to online shops, we mapped out the major
apartment buildings that host online shops. Then we randomly selected a few buildings,
knocked on all the doors in the chosen buildings and interviewed the owners of online shops.
2.3 Business Models of Online Shops and Wholesale Stores
In Baigou, the traditional wholesale stores and online shops have different business models.
As shown in Figure 2, many store owners in the wholesale market also operate a production
workshop or factory nearby. Their products have to pass several layers of intermediaries
before reaching consumers. By comparison, few online shop owners own a workshop or
factory. They operate in the following way: first visit the online suppliers, select a few
favorite samples displayed in the stores of online suppliers, and upload the photos of chosen
products onto their virtual shops. After receiving an order online, they go to the stores of
online suppliers, buy the merchandise by cash, package it, and have it shipped directly to
consumers. Because online shop owners do not need to operate a workshop/factory or rent
a physical store, their cost of doing business is much lower than a traditional store in the
wholesale market. Moreover, payments are guaranteed by Ali-pay or other third-party
payment systems, hence it is no longer necessary for online shop owners to spend time
6
building connections with other intermediaries to secure trade credit, as what traditional
wholesalers have done for years. The lower operational cost and more timely payment
reduce the reliance of online shops on both financial and social capital.
Table 1 further presents summary statistics comparing the mode of operations
between online shops and traditional wholesale stores. Several features are apparent from
the table. First, as shown in the first row, the percentage of people from outside Baigou
among online shop owners is much higher than among wholesalers, suggesting that online
retail businesses need less social networks than traditional wholesale businesses.
Second, it is less costly to run an online shop than a real store in the wholesale market.
As indicated in the second row, online shops maintain lower monthly inventories. Moreover,
it requires smaller amount of starting capital to open an online shop than a wholesale store.
It takes 15 times more starting capital to start a wholesale store than an online shop as
indicated in the third row.
Third, it involves much more social trust to run a traditional wholesale business than
an online shop. Thanks to the third-party payment system, 66% of transactions online do
not even need trade credit as shown in row 4. Row 5 reports how many months after the
first transaction are needed to obtain the first trade credit from customers or suppliers. On
average, it takes more than a year (14 months) before both parties in the traditional
wholesale sector develop enough trust to extend trade credit. The length shortens to three
months for an online shop. As a matter of fact, 67% of them receive full payment within
one month, as shown in the last row. By comparison, the share for traditional stores is much
lower with only 38% of them reporting to receive full payment within one month.
Apparently, online sales have a much quicker cash flow than traditional trade.
The lower requirement on financial resources and less reliance on social capital
provide a more level playing field for outsiders relative to insiders in running online shops.
In traditional wholesale trading business, it is hard for an outsider to penetrate in due to
greater starting capital and more dependence on social network. In traditional wholesale
business, insiders enjoy greater advantages than outsiders in starting and running a business.
The case study in Baigou demonstrates how online trade has diminished the role of
social network in the trade business, making it easier for outsiders to enter the business of
7
online sales. However, it is not clear whether the observation in Baigou holds true for China
as a whole. We are going to test this out in the next section.
3. Empirical Analysis of E-commerce on Entrepreneurship
We first examine the impact of e-commerce on two outcome variables related to
entrepreneurship, growth in the number of new firms and the total number of survival firms
over the five-year or ten-year period. Next, we show that the effect mainly runs through the
channel of lowering entry barriers (physical and social capital). Moreover, we find that the
effect of e-commerce on the emergence of new firms is particularly strong in industries with
high contract intensity. Finally, we estimate the general equilibrium effect of e-commerce.
3.1. Data Description
We employ a unique database from the State Administration of Industry and
Commerce (SAIC) in China to calculate the number of new firms and survival firms in each
county each year. According to the “China Corporation Law,” all companies must register
in the SAIC before opening a business. The administrative database of SAIC includes the
starting date, 4-digit industry code, location, ownership, registered capital, and information
about the legal representative and shareholders. When a firm goes bankrupt or its business
license is revoked, the existing date is recorded in the database. Unlike traditional sampling
firm surveys this database covers universal firms, including small and medium enterprises
(SME).
Using the starting and exit dates, we can compute the number of new firms and
survival firms at the county and industry level each year from 1978-2015. The database
includes the personal IDs of all the investors. The first six digits of their IDs reveal the
places where they obtained their IDs before 18 years old, mostly their hometowns. Based
on the amount of investment by each investor, we can identify the largest shareholder for
each firm. By comparing the hometown addresses of the largest shareholders with the
current firm registration addresses, we can infer whether the entrepreneurs come from local
areas or not. Using this method, we can compute the number of new firms and survival firms
owned by outsiders and insiders.
8
We also use the initial registered capital of investors as a measure of capital barrier
to entry. The initial capital is a proxy of entrepreneurs’ financial capital. Banks often use it
as a key indicator to determine credit lines when extending loans to a firm.
Ali research released China’s e-commerce index of 1,864 rural counties in 2013.
This index includes both the online sales index and the online purchase index. The online
sales index is a simple average of two separate measures, the e-commerce business density
and the average size of online shops in a county. The e-commerce business density takes
the average of the ratio of the number of business-to-consumers e-shops (B2C) to the total
number of registered corporations and the ratio of the number of consumers-to-consumers
e-shops (C2C) over the total number of self-employed entrepreneurs. The average size of
online shops is defined as the total online sales divided by the total number of online shops
at the county level. The online purchase index is defined as 0.5 * (number of online
consumers / population + total value of online purchase / number of consumers). We
transform them to z-scores in our empirical analyses.
We also use three factor intensity indexes — capital intensity index, skill intensity
index, and contract intensity index, to capture industry-specific characteristics. Like Nunn
(2007), we focus on manufacturing industries. The capital intensity and skill intensity
measured at the industrial level (SIC87) are obtained from the NBER-CES Manufacturing
Industry Database (Becker et al., 2013). Capital intensity takes the average of the ratio of
capital stock over the value added from 2005 to 2010. Skill intensity is the average of the
ratio of nonproduction workers’ wages over total wages in the period 2005-2010. We make
the concordance between SIC87 and Chinese Industry Classification 2011 (CIC2011) via
the International Standard Industry Classification (ISIC Rev3). Because ISIC Rev3 is
coarser than SIC87, a sector in ISIC Rev3 may correspond to multiple sectors in SIC87.
Thus when converting the index from SIC87 to ISIC Rev3, we compute the weighted
average of capital intensity using the total employment of each SIC87 industry as a weight
for a ISIC Rev 3 sector if it corresponds to more than one SIC87 sectors.2 We use a similar
procedure to obtain skill intensity. The contract intensity index is calculated based on the
1997 US Input-Output Table taken from Nunn (2007). The index is defined as the fraction
of inputs which are neither traded in commodity exchange markets, nor is public price
2 Following Antras and Chor (2013).
9
information available. Similarly, we matched IO1997 codes with CIC2011 codes through
ISIC Rev3.
After obtaining these factor intensity indexes at a 4-digit CIC2011 level, we can
analyze the county-sector cells. However, there is a problem using 4-digit CIC2011 codes
− some cells are empty because not all of the counties have new firms in every 4-digit
industry. As a compromise, we use 3-digit CIC2011 in all the analyses. We end up with 176
industries, comparable to the number of industries used in Nunn (2007), which covers 183
sectors.
The calculation of factor endowment follows Antweiler and Trefler (2002) and
Nunn (2007). Capital endowment is defined as the natural log of average capital per worker
using data from the China Economic Census 2008. The human capital variable is defined as
the natural log of the ratio of workers graduated from high school to those not based on the
China Population Census 2010. Table 2 reports summary statistics of all the variables.
3.2. Empirical Estimations
We follow the estimation strategy laid out in Ghani, Goswami, and Kerr (2016) to examine
the impact of e-commerce on the growth in number of new firms and survival firms at the
county level:
Δlog (Y𝑖𝑖𝑖𝑖) = 𝑎𝑎 ∗ log (Y𝑖𝑖0) + 𝛽𝛽 ∗ 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖 + 𝛾𝛾 ∗ 𝑋𝑋𝑖𝑖0 + 𝜀𝜀𝑖𝑖 (1)
The outcome variable Y𝑖𝑖𝑖𝑖 refers to the total number of newly registered firms or total
number of firms at year t in county i. 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖 is the standardized online sales index.
Coefficient 𝛽𝛽 measures the impact on outcome variables if the e-commerce index increases
by one standard deviation. Our sample includes rural counties where the online sales index
in 2013 is available and has at least one newly registered firm in 2014.
Following Ghani, Goswami, and Kerr (2016), the number of new firms (total firms)
in the initial year is included to capture the phenomenon of “mean reversion” across counties
(when the initial base is low, naturally the subsequent growth rate is high). The results are
robust no matter whether the initial value is included or not. Vector 𝑋𝑋𝑖𝑖 contains other factors
which may affect the outcome variables, including population size, the share of prime-age
population (15-64), the male-female sex ratio in the age group 0 and 14, the share of urban
10
population, illiteracy rate among population older than 15, and a clustering index.3 Except
for the clustering index, the control variables are largely the same with Ghani, Goswami,
and Kerr (2016). First, we estimate the effect of e-commerce on 5-year and 10-year growth
rates in the number of new firms in Table 3. The first three columns of Table 3 report the
estimation over the five-year period 2009-2014, while the last three present results for the
10-year period of 2004-2014. The first and fourth column are the most parsimonious
specification, including only the e-commerce index and the initial value of the number of
new firms in logarithmic form. The coefficient for the e-commerce in column (1) is 0.051,
which is highly significant. It means that an increase in one standard deviation in e-
commerce index brings about additional 5.1 log points of new firms in 2014, accounting for
4.22% (0.051/1.207) of the average short-run growth rate in the number of new firms from
2009 to 2014. In the second column, we add a set of control variables at the county level.
The coefficient for the e-commerce index increases to 0.068 and remains highly significant.
In the third column, we further control for province fixed effects. There are some advantages
and disadvantages of including province fixed effects. On the positive side, including
province fixed effects can help capture a province-specific growth trajectory. However,
since the growth regressions are essentially dynamic models with the lagged values of the
dependent variable on the right-hand side, the estimates may become biased when province
fixed effects are included. Nonetheless, it is worthwhile to check the robustness of the
results to the inclusion of province fixed effects. As shown in the third column, the
coefficient is 0.059, a result that is highly significant.
In the last three columns, we repeat the excise of the first three columns by using a
longer sample period from 2004 to 2014. In the fourth column (the most parsimonious
specification), the coefficient for the e-commerce variable is 0.072, larger than the five-year
effect as shown in column (1). In the fifth column (with county controls but without
province fixed effect), the coefficient for the e-commerce index drops to 0.048. After
controlling for province fixed effects in column (6), the coefficient for the e-commerce
index increases to 0.075, meaning that if a county’s e-commerce index is one standard
3 For the period 2009-2014, the demographic variables are from China Population Census 2010; the clustering measure is computed based on China Economic Census 2008 following Ruan and Zhang (2015). For the period 2004-2014, the demographic variables are obtained from China Population Census 2000 and the clustering measure is computed based on China Economic Census 2004.
11
deviation above the average, the total number of new firms in 2014 would be 7.5 log point
higher and the growth rate of number of new firms in the period 2004-2014 would be 3.98%
(0.075/1.884) higher than the average.
Second, we look at the effect of e-commerce on changes in the total number of firms.
Even though e-commerce brings about more new firms, it may also destruct some existing
firms. Therefore, it is useful to evaluate the impact on growth in the total net number of
survival firms. Table 4 presents the estimation results on growth in the total number of firms
over five-year and ten-year periods. The coefficient for the e-commerce index ranges from
0.023 to 0.050, still highly significant at the 1% level. In general, the coefficient in Table 4
is smaller than that in Table 3 probably because e-commerce has also yielded more firm
failures than before. In China, e-commerce did not start until 2003, the year when
Taobao.com was established. Therefore, the level of e-commerce development in 2004 was
extremely low. The actual increase in the e-commerce index is likely more than two standard
deviations from 2004 to 2014. This means that e-commerce development between 2004 and
2014 has contributed to 5% (2*0.05/1.989) of the actual increase in the total number of firms
in the period based on the estimate in column (6).
Although we have controlled for province fixed effects and some initial
characteristics at the county level, there is still a possibility that the e-commerce index
actually captures some omitted variables, which happen to be associated with extensive firm
growth. To remedy this concern, we run a falsification test in Table 5. Because e-commerce
was not developed until 2003, the growth rate of the number of newly registered firms (or
total number of firms) from 1997 to 2002 should have little to do with the e-commerce index
observed in 2013. If the e-commerce index captures some unobserved factors contributing
to extensive firm growth in a county, we should observe that the index also matters to
extensive firm growth from 1997 to 2002 prior to the e-commerce era. We repeat the
estimations of column (3) in Tables 3 by replacing sample period 2009-2014 with sample
period 1997 -2002 prior to the era of e-commerce. In columns (1) and (3), counties with zero
number of new firms (total survival firms) in 1997 are dropped from the sample. In columns
(2) and (4), we replace the zero values in 1997 with the value of the bottom 1 percentile
among the counties which have at least one firm. In doing so, the sample size is kept the
12
same as in the main specifications in Tables 3. None of the coefficient for e-commerce in
Table 5 is significantly positive. The results in Table 5 mute the concerns of omitted variable
bias.
4. Underlying Mechanisms
In this section, we further explore the underlying mechanisms behind the observed effect
of e-commerce on extensive firm growth.
4.1. Heterogeneous Effects on Outsiders and Insiders
Our hypothesis predicts that e-commerce benefits outsiders more rather than locals.
Outsiders are generally more likely to lack the necessary social networks which underpin
personal trade than local people (Banerjee and Munshi, 2004). Therefore, it is harder for
them to enter the traditional trading business. Thanks to the timely payments and the
impersonal trade embedded in e-commerce, the importance of social network fades,
providing outsiders with a level playing field to enter the business of online sales. Next we
test if the introduction of e-commerce particularly benefits those people who are from other
areas as predicted by our hypothesis.
Figure 4 plots the growth of new firms owned by local people (insiders) and non-
local people (outsiders) versus the standardized e-commerce index at the county level. It is
apparent that there is a positive association between extensive growth of firms owned by
outsiders and local e-commerce index. The relationship is much weaker for the extensive
growth of local firms. Since the figure does not control for any other covariates except for
the initial value, the pattern is just suggestive.
In Table 6, we repeat the third column in Table 3 for the subsamples of new firms
established by local people (insiders) and people from other areas (outsiders). As shown in
the first two columns in Table 6, the coefficient for the e-commerce index in both
regressions are statistically positive, but it is larger in the regression on outsiders (0.070)
than that on insiders (0.036). To check if the difference is statistically significant, in the
third column, we combine the subsamples and add a dummy for outsiders as well as its
interaction with the e-commerce index. The positively significant coefficient for the
13
interaction term implies that e-commerce has indeed provided an opportunity for people
outside the local area to start their businesses. In columns (4)-(6), we replace the dependent
variable with the total number of firms. The main results still hold. The impact of e-
commerce on extensive firm growth is larger for outsiders than for insiders.
4.2. The Impact of E-commerce on Growth in Private Firms and State-Owned
Enterprises (SOEs)
Unlike private firms, state-owned enterprises have easy access to state bank credit. They do
not need to cultivate personal connections in order to obtain bank credit. Because they are
less constrained by working capital, they are more likely to extend trade credit than receive
trade credit in industrial clusters (Long and Zhang, 2011). If the primary role of e-commerce
on the extensive firm growth is through the channels of lowering entry barriers and reducing
reliance on social networks, we would expect to see a much more muted effect on SOEs
than private firms.
To check this out, in Table 7, we run separate regressions for growth in number of
private firms and SOEs following the specification of columns (3) of Table 3-4 over the
period 2009-2014. We use the number of new firms as the dependent variable in columns
(1)-(3) and the total number of firms as the alternative dependent variable in columns (4)-
(6). As shown in columns (1) and (4) for private firms, the coefficients for the e-commerce
index are 0.059 and 0.038, both of which are statistically significant at the 1% level. By
contrast, they are insignificant in regressions on SOEs as revealed in columns (2) and (5).
In other words, e-commerce development matters little to the extensive growth of SOEs. To
further check if the coefficients between columns (1) and (2) and between columns (4) and
(5) are statistically different to each other, we further run two regressions using the
combined sample by adding a dummy variable for private firms and its interaction term with
the e-commerce measure for two sample periods, 2009-2014 and 2004-2014, respectively.
As indicated in columns (3) and (6), the interaction terms are highly positive and significant.
4.3. The Impact of E-commerce on Firm Starting Capital
14
Having shown that e-commerce promotes extensive firm growth, we next examine if e-
commerce lowers the threshold of starting a business, in particular for outsiders. We use the
initial registered capital as a measure of starting capital.
Table 8 examines the effect of e-commerce on the log of starting capital of newly
registered firms in 2014. The first two columns are regressions on the subsample of firms
established by outsiders and insiders. The control variables are the same as the third column
of Table 3. In addition, industry fixed effects are included. As shown in the first column,
the impact of e-commerce on the starting capital of firms owned by outsiders is negative
and insignificant. In column (2), the coefficient for e-commerce in regression on starting
capital for local firms is positive, yet indifferent from zero. In the third column, we regress
on the combined sample by adding a dummy for non-local firms and its interaction term
with the e-commerce index. The coefficient for the dummy variable “Outsiders” is positive
and significant, indicating that if there is no e-commerce, firms owned by outsiders have to
put 44% more registered capital than local firms. The rise of e-commerce has slightly offset
the disadvantage of non-local firms as shown by the negative coefficient for the interaction
term between the e-commerce index and a dummy for outsiders.
4.4. The Effect of E-commerce in Reducing Reliance on Formal Institutions
The formal legal system is instrumental in helping businesses to enforce contracts,
particularly in contract-intensive industries, such as long-distance trade. As shown in Nunn
(2007), in countries with more sound institutions, relationship-specific industries (high
contract intensity) perform better in export than those with weak institutions. However, it
takes a long time to develop formal institutions in developing countries. As shown in the
case study of Baigou, e-commerce cuts many layers of intermediaries and reduces the need
on trade credit, thereby greatly weakening the role of relationships in trade. In this sense, e-
commerce plays a similar role as formal institutions in facilitating transactions in
relationship-intensive industries, such as the wholesale and retail industry. To test this idea,
we replace the contract enforcement variable (legal system, a proxy for formal institutions)
used in Nunn (2007) with the e-commerce index and interact it with contract intensity at the
industry level. In essence, our specification below is similar to Nunn (2007):
Δlog (Y𝑖𝑖𝑖𝑖𝑖𝑖) = 𝛾𝛾 ∗ log (Y𝑖𝑖0) + 𝑎𝑎𝑖𝑖 + 𝑎𝑎𝑖𝑖 + 𝛽𝛽1𝑧𝑧𝑖𝑖𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖 + 𝛽𝛽2ℎ𝑖𝑖𝐻𝐻𝑖𝑖 + 𝛽𝛽3𝑘𝑘𝑖𝑖𝐾𝐾𝑖𝑖 + 𝜀𝜀𝑖𝑖𝑖𝑖 (2)
15
Where Δlog (Y𝑖𝑖𝑖𝑖𝑖𝑖) is the growth rate of number of new firms in county i and industry s from
2009 to 2014; 𝑧𝑧𝑖𝑖 is a measure of contract intensity in industry s; 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖 is the
standardized e-commerce sales index; ℎ𝑖𝑖 is the ratio of wages of skilled workers to total
wage bills, representing the skill intensity of production in industry s; 𝑘𝑘𝑖𝑖 is defined as the
ratio of fixed capital to valued added, measuring the capital intensity of production in
industry s; 𝐻𝐻𝑖𝑖 denotes county i’s human capital; 𝐾𝐾𝑖𝑖 is county i’s capital level, measured as
the natural log of average capital per worker; 𝑎𝑎𝑖𝑖 and 𝑎𝑎𝑖𝑖stand for county fixed effects and
industry fixed effects. Following Ghani, Goswami, and Kerr (2016), we include the number
of new firms of county i and industry s in the initial year, to capture the phenomenon of
“mean reversion” across counties.
Table 9 presents ordinary least square (OLS) estimations for equation (2). The first
column is the most parsimonious specification for it includes only the interaction of the
contract intensity variable at the industrial level and the e-commerce index at the county
level. Although there are 1,861 counties and 176 industries, not all industries are present in
all of the counties. The number of county-industry cells with positive values in 2014 is
38,917. To measure the growth rate of firms for as many cell as possible, for cells with
positive entries in 2014 but zero values in 2019, we replace the zero values in 2009 by the
value at the bottom 1 percentile of all the observations with positive entries in 2009. The
estimated coefficient for the e-commerce interaction term in column (1) is 0.086 and highly
significant.
In the second column, following Nunn (2007) we add two more interaction terms:
capital intensity∗physical capital; skill intensity∗human capital. The coefficient for the
interaction between contract intensity and e-commerce index drops slightly to 0.081 and
remains statistically significant. The interaction term between capital intensity and capital
endowment is not significant. The interaction term between skill intensity and the human
capital is 0.085, significant at the 1% level. The effect of e-commerce in fostering extensive
firm growth in relationship-intensive industries is the same order of magnitude as that of
human capital in promoting growth in skill intensive industries.
To examine if the effect is more pronounced for firms owned by outsiders than those
by local people, in columns (3) and (4), we repeat the excise in column (2) for two
16
subsamples, firms owned by outsiders and insiders. In the regression on the subsample of
outsiders (column 3), the coefficient for the interaction term between contract intensity and
e-commerce rises to 0.101, highly significant and greater than that in column (1) for the
whole sample. The coefficients for other two interaction terms are insignificant. In the
regression on the insider subsample in column (4), the coefficient for the e-commerce
interaction term drops to 0.061, much smaller than that in the outsider subsample. In the last
column, we combine the insider and outsider subsamples and include a triple interaction
term of contract intensity, e-commerce index, and a dummy for outsiders. The coefficient
for the new triple interaction term is 0.040, which is significant at the 1% level. This means
that e-commerce is more conducive to the growth of new firms established by outsiders,
particularly in relationship-intensive industries.
4.5. Effect of Online Purchase Index and Overall E-commerce Index
In the previous tables, the e-commerce index is measured by online sales index. E-
commerce not only affects production but also consumption. However, the places of online
sales and purchases are not necessarily the same. Although we have shown a positive effect
of online sales index on extensive firm growth at the county level, there is a possibility that
greater online purchases elsewhere may destruct local retail businesses. The net effect is far
from clear. We use two ways to check this possibility. First, in Table 10, we look at the
impact of online purchases and the combined e-commerce index (the average of online
purchase and online sales indexes) at the county level on the growth of number of new firms
and total number of survival firms between 2009 and 2014 following the same specification
of column (3) in Tables 3. As shown in columns (1) and (3), the coefficient for the online
purchase index is significantly positive. In other words, even in areas with active online
purchases, we have observed extensive firm growth.
In columns (2) and (4), we use the combined e-commerce index as an e-commerce
measure. The results still hold. An increase in the combined e-commerce index by one
standard deviation in a county would lift the growth rate of the number of new firms and
total firms between 2009 and 2014 by 11.2% and 6.1%, respectively.
4.6. Spillover Effect of E-commerce
17
All the previous regressions are at the county level. It is possible that a county with well-
developed e-commerce pulls entrepreneurs from nearby counties, draining their total
number of firms. It is not clear whether e-commerce increases the total number of firms or
not at a more aggregate level. To check this out, we next run regressions at more aggregate
levels, such as at the prefecture or province level.
In Table 11, we regress the growth of number of new firms at the prefectural (or
provincial) level, respectively, on the maximum value of the e-commerce index at the
county level within a prefecture (or province). The set of control variables is the same as in
column (3) in Table 3, except that they are at the prefecture or provincial level. The
coefficient for the e-commerce index in regression at the prefecture level (column 1) is 0.039,
significant at the 5% level. When running the regression at the province level, the coefficient
declines slightly to 0.031. All in all, the spatial spillover effect within a prefecture or
province, if any, is minimal.
In columns (3) and (4), we directly test if a county with the highest e-commerce
development index in a prefecture (or province) affects other counties within the same
prefecture (or province). Specifically, we drop counties with the highest e-commerce index
from the sample in the calculation of extensive firm growth at the prefectural level (or
provincial level). The coefficient for the maximum e-commerce index in a prefecture or
province is insignificant. This indicates that counties with the highest degree of e-commerce
development has not dried out entrepreneurship in other counties within the same prefecture
or province.
5. Conclusion
In history, the transition from personal trade to impersonal trade takes a long time. Sound
institutions are foundations for impersonal trade. However, it is a daunting task to build up
sound institutions in the first place. In this paper, we used China as an example to show how
e-commerce has offered a faster route for the transition to impersonal trade.
E-commerce has lowered the entry barriers to capital and weakened the reliance on
social networks in business. Therefore, it offers a new opportunity for potential entrepreneurs
who were previously constrained by a lack of financial and social capital. Using the universe
18
firm registry database, we have shown that e-commerce has particularly benefited non-local
entrepreneurs, taping the previously inhibited entrepreneurial talent.
The impact of e-commerce may differ across industries and regions. We will leave
the investigation on heterogeneity as a future research topic.
19
References
Alibaba Research (2013). The development of e-commerce in Chinese counties. Alibaba,
2013
Alibaba Research (2013). Simple report 2.0 of “Taobao Towns”. Alibaba, 2013
Alibaba Research (2015). Internet plus: New engine to Chinese economy, Alibaba, 2015
Antràs, P., and Chor, D. (2013). Organizing the global value chain. Econometrica, 81(6):
2127-2204.
Banerjee, A., and Munshi, K. (2004). How efficiently is capital allocated? Evidence from
the knitted garment industry in Tirupur. Review of Economic Studies, 71(1):19-42.
Becker, R., Gray, W., and Marvakov, J. (2013). NBER-CES Manufacturing Industry
Database: Technical Notes. NBER Working Paper, 5809.
Coase, R. H. (1937). The nature of the firm. Economica, 4(16): 386-405.
Economist (2013). The world’s greatest bazaar: Alibaba. Economist, March 23.
Fafchamps, M. (2004). Market institutions in sub-Saharan Africa: Theory and
evidence. MIT Press Books
Ghani, E., A. G. Goswami, and Kerr, W. R. (2016). Highway to Success: The Impact of
the Golden Quadrilateral Project for the Location and Performance of Indian Manufacturing.
The Economic Journal, 126(591): 317-357.
Goldmanis, M., Hortaçsu, A., Syverson, C., and Emre, Ö. (2010). E‐Commerce and the
Market Structure of Retail Industries. The Economic Journal, 120(545): 651-682.
Greif, A. (1993). Contract enforceability and economic institutions in early trade: The
Maghribi traders' coalition. The American economic review, 83(3), 525-548.
Lendle, A., Olarreaga, M., Schropp, S. and Vézina, P.-L. (2016), There Goes Gravity: eBay
and the Death of Distance. The Economic Journal, 126(591): 406–441.
Long, C. and Zhang, X. (2011). Cluster-based industrialization in China: Financing and
performance. Journal of International Economics, 84(1), 112-123.
20
Lucking-Reiley, D., and Spulber, D. F. (2001). Business-to-business electronic commerce.
Journal of Economic Perspectives, 15(1): 55-68.
Munshi, K. (2014). Community networks and the process of development. Journal of
Economic Perspectives, 28(4): 49-76.
Nunn, N. (2007). Relationship-specificity, incomplete contracts, and the pattern of
trade. The Quarterly Journal of Economics, 122(2): 569-600.
North, D. C. (1990). Institutions, Institutional Change and Economic Performance.
Cambridge university press.
Qian, Y. (1994). A Theory of Shortage in Socialist Economies Based on the" Soft Budget
Constraint". American Economic Review, 84(1): 145-156.
Ruan, J. and Zhang, X. (2009). Finance and Cluster‐Based Industrial Development in
China. Economic Development and Cultural Change, 58(1): 143-164.
Ruan, J. and Zhang, X. (2015). A Proximity-based Measure of Industrial Clustering. IFPRI
Discussion Paper No. 1468.
Williamson, O. E. (1979). Transaction-cost economics: the governance of contractual
relations. Journal of law and economics, 22(2): 233-261.
Xing, Y. (2009). Analysis on the Baigou Pattern [D], Hebei University
Zhang, X. and Zhu, W. (2015). The spatial patterns of E-commerce in China. Working Paper.
21
Table 1: The Characteristics of Online Shops and Traditional Wholesales
(1) (2) T-test
Online Traditional
Obs. Mean Obs. Mean (1)-(2)
Outsiders (%) 180 79% 107 42% 0.37***
Inventory (piece) 113 515.62 95 1,145.35 -629.73***
Start capital (10 thousand Yuan) 181 2.48 108 39.71 -37.23***
Fraction of businessman without trade credit 180 66% 108 0% 66%***
Length before obtaining trade credit (Month) 42 3.21 31 13.77 -10.57***
Receiving full payment in a month (%) 181 67% 108 38% 0.29***
Source: Authors’ survey in Baigou.
22
Table 2: Descriptive statistics
(1) (2) (3) VARIABLES N mean sd No. of new firms in 2014 (log) 1,861 5.549 1.138 No. of new firms in 2009 (log) 1,864 4.337 1.279 No. of new firms in 2002 (log) 1,781 3.215 1.404 No. of new firms in 1997 (log) 1,405 1.882 1.336 No. of survival firms in 2014 (log) 1,863 6.933 1.165 No. of survival firms in 2009 (log) 1,864 5.916 1.286 No. of survival firms in 2002 (log) 1,837 4.358 1.535 No. of survival firms in 1997 (log) 1,705 2.664 1.605 Log(new firms in 2014/new firms in 2009) 1,861 1.207 0.605 Log(new firms in 2014/new firms in 2004) 1,827 1.884 0.827 Log(new firms in 2002/new firms in 1997) 1,402 1.712 0.880 Log(survival firms in 2014/survival firms in 2009) 1,863 1.015 0.407 Log(survival firms in 2014/survival firms in 2004) 1,853 1.989 0.673 Log(survival firms in 2002/survival firms in 1997) 1,705 1.898 0.694 E-commerce (standardized) 1,864 0.001 1.004 Log(population) 1,864 12.73 0.811 Cluster index 1,864 -0.200 0.826 Share of urban population (%) 1,864 34.58 13.86 Illiteracy rate (%) 1,864 6.847 6.270 Share of prime age (15-64) (%) 1,864 72.52 4.535 Male/female sex ratio (0-14) 1,864 1.162 0.347 Log(initial capital) 878,887 6.614 1.743 Contract intensity 176 0.509 0.186 Capital intensity 176 0.788 0.374 Skill intensity 176 0.431 0.106
Notes: Number of firms are aggregated based on China firm registry database. Number of new firms refers to the number of newly established firms. Number of survival firms is defined as the total number of firms shown active status in the firm registration database by the end of a calendar year. Initial capital is defined as the self-reported registry capital at the time of establishment. E-commerce index refers to online sales index, obtained from Ali-research and we have transformed it into z-score. Cluster index is taken from Ruan and Zhang (2015) based on China Economic Census 2008. Total population, the share of the urban population, illiteracy rate, the share of the prime-age population, and male/female sex ratio are from aggregate data at the county level of China Population Census 2010. The contract intensity measure is obtained from Nunn (2007). The capital intensity and skill intensity are obtained from the NBER-CES Manufacturing Industry Database (Becker et al., 2013), following Nunn (2007). Capital intensity is defined as the ratio of capital stock to value added. Skill intensity is the ratio of wages of skilled workers to total wages.
23
Table 3: The effect of e-commerce on growth in number of new firms
(1) (2) (3) (4) (5) (6)
VARIABLES Log(no. of new firms in 2014/
no. of new firms in 2009) Log(no. of new firms in 2014/
no. of new firms in 2004) No. of firms in -0.231*** -0.319*** -0.450*** -0.367*** -0.545*** -0.687*** initial year (log) (0.013) (0.024) (0.023) (0.016) (0.024) (0.024) E-commerce (standardized) 0.051*** 0.068*** 0.059*** 0.072*** 0.048*** 0.075***
(0.011) (0.012) (0.012) (0.015) (0.015) (0.015) Log(population) 0.270*** 0.377*** 0.436*** 0.530***
(0.028) (0.027) (0.028) (0.029) Cluster index -0.051** 0.053** 0.110*** 0.141***
(0.025) (0.026) (0.030) (0.033) Share of urban 0.001 0.003*** 0.280** 0.473*** population (%) (0.001) (0.001) (0.125) (0.144) Illiteracy rate (%) 0.005 -0.010*** 0.006** -0.007***
(0.003) (0.003) (0.002) (0.002) Share of prime age -0.001 0.006 0.007* 0.010** (15-64) (%) (0.003) (0.005) (0.004) (0.005) Male/female sex ratio -0.000 0.004 0.003 0.001 (0-14) (0.010) (0.015) (0.002) (0.002) Province fixed effect No No Yes No No Yes
Observations 1,861 1,861 1,861 1,861 1,861 1,861 Adjusted R-squared 0.206 0.271 0.467 0.323 0.464 0.617 AIC 2984 2831 2221 3884 3456 2798
Notes: For variable definitions, please see Table 2. Robust standard errors are in parentheses. The symbols *, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively.
24
Table 4: The effect of e-commerce on growth in the number of total firms
(1) (2) (3) (4) (5) (6)
VARIABLES Log(total no. of firms in 2014/
total no. of firms in 2009) Log(total no. of firms in 2014/
total no. of firms in 2004) No. of firms in -0.152*** -0.204*** -0.230*** -0.314*** -0.458*** -0.528*** initial year (log) (0.010) (0.016) (0.020) (0.011) (0.018) (0.019) E-commerce 0.023*** 0.031*** 0.038*** 0.050*** 0.027*** 0.050*** (standardized) (0.008) (0.007) (0.007) (0.011) (0.010) (0.010) Log(population) 0.165*** 0.192*** 0.308*** 0.352***
(0.018) (0.019) (0.022) (0.023) Cluster index -0.020 -0.004 0.135*** 0.134***
(0.017) (0.017) (0.023) (0.024) Share of urban 0.001* 0.001 0.255*** 0.299*** population (%) (0.001) (0.001) (0.095) (0.106) Illiteracy rate (%) 0.006** -0.003 0.003* -0.002
(0.002) (0.002) (0.002) (0.002) Share of prime age -0.005** 0.000 0.001 -0.000 (15-64) (%) (0.003) (0.003) (0.003) (0.004) Male/female 0.018* 0.004 0.004** -0.000 sex ratio (0-14) (0.011) (0.005) (0.002) (0.001) Province fixed effect No No Yes No No Yes
Observations 1,861 1,861 1,861 1,861 1,861 1,861 Adjusted R-squared 0.211 0.283 0.529 0.427 0.551 0.677 AIC 1446 1275 464.1 2777 2328 1689
Notes: See Table 2. Robust standard errors are in parentheses. The symbols *, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively.
25
Table 5: Falsification test: the effect of e-commerce on growth in number of new firms
and total survival firms from 1997 to 2002
(1) (2) (3) (4)
VARIABLES Log(no. of new firms in 2002/
no. new of firms in 1997) Log(total no. of firms in 2002/
total no. of firms in 1997) No. of firms in initial year (log) -0.639*** -0.617*** -0.461*** -0.462***
(0.021) (0.019) (0.016) (0.014) E-commerce (standardized) 0.011 0.023 0.021 0.020
(0.017) (0.019) (0.013) (0.015) Log(population2000) 0.321*** 0.475*** 0.277*** 0.385***
(0.039) (0.035) (0.028) (0.030) Cluster index 0.204*** 0.179*** 0.101*** 0.071**
(0.036) (0.036) (0.027) (0.028) Share of urban population (%) 1.063*** 0.986*** 0.626*** 0.662***
(0.148) (0.141) (0.119) (0.120) Illiteracy rate (%) -0.022*** -0.021*** -0.014*** -0.020***
(0.004) (0.002) (0.003) (0.002) Share of prime age (15-64) (%) 0.027*** 0.038*** 0.024*** 0.030***
(0.007) (0.007) (0.005) (0.005) Male/female sex ratio (0-14) -0.002 -0.004 -0.004** -0.005**
(0.003) (0.002) (0.002) (0.002) Province fixed effect Yes Yes Yes Yes
Observations 1,387 1,861 1,688 1,861 Adjusted R-squared 0.574 0.543 0.579 0.565 Notes: See table 2. In columns (1) and (3), counties with zero number of new firms or total survival firms in 1997 are dropped from the sample. In Columns (2) and (4), we replace the zero values in 1997 with the value of the bottom 1 percentile among other counties. We also substitute the cluster index in 1995 by that in 2004 for the county without cluster index in 1995. In doing so, the sample size is kept the same as that in the main specifications in Tables 3-4.
26
Table 6: The effect of e-commerce on the growth in number of new firms by outsiders or insiders from 2009 to 2014
(1) (2) (3) (4) (5) (6)
Log(no. of new firms in 2014/
no. new of firms in 2009) Log(total no. of firms in 2014/
total no. of firms in 2009) VARIABLES Outsider Insider Combined Outsider Insider Combined E-commerce (standardized) 0.076*** 0.032** 0.023* 0.054*** 0.026*** 0.011
(0.013) (0.016) (0.013) (0.008) (0.010) (0.008) Outsider -0.387*** -0.192***
(0.022) (0.015) E-commerce * Outsider 0.064*** 0.057***
(0.013) (0.008)
Control variables Yes Yes Yes Yes Yes Yes Province fixed effect Yes Yes Yes Yes Yes Yes Observations 1,861 1,861 3,722 1,861 1,861 3,722 Adjusted R-squared 0.367 0.465 0.392 0.326 0.540 0.403
Notes: We define a firm as “outsider” (“insider”) if its largest shareholder’s birth county is different (the from (the same as) the firm’s registered location. The combined sample consists of the two subsamples of “outsider” and “insider”.
27
Table 7: The effect of e-commerce on the growth in number of new private firms and SOE from 2009 to 2014
(1) (2) (3) (4) (5) (6)
Log(no. of new firms in 2014/
no. new of firms in 2009) Log(total no. of firms in 2014/
total no. of firms in 2009) VARIABLES Private SOE Combined Private SOE Combined E-commerce 0.059*** 0.006 -0.066** 0.038*** 0.008 -0.003 (standardized) (0.012) (0.025) (0.028) (0.007) (0.007) (0.006) Private dummy 3.440*** 1.944***
(0.053) (0.061) Private* E-commerce 0.210*** 0.050***
(0.028) (0.008) Control variables Yes Yes Yes Yes Yes Yes Province fixed effect Yes Yes Yes Yes Yes Yes
Observations 1,861 1,861 3,722 1,861 1,861 3,722 Adjusted R-squared 0.467 0.533 0.696 0.529 0.405 0.772
Notes: See Table 2 for variable definitions. In counties with zero number of SOEs in 2009, we replace it with the value of the bottom 1% percentile among other counties in order to keep the sample size the same as for private firms. The results still hold if dropping counties without newly registered SOEs in 2009. Robust standard errors are in parentheses. The symbols *, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively.
28
Table 8: The effect of e-commerce on the initial capital of the legal representative of new firms in 2014
(1) (2) (3) Log(Initial capital)
VARIABLES Outsiders Insiders Combined E-commerce (standardized) -0.007 0.002 0.010
(0.009) (0.013) (0.011) Outsiders 0.443***
(0.016) E-commerce * Outsiders -0.023***
(0.006) Control variables Yes Yes Yes Industry fixed effect Yes Yes Yes Province fixed effect Yes Yes Yes
Observations 273,755 597,930 871,685 Adjusted R-squared 0.193 0.184 0.191
Notes: Initial capital refers to the registry capital at the time of registration. We define a firm as “outsider” if its largest shareholder’ birth county is different from the firm’s registered location. Robust standard errors are clustered in the county level.
29
Table 9: Contract intensity, e-commerce, and firm entry
(1) (2) (3) (4) (5) Log(no. of new firms 2014/no. of new firms 2009)
VARIABLES Total Total Outsider Insider Total Contract intensity*Ecommerce 0.086*** 0.081*** 0.101*** 0.061*** 0.050**
(0.016) (0.016) (0.019) (0.021) (0.021) Capital intensity* physical capital 0.025 -0.008 0.067* 0.029
(0.033) (0.024) (0.035) (0.021) Skill intensity* human capital 0.085*** 0.034 0.083*** 0.057***
(0.027) (0.021) (0.030) (0.018) Outsiders -0.245***
(0.018) CI*Ecommerce*Outsiders 0.040**
(0.019) County fixed effect Yes Yes Yes Yes Yes Industry fixed effect Yes Yes Yes Yes Yes Observations 38,917 38,917 38,917 38,917 77,834 Adjusted R-squared 0.214 0.214 0.161 0.195 0.163
Note: The dependent variable is the growth in log number of new manufacturing firms for each county at the SIC-3 level. The column labelled “total” refer to the combined sample of two subsamples of “insider” and “outsider.” Standardized beta coefficient is reported and robust standard errors are in parentheses. The symbols *, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively.
30
Table 10: Effect of purchase and combined e-commerce index on extensive firm growth
(1) (2) (3) (4)
Log(no. of new firms in 2014/
no. of new firms in 2009) Log(total no. of firms in 2014/
total no. of firms in 2009) VARIABLES Purchase Combined Purchase Combined E-commerce(Standardized) 0.107*** 0.112*** 0.053*** 0.061***
(0.022) (0.019) (0.015) (0.012) Control variable Yes Yes Yes Yes Province fixed effect Yes Yes Yes Yes
Observations 1,861 1,861 1,861 1,861 Adjusted R-squared 0.469 0.471 0.529 0.530
Note: the e-commerce purchase index measures the total amount of online purchase at the county level. The e-commerce combined index measures both the online sales and online purchase at the county level. Robust standard errors are in parentheses. The symbols *, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively.
31
Table 11: Spillover effect of e-commerce on growth in newly registered firms at the prefectural and provincial level
(1) (2) (3) (4)
SAMPLE Total rural counties Excluding the highest county
Cells Prefecture Province Prefecture Province Max(E-commerce) 0.039** 0.031* 0.007 -0.031
(0.017) (0.016) (0.018) (0.041) Control variables Yes Yes Yes Yes Province fixed effect Yes No Yes No
Observations 317 30 317 30 Adjusted R-squared 0.500 0.681 0.495 0.414
Note: Max(E-commerce) takes the value of e-commerce index in the county which ranks the top in e-commerce index within a prefecture or province. The sample covers only rural counties. When calculating the extensive firm growth at the prefectural (provincial) level in column 3 (4), the county with highest e-commerce index within a prefecture (province) is excluded.
32
Figure 1: The spatial distribution of online sales index
Source: The online sales index is at the prefectural level. We use the highest value of e-commerce index
among counties within a prefecture to represent online sales index at the prefectural level.
33
Figure 2: The structure of production for suitcases/bags under traditional and e-commerce business models
34
Figure 3: The starting capital for outsiders and insiders among three types of
business
Note: Outsiders means that shop owners are born in counties outside Baigou.
35
Figure 4: Growth of entrants of insiders and outsiders
Note: Growth of entrants is Log (no. of new firms in 2014/ no. of new firms in 2009). The lines are Locally
Weighted Scatterplot Smoothing (LOWESS), conditional on the total number of entrants in the initial
period. The top and bottom 1 percentile of e-commerce index are trimmed to remover outliers.