Social 1-1

Journal of Comparative Economics xxx (2015) xxx–xxx
Contents lists available at ScienceDirect
Journal of Comparative Economics

journal homepage: www.elsevier.com/locate/jce
Social interaction, Internet access and stock market

participation—An empirical study in China q
Pinghan Liang, Shiqi Guo ⇑
Research Institute of Economics and Management, Southwestern University of Finance and Economics, China
a r t i c l e i n f o a b s t r a c t
Article history: Liang, Pinghan, and Guo, Shiqi—Social interaction, Internet access and stock market par-
Received 17 February 2014 ticipation—An empirical study in China
Revised 14 February 2015
Available online xxxx Social interaction plays an important role in transmitting relevant information to potential
investors. However, the informational role of social interaction might be affected by other
JEL classification: information channels, which is to a large extent ignored in previous studies. Using a
C31 national representative household finance survey data covering more than 8000 Chinese
D14 households, we demonstrate that social interaction alone positively affects household
O16 stock market participation, but Internet access mitigates the influence of social interaction.
Z13 In particular, among households with the access to Internet, sociable households in effect
are associated with a 6 percentage-point decrease in the probability to participate in the
Keywords:
stock market. This finding supports the substitution between Internet access and social
Social interaction
Internet access
interaction as information channels. Moreover, we also identify the social multiplier effect
Information of social interaction: sociable households living in the communities with higher stock mar-
Stock market participation ket participation rate are more likely to invest in stocks. Journal of Comparative Economics
xxx (xx) (2015) xxx–xxx. Research Institute of Economics and Management, Southwestern
University of Finance and Economics, China.
Ó 2015 Association for Comparative Economic Studies. Published by Elsevier Inc. All rights
reserved.
1. Introduction
Information is key for decision-making. Individuals usually rely on their social networks, e.g., relatives, friendships,
neighborhood, etc., to acquire information for making important decisions, such as searching for jobs (Granovetter, 1973),
participating in criminal activities (Glaeser et al., 1996), voting (Katz and Lazarsfeld, 1955), and entering the stock market
(Cohen et al., 2008). On the one hand, there is evidence that a stock market with wide participation improves the efficiency
of resource allocation, facilitates financial development, and consequently causes economic growth (Levine, 1997, 2005). On
q
The authors are grateful to the insightful comments from an anonymous referee as well as the editor. The valuable suggestions from Zaichao Du, Yi
Huang, Han Li, Jingye Shi, Yan Yuan, Zhichao Yin, Ligang Zhong, as well as the attendees of 2013 China Household Finance Research Workshop (Chengdu),
2013 Doctoral Meeting on Quantitative Economics (Xiamen), Geneva-China Workshop on International Macroeconomics and Finance (2014), Ronald Coase
Institute Workshop (2014). CHFS (2011) data are kindly provided by China Household Finance Survey and Research Center, Southwestern University of
Finance. This paper is partially supported by the China Scholarship Council. All remaining errors are our own.
⇑ Corresponding author.
E-mail addresses: liangph@swufe.edu.cn (P. Liang), stevenguo_swufe@qq.com (S. Guo).
http://dx.doi.org/10.1016/j.jce.2015.02.003
0147-5967/Ó 2015 Association for Comparative Economic Studies. Published by Elsevier Inc. All rights reserved.
Please cite this article in press as: Liang, P., Guo, S. Social interaction, Internet access and stock market participation—An empirical study in
China. Journal of Comparative Economics (2015), http://dx.doi.org/10.1016/j.jce.2015.02.003
2 P. Liang, S. Guo / Journal of Comparative Economics xxx (2015) xxx–xxx
the other hand, Mankiw and Zeldes (1991), and Vissing-Jorgensen (2002) demonstrate that limited stock market partic-
ipation is also closely linked to the equity premium puzzle (Mehra and Prescott, 1985): the observed long-run returns on
stocks are much higher than returns on risk-free assets. Hence, the potential reliance of household financial decisions on
information acquisition from the interaction with others might influence important aggregate outcomes.
As many researchers observe (Shiller, 1989; Hong et al., 2004; Ivkovic and Weisbenner, 2007; Brown et al., 2008; Li,
2014), social interaction might serve as a channel to disseminate market-related information, consequently affecting house-
hold stock market participation. However, since the information transmitted by social interaction is usually biased (Shiller,
1984, 1990; Kaustia and Knupfer, 2012), making financial decisions based on information from social interaction might
result in investors’ heterogeneous beliefs, consequently contributing to asset price bubbles (Scheinkman and Xiong, 2004;
Hong and Stein, 2007) and large fluctuations of financial markets. Therefore, understanding how stock market participation
decisions are affected by social interaction has important implications for macroeconomics and finance.
Social interaction might exhibit two effects in affecting household financial decisions: informational effect and social multi-
plier effect. It is well-acknowledged that social interaction may serve as a mechanism for information exchange by means of
word-of-mouth communication or ‘‘observational learning’’ (Ellison and Fudenberg, 1995). Specifically, word-of-mouth
communication makes it easy and convenient for potential investors to learn about opening accounts, making transactions,
etc., and to obtain relevant information by talking with experienced friends and neighbors. This is referred to as the
informational effect of social interaction on stock market participations, which reflects an individual’s active use of
information. On the other hand, if an individual’s behavior is affected by the behavior of neighbors, then a more sociable
person would be influenced more by peers. Consequently, a sociable person in a high stock market participation community
is more likely to participate in the stock market. In other words, social interaction increases the correlation between
community-level stock market participation rates and individual participation. We call it the social multiplier effect of social
interaction, which reflects that individuals are passively influenced by the average behavior (characteristics) of the
community they live in.
A novel point is that we explore the substitution between social interaction as an information channel and other infor-
mation channels, e.g., Internet access. This also helps us to identify the informational effect of social interaction. We now live
in an age of information explosion. Therefore, we need to allocate our limited attention and information-processing capaci-
ties to different information channels (Sims, 2003; Veldkamp, 2011). As an information channel, the Internet substantively
changes our way of learning and acquiring information. If individual investors rely more on Internet to acquire information
for participating the stock market, they have to reduce using the information from social interaction, e.g., word-of-mouth
communication through channels other than the Internet. Therefore, social interaction and Internet access should substitute
with each other in affecting stock market participation decisions.
To examine the impacts of social interaction on stock market participation, we need the proxies for social interaction
which do not cover the possible Internet social network. Therefore, we are interested in the face-to-face interaction and tele-
phone interaction. China Household Finance Survey (CHFS) 2011, a national representative survey data designed for collect-
ing detailed household financial information, enables us to fulfill this task. We use the amount spent on gifts to non-family
members, as well as the expenditure on telephone communication, to measure the household sociability, respectively.
Conditional on whether the level of sociability measured by gifts or communication expenditure is in the upper 50% or
not, households are categorized into two types, ‘‘sociable’’ and ‘‘non-sociable’’, respectively. Moreover, we employ the com-
munity interview refusal rate to proxy for the community-level degree of social interaction.
Our paper makes four contributions. First, our empirical results find support for the substitution between social interac-
tion and Internet access as information channels. More specifically, among households without the Internet access, sociable
households are 1.6% more likely to enter the stock market than those non-sociable ones. However, for households with
access to the Internet, being sociable in effect is associated with a 6 percentage-point decrease in the probability to partici-
pate in the stock market. Therefore, the access of Internet mitigates the informational effect of social interaction. We think
this result might shed light on the weakening informational effect of social interaction in the Internet Age in general.
Second, after controlling for the informational effect of social interaction, we also demonstrate that for a household with
the average Internet access level (22%), sociability does not significantly increase stock market participation. Hence, in con-
trast with Kaustia and Knupfer (2012), we support the view that negative returns might discourage market participation.
Third, we provide evidence for the social multiplier effect of social interaction. We show that the effect of the community-
average stock market participation rate on households’ stock market participation is influenced by the level of households’
sociability, i.e., sociable households are more likely to be associated with stock market participation if the neighbors
participate in the stock market. Active social interaction (having above the median level sociability) raises the probability
to participate in the stock market by 5.13% for the household living in high participation communities, but by only 0.67%
for those living in low participation communities.
Fourth, our paper also contributes to the understanding of stock market participation in China, the second largest econ-
omy in the world. By the end of 2011, Shanghai Stock Exchange, one of the two stock exchanges in China, became the world’s
6th largest stock market by market capitalization at US$2.3 trillion.1 With more than 140 million investor accounts, the stock
market has become an important way for Chinese households to allocate financial assets and diversify risks. However, due to
1
http://en.wikipedia.org/wiki/Shanghai_Stock_Exchange.
P. Liang, S. Guo / Journal of Comparative Economics xxx (2015) xxx–xxx 3
the lack of data, the previous studies on Chinese individual stock market participation only focus on the households’ behavior in
certain coastal provinces (Li, 2006a), a few large cities (Li, 2006b; Zhou et al., 2011), or provincial-level aggregate behavior (Liu
et al., 2013). To the best of our knowledge, our paper is the first to provide a comprehensive analysis of stock market partic-
ipation decisions among Chinese households.
The rest of this paper proceeds as follows: Section 2 reviews the existing literature. In Section 3, we describe our data set,
explain the variables, and present our theoretical hypotheses. Section 4 presents the empirical results. Section 5 concludes.
2. Literature review
It is well-acknowledged that household stock market participation is much lower than would be predicted by the
Consumption-based Asset Pricing Model and other models. Many studies have been conducted to investigate the determi-
nants underlying household participation in the stock market. For instance, Vissing-Jorgensen (2003) shows that since it is
easier for wealthier households to overcome the fixed costs of participating, participation strongly increases in wealth.
Bernheim and Garrett (2003) and Hong et al. (2004) establish that the holding of stocks and other financial assets is increas-
ing with household educational level. Hong et al. (2004) suggest that the race of households is linked with stock market par-
ticipation. Guiso et al. (2004, 2008) have studied the effect of trusting behavior on portfolio choice decisions, and shown that
countries with high generalized trust level exhibit on average high stock market participation rates because trust in others
matters for the subjective expected return.
As Grossman and Stiglitz (1980) illustrate, information plays a key role in financial markets. Bogan (2008) suggests that
the Internet could mitigate three causes for low stock market participation: transaction costs, information costs, and limited
access. He demonstrates that Internet/computer usage substantially raises U.S. households’ stock market participation rates
by reducing transaction costs. Using a German panel household survey, Glaser and Klos (2013) suggest that the positive
effect of Internet usage on stock market participation depends on households’ financial literacy. In China, access to the
Internet to a large extent reflects the availability of stock market information. Hence, the penetration of the Internet should
play an important role in influencing stock market participation of Chinese households.2 A novel part of this paper is that we
consider Internet access as one information channel, and jointly examine the information role of Internet access and that of
social interaction. Especially, we provide a hypothesis about the substitution relationship between these two information chan-
nels and empirically examine it.
Manski (1993, 2000) distinguishes three elements of social interaction. The first is endogenous interaction, which cap-
tures that ‘‘the propensity of an agent to behave in some way varies with the behavior of the group’’. Specifically, endogenous
interaction can take effect through mechanisms of word-of-mouth communication, common topic pleasure,3 and social
norm.4 The second is contextual interaction (exogenous interaction), which shows ‘‘the propensity of an agent to behave in
some way varies with exogenous characteristics of the group members’’. For example, the losses (gains) in stock investment
of neighbors are likely to demonstrate stock investment unattractiveness (attractiveness). The third one is the correlated effect,
which attributes the similar behaviors of the group members to their similarities in individual characteristics and institutional
environment. However, it is notorious that empirically identifying social interaction is subject to a serious endogeneity prob-
lem—since households are not assigned randomly into communities, their unobserved characteristics may result in the behav-
ioral consistency.
Among the mechanisms underlying endogenous interaction, word-of-mouth communication represents a clear informa-
tion channel. As Ellison and Fudenberg (1995) note, economic agents have to ‘‘rely on whatever information they have
obtained via causal word-of-mouth communication’’. It is convenient for individuals to learn about opening accounts,
making deals obtaining related information by talking with friends and neighbors who are experienced in stock market
investment.5 By analyzing a cross-sectional data of U.S. investors, Hong et al. (2004) suggest that endogenous interaction
provides individuals with a channel to obtain information and allow observational learning. Furthermore, Brown et al.
(2008) use an instrumental variable approach to establish the observed causality between an individual’s decision to invest
in the stock market and the average participation rate of his community (community effect). They suggest that word-of-mouth
communication is the source of this ‘‘community effect’’. Using a field experiment conducted with a financial brokerage,
Bursztyn et al. (2014) demonstrate that both learning effect and social norms underlie the social influence in financial decisions.
Pool et al. (forthcoming) use the residential address history of fund managers to identify the role of information transmission
among neighbors and correlated effect. In this paper, we confirm that a higher level of social interaction is associated with a
higher probability to participate in the stock market. Moreover, we take an alternative perspective in addressing the magnitude
2
According to ‘‘China Statistical Bulletin of National Economic and Social Development 2010’’, there are 420 million Internet users in China in 2010, with the
Internet penetration rate 31.8%.
3
Individuals might enjoy pleasures from the conversations based on common interests (Becker, 1991). For example, an investor can get pleasure from a
conversation about stock issues with his peers who also own stocks (Hong et al., 2004).
4
The ‘‘social norm’’ mechanism is also called the ‘‘keeping up with the Joneses’’ effect (Abel, 1990; Brown et al., 2008). This phenomenon provides insights
for many studies—the conformity model by Bernheim (1994), the consumption-based model with external habit formation by Campbell and Cochrane (1999),
and the theoretical work based on competition on local resources of Demarzo et al. (2004).
5
Liu et al. (2013) distinguish word-of-mouth communication from observational learning in analyzing Chinese stock market participation. However,
identifying these two models is not the priority of our paper.
of the influence of sociability. Our finding that Internet access decreases the marginal effect of the word-of-mouth communica-
tion supports the view that social interaction acts as an informational channel, and illustrates the limitation of social
interactions in the Internet Age.
Previous studies also suggest that social interaction might be a biased information transmission channel. Shiller (1984,
1990) suspects that the spreading of investment success stories within social networks might partly explain the patterns
of stock market fluctuations. Kaustia and Knupfer (2012) suggest that positive returns that local peers experienced encour-
age stock market entry, whereas negative returns do not make entry less likely. Using China’s provincial-level aggregate data,
Liu et al. (2013) show that the word-of-mouth effect is stronger in the bull market. However, by using various proxies of
social interaction, Li (2006a, 2006b) suggests that the substantial negative returns incurred on Chinese investors might be
responsible for the negative effect of contextual interaction on stock market participation in China. Our paper finds evidence
that social interaction discourages stock market participation when Internet access is available. This supports the negative
effect of contextual interaction.
Finally, our work is also related to the literature on social trust and social capital. Since social interaction and trust are two
essential components of social capital (Durlauf and Fafchamps, 2004), they may mix together to influence households’
investment behavior. It is found that stock market investors tend to exhibit more generalized trust (Guiso et al., 2003). In
addition, households with more reliable information should be less affected by the cultural stereotypes (Guiso et al.,
2008). Following Guiso et al. (2008) and El-Attar and Poschke (2011) show that trust level affects the household asset al-
location between housing and risky financial assets in Spain. Georgarakos and Pasini (2011) separate trust from sociability
in determining stock market participation. They show that these two factors both have distinct and sizeable positive effects
on stock market participation in European countries, and sociability is likely to partly balance the discouragement effect on
stockholding induced by low generalized trust level. It is not the priority of this paper to distinguish the role of trust from
that of sociability. However, from the perspective that sociability might be closely related to social trust level, either general-
ized or specified, our finding of the substitution between sociability and Internet access might complement the empirical
results that the usage of modern communication devices might contribute to the decline of social capital in developing coun-
tries (Olken, 2009).
3. Data and hypotheses
3.1. Data description
We use the data from the China Household Finance Survey 2011 (CHFS 2011), which was jointly conducted by the
People’s Bank of China and the Southwestern University of Finance and Economics.6 This survey aims to be a national
representative by collecting micro-level information about household finance. It employs a stratified three-stage probability
proportion-to-size random sample design, and the respondents are representative of all the residences in China. The
respondents would receive a small gift (value about $15) for answering the questionnaire. The average time to interview
a household is 2 h.
This survey was carried out in July and August, 2011, which covers the detailed financial information of 8438 households
located in 320 communities across 25 of China’s provinces in the end of 2010. We drop 370 observations (4% of the total
sample), including 346 households with negative or zero gross annual income and 24 households who invest in the compa-
nies where their family members are working or once worked. All the variables presented in Table 1 are from the question-
naires and are relevant indicators of the survey.7
The CHFS dataset includes information on each family member—Family Size takes values from 1 to 18—so we have 29,324
individual observations in total. Following Hong et al. (2004), we take the highest age and education level within the family
as the household ‘‘age’’ and ‘‘education’’. Two education dummies, Middle Education (55.18%) and College Education (31.43%),
are constructed, with the former representing junior high, high school or secondary/vocational school educational level, and
the latter including college/vocational, undergraduate, master and PhD degrees. The dummy variable Finance takes the value
of one (1.95%) if at least one family member works in the financial industry. Income (Log) is the logarithm of household gross
annual income in 2010, while Wealth (Log) is the logarithm of household wealth.8 The dependent variable Stock equals to one
for those households entering stock markets (opening an account in stock markets), and zero for the others. In the sample, 8.68%
of the households participate in the stock market.9
We use three proxy variables to measure the level of social interaction (sociability): the interview response rate of a com-
munity, a household’s expenditures on cash-gift, and a household’s communication expenditure. Here we will explain these
variables in detail.
6
See Gan et al. (2014) for detailed description about this dataset.
7
For the English version of questionnaires, see http://chfs.swufe.edu.cn/upload/files/CHFS-English.pdf.
8
The household wealth is calculated as the total household assets in production and management projects, land and real estate, vehicles, other non-financial
assets and financial assets.
9
The stock market participation rate is 11% in a household survey data of Guangdong Province in 2004 (Li, 2006a); 32% in a survey of urban residents of 15
Chinese cities in 2007 (Li and Guo, 2009); and 23% in a household survey covering Guangdong, Beijing and Shanghai in 2008 (Zhou et al., 2011). Given that CHFS
covers a national representative sample of households including both urban and rural residents, this participation rate also looks plausible.
Table 1
Descriptions of variables.
Variable N Mean Median St. dev. Min Max Explanation

Stock 8062 0.0868 0 0.282 0 1 Binary: Whether household owns stock
investment accounts
Community stock 8068 0.0869 0.0357 0.123 0 0.609 Average stock participation rate in the community
Social interaction and Internet access variables
Response 8068 0.920 1 0.136 0.302 1 Response rate of the community
Cash-gift 6029 0.500 1 0.500 0 1 Binary: Whether the ratio of cash-gift expenditure
to total family income is above the sample median
Communication 8068 0.503 1 0.500 0 1 Binary: Whether the ratio of communication
expenditure to total family income is above the
sample median
Internet 8052 0.223 0 0.416 0 1 Binary: Whether household not only owns
computers but also use Internet as an information
source
Cash-gift Internet 6021 0.0977 0 0.297 0 1 Interaction term of Cash-gift and Internet
Communication Internet 8052 0.118 0 0.323 0 1 Interaction term of Communication and Internet
Households characteristics
Income (Log) 8068 10.11 10.30 1.341 1.386 14.91 Gross annual household income taking log
1st Quintile of income distribution 8068 0.200 0 0.400 0 1 Binary: Whether household income lies within the
range of lowest 20%. (Baseline group)
2nd Quintile of income distribution 8068 0.201 0 0.401 0 1 Binary: Whether household income lies within the
range of 20–40%
3rd Quintile of income distribution 8068 0.199 0 0.399 0 1 Binary: Whether household income lies within the
range of 40–60%
4th Quintile of income distribution 8068 0.200 0 0.400 0 1 Binary: Whether household income lies within the
range of 60–80%
5th Quintile of income distribution 8068 0.200 0 0.400 0 1 Binary: Whether household income lies within the
range of highest 20%
Wealth (Log) 8058 11.96 12.22 1.839 1.609 16.57 Household wealth taking log
Education 7999 4.370 4 1.814 1 9 Categorical: Highest education level within the
household. (1 = never attended school. . .9 = PhD)
Primary education 7999 0.134 0 0.341 0 1 Binary: Whether the highest education level
within a household is Never Attended School or
Primary School. (Baseline group)
Middle education 7999 0.552 1 0.497 0 1 Binary: Whether the highest education level
within a household is junior high, high school or
secondary/vocational school
College education 7999 0.314 0 0.464 0 1 Binary: Whether the highest education level
within a household is college/vocational,
undergraduate degree, masters degree or PhD
Age 8068 56.23 56 15.15 17 112 Highest age within a household
Travel 8017 0.443 0 0.497 0 1 Binary: Whether household have expenditure on
travel last year
House 8067 0.912 1 0.284 0 1 Binary: Whether household owns houses
Family size 8068 3.493 3 1.551 1 18 Number of family numbers
Children 8068 0.552 0 0.755 0 9 Number of children within the household
(younger than 16)
Old 8068 0.382 0 0.678 0 3 Number of old people within the household (older
than 65)
Unemployed 8068 0.119 0 0.404 0 6 Number of unemployed people within the
household
Finance 8068 0.0195 0 0.138 0 1 Binary: Whether its family members work in the
financial industry
Job category 1 8068 0.322 0 0.467 0 1 Binary: Whether is employed in a work unit
Job category 2 8068 0.0876 0 0.283 0 1 Binary: Whether is in individual or privately
owned business, entrepreneur
Job category 3 8068 0.264 0 0.441 0 1 Binary: Whether farming at home
Job category 4 8068 0.000992 0 0.0315 0 1 Binary: Whether was once retired, and is now
reengaged by original work unit as needed
Job category 5 8068 0.0357 0 0.186 0 1 Binary: Whether is freelance
Job category 6 8068 0.00483 0 0.0694 0 1 Binary: Whether belongs to other job categories
Job category 7 8068 0.284 0 0.451 0 1 Binary: Whether the job category is not recorded
Personal traits
Risk attitude 7940 3.847 4 1.233 1 5 Categorical: Which investment to choose. (1 = high
risk, high return projects;. . .5 = not willing to take
on any risks)
Risk loving 8068 0.132 0 0.339 0 1 Binary: Whether to choose High risk, high return
or Slightly higher risk, lightly higher return
(continued on next page)
Table 1 (continued)
Variable N Mean Median St. dev. Min Max Explanation

projects. (Baseline group)
Risk neutral 7940 0.260 0 0.439 0 1 Binary: Whether to choose Average risk, average
return projects
Risk averse 7940 0.606 1 0.489 0 1 Binary: Whether to choose Slightly lower risk,
slightly lower return projects or Not willing to take
on any risks
Seatbelt 7960 1.840 2 0.849 1 3 Categorical: Whether to normally wear seatbelts
while driving. (1 = Yes; 2 = It depends; 3 = No)
Traffic rule 8004 1.282 1 0.653 1 4 Categorical: Whether to follow traffic rules and
wait for the stop light when crossing the road.
(1 = always comply. . .4 = rarely obey)
Expectation 7969 2.084 2 0.825 1 5 Categorical: In what way to expect that China’s
economy is more likely to change in the next three
to five years. (1 = Much better. . .5 = Much worse)
Donation 7781 0.499 0 0.500 0 1 Binary: Whether ratio of Wenchuan earthquake
donation to total income is above the sample
median
Regional characteristics
Urban 8068 0.611 1 0.488 0 1 Binary: Whether in urban areas. (Baseline group)
Rural 8068 0.389 0 0.488 0 1 Binary: Whether in rural areas
Western 8068 0.229 0 0.420 0 1 Binary: Whether in eastern region. (Baseline
group)
Central 8068 0.303 0 0.460 0 1 Binary: Whether in central region
Eastern 8068 0.468 0 0.499 0 1 Binary: Whether in western region
Public security 8056 2.503 2 0.887 1 5 Categorical: Comments on local public security.
(1 = extremely good. . .5 = very poor)
Community income (Log) 8068 10.65 10.60 0.699 8.768 12.70 Average income of the community taking log
Community wealth (Log) 8068 12.71 12.62 0.940 10.45 15.17 Average wealth of the community taking log
Community unemployment 8068 0.0350 0.0238 0.0394 0 0.205 Average unemployment rate of the community
Community education 8068 4.371 4.292 1.006 2.389 7.638 Average education level of the community
Community age 8068 56.23 56.95 5.415 41.95 71.11 Average age of the community
Community internet 8068 0.222 0.167 0.214 0 0.854 Proportions of the households not only owning
computers but also using Internet as an
information source within the community
Community stock (Exclusive) 8062 0.0868 0.0357 0.123 0 0.622 Average stock participation rate of other
households within the community
Community Stock Category 8068 0.193 0 0.824 1 1 Categorical: Takes the value of 1 if no households
owns stocks in the community he lives (45.5%). For
the remaining communities whose stock market
participation rates are positive, we divide them
into two groups by their median and assign them
the values of 0 (28.4%) and 1 (26.2%) respectively
Response community Internet 320 0.155 0.100 0.168 0 0.838 Interaction term of Response and Community
Internet
Cash-gift Community Stock 6029 0.143 0 0.578 1 1 Interaction term of Cash-gift and Community
Category Stock Category
Communication Community 8068 0.116 0 0.585 1 1 Interaction term of Communication and
Stock Category Community Stock Category
We think the interview response rate of a community might be a plausible proxy of the community-level extent of social
interaction. During the survey stage, most interviewers are introduced to the respondents by the local social workers who
have lived in the neighborhood for a long period. And only if an interviewer is refused at least three times by a household
would that household be replaced by another in the sample pool. Hence, it is reasonable to say that the non-cooperative
household in the survey has little interaction with the neighbors in the local community;10 a high refusal rate of a community
thus reflects a low level of social interaction (or trust) within that community.
The overall refusal rate of CHFS is 11.6%, with 16.5% in urban areas and 3.2% in rural areas, which is quite low in compar-
ison with other major surveys.11 The variable Response, which represents the interview response rate of the community, varies
10
The payments are constant in all areas, a potential challenge is that the wealthy households might refuse to respond if they feel the compensation is not
worthy of the effort. In Table 2 we could observe that the community level response rate is negatively correlated with wealth and income. However, the highest
urban household wealth in the data is 1 billion RMB ($160,000,000), and the highest rural household wealth amounts to 100,000,000 RMB ($16,000,000), thus it
is unlikely that the value of compensation is the main driving force underlying refusal behavior.
11
The overall refusal rate of China Health and Retirement Longitudinal Survey (CHARLS) in 2008 is 15.2–20.7% in the urban sample and 10.1% in the rural
sample, respectively. The Survey of Consumer Finance (SCF, USA) in 2007 has the refusal rates of 32.2% and 67.3% respectively in the AP sample and the listed
sample. The refusal rates of Consumer Expenditure Survey (CES, USA) in 2005 are 25.5% in the interview sample and 29% in the diary sample. And the Survey of
Household Income and Wealth (the SHIW, Italy) in 2008 has an overall refusal rate of 43.9%.
from 0.3 to 1. In effect, 52% of surveyed households live in the communities with no interview refusals. It has been debated that
some personal traits may be associated with sociability as well as the propensity to participate in the stock market (Hong et al.,
2004). In this sense, the use of response rate as a proxy for social interaction avoids such potential endogenous problems
because it measures the overall social interaction level of a community which is exogenous for households.
The expenditure on cash-gift reflects a household’s pattern of favor exchange in Chinese society and it has been employed
by many researchers to measure the size of household’s social network in China (Yang et al., 2011; Ma and Yang, 2011). In
the survey, the respondents answered the value of cash (or cash equivalents) given to non-family members for weddings and
funerals (including birthday gifts) in 2010. Since the expenditure on cash-gift is largely affected by household income and
local culture, we use the ratio of cash-gift expenditure to total family income. Among the 6029 households with records
of cash-gift expenditure, dummy Cash-gift equals to one if the ratio of cash-gift expenditure to total family income is above
the sample median (3%).
The expenditure on communication is recorded as the expenditure a household spent on telephone and Internet fees last
month. The main focus of this paper about social interactions is on its information role, which might be captured by the
expenditure on communication. It equals to one if the ratio of communication expenditure to total family income is above
the sample median (0.3%). Consequently, among those defined as high communication expenditure households, 2462 house-
holds (49.98% of the urban sample) live in the urban areas and 1598 households (50.86% of the rural sample) stay in rural
areas. Since the expenditure on Internet is included, this measure might be positively correlated with the usage of Internet.
However, as Table 2 shows, the correlation coefficient between this variable and our measure of Internet usage is only 0.03,
but the correlation coefficient between this and Cash-gift is 0.23. Hence, we think this variable is a plausible measure of
sociability.12
We consider both computer ownership and Internet usage to measure Internet access. In the survey, the respondents are
asked separately whether they own computers (Question C8001), and whether they use Internet as the main source of infor-
mation (Question A4001).13 The dummy variable Internet equals to one if a household not only owns computers but also uses
the Internet as the main source of information, and zero for other cases. 22.3% of households actively use Internet to acquire
information.
To overcome the potential endogeneity problems arising from omitted variables, we make several attempts. Firstly, since
individual’s characteristics can be related with both active social interaction and stock market participation, we control for
respondent’s personal attitudes and traits. We take two dummies with respect to risk attitude—Risk Neutral and Risk Averse—
with the benchmark group of risk lovers. As Hong et al. (2004) suggest, sociable people may be bolder and possibly more
willing to invest in stocks without full knowledge about the stock market. To control for prudence and rashness, we use
Traffic Rule and Seatbelt to measure whether the individual follows traffic rules, and whether he normally wears seatbelts
while driving. Optimism is also associated with social interaction and may influence stock market participations as well
(Hong et al., 2004). Based on CHFS questionnaires, we use Expectation to be a proxy for optimism. In addition, we also control
for Donation to Wenchuan Earthquake, which reflects an individual’s other regarding preferences.
Secondly, a household with little extra work time may not have sufficient time to invest in stocks, and could only acquire
information by interacting with neighbors and friends. Therefore, we include the number of children, old and unemployed
members, whether they spent money on travel last year and whether they own houses (Hong et al., 2004; Zhou et al., 2011)
to measure a household’s free time.
Finally, the level of social interactions might have regional characteristics (Durlauf and Fafchamps, 2004). These regional
characteristics might be representative of local culture, and can also affect economic outcomes such as stock market partic-
ipation (Guiso et al., 2006). Therefore, we include dummy variables Central and Eastern to represent the central and eastern
regions in China, dummy variable Rural and other relevant variables such as household’s comments on local public security,
average community income level, and local unemployment rate.14 All the variables are listed and described in details in
Table 1.
3.2. Theoretical hypotheses
Previous studies in the U.S. and China have established the positive effect of social interactions on stock market partic-
ipation (Hong et al., 2004; Li, 2006a, 2006b). However, the mechanisms behind it are still ambiguous. First, the lack of data
and indicators make it difficult to identify contextual interaction, as well as the different mechanisms of endogenous
12
As an anonymous referee points out, mobile could be used to connect to the Internet to make online surfing, which might also be related to Internet access.
According to the 26th Statistical Report about China’s Internet Development issued by CNNIC (China Internet Network Information Center, http://www.cnnic.
net.cn/) on 2010, there were 277 million mobile internet users in the mid of 2010, more than one third of them were students, the majority of them were under
30, and the main function of mobile internet was instant communication (61.5%), which is a clear kind of social interactions, while the commercial use of
mobile internet was rare (6.1% for online payment through mobile). On the other hand, 77% users only used the online surfing functions of mobile in the leisure
time. Therefore, in 2010, the year the survey was conducted, mobile was not the main access to Internet for the heads of households, and the role of mobile
internet as a financial information channel for individual investors was also limited.
13
A4001 explicitly asks the main sources of information acquisition, the options include newspaper, mobile, TV, radio, Internet, relatives/friends/colleagues,
etc.
14
We did not add dummies for province in order to avoid multicollinearity.
8
P. Liang, S. Guo / Journal of Comparative Economics xxx (2015) xxx–xxx

Table 2
Correlations of main variables.
Stock Response Cash-gift Communication Internet Income Wealth Education Age Risk Seatbelt Traffic Donation Expectation
(Log) (Log) attitude rule
Stock 1
Response 0.1954* 1
Cash-gift 0.1064* 0.0995* 1
Communication 0.0504* 0.00360 0.2271* 1
Internet 0.3378* 0.2322* 0.1314* 0.0300* 1
Income (Log) 0.2538* 0.1889* 0.3533* 0.3467* 0.3253* 1
Wealth (Log) 0.2813* 0.1828* 0.1365* 0.0451* 0.3185* 0.4594* 1
Education 0.2829* 0.2174* 0.1489* 0.0224* 0.4329* 0.4338* 0.3519* 1
Age 0.0870* 0.0853* 0.0457* 0.1030* 0.3010* 0.1485* 0.0894* 0.2127* 1
Risk Attitude 0.1764* 0.0865* 0.0367* 0.0529* 0.2932* 0.1755* 0.1628* 0.2003* 0.2653* 1
Seatbelt 0.0671* 0.0957* 0.0494* 0.0261* 0.1056* 0.1009* 0.1164* 0.1019* 0.0636* 0.0952* 1
Traffic Rule 0.00440 0.00500 0.00330 0.00390 0.0271* 0.0176 0.0374* 0.0221* 0.0315* 0.0217 0.1262* 1
Donation 0.1357* 0.1089* 0.0700* 0.2027* 0.2111* 0.0407* 0.1613* 0.2672* 0.0948* 0.1017* 0.1064* 0.0688* 1
Expectation 0.1441* 0.1454* 0.0606* 0.000300 0.1823* 0.1068* 0.1131* 0.1842* 0.0998* 0.0444* 0.0223* 0.0397* 0.0643* 1
*
Indicates a 0.05 significance level of correlation coefficient.
interactions. Second, social interaction may interact with other factors. For instance, Hong et al. (2004) show the peer effects
to be stronger for the households who are white, educated and wealthy.
On the one hand, individuals might actively use their social networks as an instrument to acquire information, hence
social interactions serve as an information channel. This is the informational effect of social interactions. Informational
effects imply that individuals’ sociability positively contributes to stock market participation. On the other hand, individuals
might be influenced by the behavior/characteristics of peers in their social networks, hence social interactions may also serve
as a channel through which peers influence an individual’s behavior. This is the social multiplier effect of social interactions.
This means that sociability expands the influence of peers on individual behavior, so a sociable individual is more likely to be
influenced by the neighborhood participation decisions.
We first address the informational effect of social interactions. Recent theories on individual information choice (Sims,
2003; Veldkamp, 2011) suggest that households are subject to information-processing capacity constraint and have to allo-
cate attention among different information sources. Therefore, as an information channel, the role of social interactions
should be negatively affected by the usage of other information channels. In effect, the possible substitution relationship
between other information channels and social interactions could be considered as evidence in favor of the informational
effect of social interactions. Therefore, our first hypothesis highlights the substitution effect between sociability and
Internet access.
Hypothesis 1. As two channels to transmit stock market information, both Internet access and social interaction can
increase stock market participation, but they substitute for each other.
Informational effects of social interactions capture the role of word-of-mouth communication mechanisms within
endogenous interactions. The effect of contextual interactions on Chinese stock market participation, however, is ambiguous.
On the one hand, both the cases of gains and losses are likely to produce demonstration effects, which will promote and dis-
courage stock market investments, respectively. On the other hand, the absence of the investment performance of an
individual’s neighbors in data causes difficulty in measuring the effect of contextual interactions. In effect, traditionally it
is difficult to identify these different mechanisms underlying social interactions. However, if we could control for the effect
of word-of-mouth communication, we could reasonably believe that what we observe is the aggregate effect of the remain-
ing mechanisms.
As Li (2006a, 2006b) suggests, contextual interactions negatively affect stock market participation in China since most
individual investors lost money in the stock market, and only the negative experience in the stock market is demonstrated
to the neighbors. According to CHFS 2011, only 10% of households who own a stock market account gain from investment. In
the year information was collected (2010), the stock market index in Shanghai Exchange decreased 470 points (14%). And the
index in the end date of 2010 was 2808.08 points, significantly lower than the historical height of 6124 points on October 16,
2007. Moreover, most individuals have loss aversion (Kahneman and Tversky, 1979). Therefore, the negative effect of con-
textual interactions is likely to dominate any other remaining positive effects of social interactions, including the common
topic pleasure, social norm and the demonstration effect created by the cases of gains. Hence, we set up our second hypothe-
sis regarding the effects of other mechanisms.
Hypothesis 2. The above ‘‘substitution relationship’’ causes social interactions to discourage stock market participation
when Internet access is available.
Our third hypothesis focuses on the social multiplier effect. It is shown that an individual is more likely to participate in
the stock market provided that a higher fraction of individuals in the neighborhood are stock market investors (Brown et al.,
2008). Hong et al. (2004) and Brown et al. (2008) both find that if there are more social interactions, there is a stronger
relationship between community-level variables and individual participation decisions. Hence, frequent social interactions
exacerbate the influence of neighbors on individual behavior, which could be referred to as ‘‘social multiplier effect’’ in terms
of Hong et al. (2004). In this paper, we examine the magnitude of social multiplier effect on the individual level. According to
Manski (1993), this effect could also be driven by common topic pleasure and social norm, in additional to word-of-mouth
communication.15 Hence we state our Hypothesis 3 as follows:
Hypothesis 3. There exists the ‘‘social multiplier effect’’—a sociable household is more likely to invest in stocks if he lives in
a community with higher participation rates.
Hypothesis 3 predicts that the marginal effect of social interactions on stock market participation is greater in a high-par-
ticipation community. The social multiplier effect implies that individuals will also be influenced by the peers through social
interactions. Hence, social interactions could also serve as a channel of receiving influences.16
15
Brown et al. (2008) actually take a broader definition of ‘‘word-of-mouth interactions’’, which is in effect equivalent to endogenous interactions here.
16
As an anonymous referee points out, there is the possibility that Internet access also serves as a channel through which other people’s behavior affect
individual investor. However, in our data we could not categorize the usage of Internet between acquiring information and receiving influence from the peers.
4. Empirical results
4.1. Descriptive statistics
In Table 2, we present the correlations of the main variables. First of all, the three sociability variables are correlated with
each other. The correlation between Cash-gift and Communication is 0.23, while their correlations with Response are weak—
the coefficients are 0.1 and 0.004 respectively. It seems all of them capture different aspects of social interactions. It is note-
worthy that Response is measured on the community level while Cash-gift and Communication are measured on household
level, so it can be misleading to simply look at their correlations. We also observe weak correlations between the social vari-
ables and individual traits variables.
Second, Stock is highly correlated with Internet, with the correlation coefficient 0.34. Stock is negatively correlated with
sociability variables, with the coefficients 0.2, 0.11 and 0.05. Also, Stock is positively correlated with Income, Wealth and
Education, and negatively correlated with Age and Risk Attitude.
Fig. 1 provides a simple tabulation of average stock market participation across different groups, which allows us a rough
look at the overall stylized facts about participation in China. From Fig. 1-1 to -6, we divide the sample into high and low
groups regarding Income, Education, Internet access, Cash-gift, Communication and Response, with the blue bar representing
the ‘‘high’’ group and red bar the ‘‘low’’ group. For each indicator, we make further comparisons of urban–rural areas and
eastern–central–western regions to capture the relationship of stock market participation and regional imbalanced develop-
ment in China.
There are considerable regional differences in stock participation rates. The participation rates fall from the coastal
areas (the eastern) to inland (the central and western). This still holds when we look at urban and rural subsamples
separately. For urban–rural comparisons, urban areas have higher participation than rural areas. As shown in Fig. 1-2,
the high education households and high income households are much more likely to participate in the stock market.
Stock market participation also increases in access to the Internet, both for the region and urban–rural division. Take,
for example, the eastern households. The participation rate is 5.9% for those without Internet access, and 30% for those
with Internet access. Again this confirms the significant correlation coefficient between Internet access and stock market
participation in Table 2.
However, in line with the negative correlations between sociability variables and stock market participation reported in
Table 2, stock market participation is higher in the low social interaction group in Fig. 1. This is in sharp contrast with the
views that social interactions are associated with high stock market participation (Hong et al., 2004; Brown et al., 2008).
Previous studies in China (Li, 2006a, 2006b) suggest that the widespread losses in the stock market investment in China
discourage stock market investment due to the contextual interaction mechanism. Here we propose another explanation
that the high correlation between Internet access and stock market participation ‘‘crowds out’’ the positive effect of social
interaction because, as information channels, Internet and sociability substitute with each other—as the Hypothesis 1
indicates—in increasing stock market participation.
1-1 Variation of Stock Market Participation 1-2 Variation of Stock Market Participation 1-3 Variation of Stock Market Participation
Stock Market Participation Rate

across income across education across Internet access

.25
.25
.4
.2
.2
.3
.15
.15
.2
.05 .1
.1
.1
.05
0
0
0
East Middle West East Middle West East Middle West East Middle West East Middle West East Middle West East Middle West East Middle West East Middle West
Total Urban Rural Total Urban Rural Total Urban Rural
Top 50% income Bottom 50% income Top 50% education Bottom 50% education with Internet access without Internet access
1-4 Variation of Stock Market Participation 1-5 Variation of Stock Market Participation 1-6 Variation of Stock Market Participation

across cashgift expenditures across communication expenditures across respond rates

.25
.2
.2
.2
.15
.15
.15
.1
.1
.1
.05
.05
.05
0
0
0
East Middle West East Middle West East Middle West East Middle West East Middle West East Middle West East Middle West East Middle West East Middle West
Total Urban Rural Total Urban Rural Total Urban Rural
Top 50% cashgift Bottom 50% cashgift Top 50% communication Bottom 50% communication Top 50% respond rates Bottom 50% respond rates
Fig. 1. Stock market participation across different groups. Notes: High Internet access group refers to the households owning computers and use the
Internet to collect information; low Internet access group refers to other households. The numbers of observations of these two groups in the whole sample,
rural sub-sample and urban sub-sample are 1792 and 6260; 175 and 2959; 1617 and 3301 respectively.
4.2. Informational effect of social interactions
To examine the impacts of social interactions and Internet access on households’ stock market participation, as well as the
substitution relation between these two information channels, we employ the following specification:
Stocki ¼ a þ b1 Sociabilityi þ b2 Sociabilityi Interneti þ b3 Internet i þ cIndiv iduali þ dHouseholdi þ uCommunityi þ e; ð1Þ
The marginal effect of sociability on stock market participation could be assessed by estimating b1 + b2 ⁄ Interneti. Similarly,
the marginal effect of Internet access is represented by b3 + b2 ⁄ sociabilityi. Hypothesis 1 predicts that the sign of b1 and b3
are positive because Sociability and Internet both could promote stock market participation. However, substitutions between
these two channels imply that the sign of b2 would be negative. Individuali refers to the personal traits of the survey respon-
der within household i,17 e.g., risk attitude, prudence and rashness, etc. Householdi represents household i’s characteristics,
including income, wealth, educational level, family size and family members. Communityi measures the characteristics of the
community household i living, e.g., the average community indicators, rural–urban and regional variables. We control these per-
sonal traits and household and community characteristics in order to alleviate the omitted variable problem.
The baseline results are presented in Table 3. The regressions from column (1) to (4) are based on household-level obser-
vations, where we use Cash-gift (Communication) to be the proxies of sociability in Column (1) and (2) (Column (3) and (4)).
Since the dependent variable is binary, we run both OLS and Probit regressions, with the marginal effects reported in the
latter case.18 All the standard errors are adjusted as the cluster-robust standard errors (Rogers, 1994) clustering on the com-
munity level because the community characteristic variables only vary with communities. Due to the missing values of vari-
ables, the numbers of observations are somewhat different across columns. In columns (2) and (4), we replace Income (Log)
with four dummies corresponding to the second, third, forth and fifth quintiles of the income distribution, and take further con-
trols for individual traits, household burdens and job categories. Finally, the regression in column (5) is based on 320 commu-
nity observations, with the dependent variable the community stock market participation. And Social interaction and Internet
access are measured by the Response and Community Internet. Other community characteristics are also controlled.
We focus on the informational effect first. The coefficients of Internet are large and statistically significant under 1% level
in all specifications. For those non-sociable households whose sociability level is in the bottom 50% (e.g., sociabilityi = 0), hav-
ing the access to Internet is associated with a 7-percentage-points-higher probability of stock market participation in the
Probit regression, and a 17-percentage-points-higher probability in the OLS regression. For the households without
Internet access, e.g., Interneti = 0, being sociable is only associated with a 2% point increase in the likelihood to participate
in the stock market. The coefficients of Sociability Internet are significantly negative at 1% level in all specifications, indi-
cating a remarkable substitution relationship between sociability and Internet access. Take the OLS result in column (2)
as an example: among those non-sociable households, those with Internet access have a 13.8% higher participation rate than
those without; however, for the sociable households, Internet access only leads to a 6.1% increase in probability
(0.1375 0.0764 ⁄ 1 = 0.0611).19 For the households without Internet access, the difference in stock market participation rate
between sociable and non-sociable households is 1.6%; but for households with the access to the Internet, being sociable in
effect is associated with a 6-percentage-points (0.016 0.076 ⁄ 1 = 0.06) decrease in the probability to participate in the stock
market! This supports our Hypothesis 1. Although the two information channels substitute for each other, because the marginal
effect of Internet access is always positive while that of sociability could be negative, we could say that access to the Internet
dominates sociability in increasing stock market participation.
As to the aggregate effect of social interactions, for a household with the average Internet access level (22%), sociability
does not substantially increase stock market participation effectively.20 As suggested by Li (2006a, 2006b), the two most influ-
ential factors of social interaction on stock market participation are word-of-mouth communication and contextual interaction.
And contextual interaction is the only possible element that may absorb the positive effect of word-of-mouth communication
by spreading the loss cases of stock market investment. Hence, we support Hypothesis 2 about the negative effect of contextual
interactions on Chinese households’ stock market participation decisions. However, when compared with Internet access, social
interactions have only a limited role in increasing stock market participation. Given that only 22 percent of Chinese households
have access to the Internet, there is still room for social interactions to increase stock market participation.
Similar results are demonstrated in column (5) that uses the community-level observations. We replace Sociabilityi in Eq.
(1) with the community response rate, and Interneti with the Community Internet. For communities with the sample average
response rate (92%), a ten-percentage-point increase in Community Internet is associated with a 2.8 percentage point increase
(10% ⁄ (0.77 0.53 ⁄ 92%) = 2.8%) in the community stock market participation rate; but for communities with no interview
refusals (191 communities in this case), a ten-percentage-point increase in the community Internet access level is only
associated with a 2.4-percentage-point increase (10% ⁄ (0.77 0.53 ⁄ 100%) = 2.4%) in community stock market participation
17
It is worth noting that the responder is not necessarily the person who makes the decision on stock market participation.
18
We also run the Logit regression, the results of which are similar and are presented in Appendix Table 3.
19
Considering the magnitude of interaction effect in nonlinear models is not equal to the marginal effect of the interaction term (Ai and Norton, 2003), we
only interpret the interaction terms of OLS regressions numerically. Though differences exist in the coefficients between the results of OLS and Probit
regressions, their signs and statistical significance are the same for most cases.
20
The aggregate impacts of social interaction on stock market participation (b1 + b2 ⁄ Interneti) from column (1) to (4) are 0.25%, 1.18%, 0.04%, 0.92%, 0.83%,
1.62%, 0.77% and 1.62%, respectively.
Table 3
Substitution relationship between social interaction and internet access.
Cash-gift Communication Response

(1) (2) (3) (4) (5)
OLS Probit OLS Probit OLS Probit OLS Probit OLS
*** ** *** * *** *** * **
Sociability 0.0228 0.0185 0.0164 0.0155 0.0174 0.0232 0.0119 0.0224 0.1316***
(0.0062) (0.0088) (0.0062) (0.0092) (0.0056) (0.0084) (0.0062) (0.0093) (0.0412)
Sociability Internet 0.0923*** 0.0306** 0.0764*** 0.0287** 0.1168*** 0.0318*** 0.0890*** 0.0284*** 0.5355***
(0.0211) (0.0132) (0.0215) (0.0139) (0.0193) (0.0102) (0.0196) (0.0108) (0.1265)
Internet 0.1615*** 0.0770*** 0.1375*** 0.0670*** 0.1921*** 0.0733*** 0.1619*** 0.0642*** 0.7772***
(0.0176) (0.0103) (0.0175) (0.0105) (0.0190) (0.0093) (0.0193) (0.0096) (0.1165)
Middle education 0.0063 0.0411* 0.0039 0.0261 0.0080 0.0447** 0.0023 0.0307
(0.0068) (0.0228) (0.0067) (0.0218) (0.0056) (0.0193) (0.0049) (0.0188)
College education 0.0551*** 0.0813*** 0.0569*** 0.0561*** 0.0464*** 0.0768*** 0.0515*** 0.0562***
(0.0106) (0.0223) (0.0114) (0.0213) (0.0087) (0.0187) (0.0096) (0.0183)
Risk neutral 0.0761*** 0.0472*** 0.0712*** 0.0447*** 0.0646*** 0.0373*** 0.0615*** 0.0368***
(0.0138) (0.0090) (0.0138) (0.0091) (0.0121) (0.0076) (0.0120) (0.0075)
Risk averse 0.0861*** 0.0610*** 0.0787*** 0.0554*** 0.0754*** 0.0504*** 0.0711*** 0.0486***
(0.0129) (0.0084) (0.0127) (0.0085) (0.0110) (0.0066) (0.0110) (0.0070)
Age 0.0007*** 0.0005** 0.0020*** 0.0016*** 0.0008*** 0.0007*** 0.0017*** 0.0014***
(0.0003) (0.0003) (0.0005) (0.0004) (0.0002) (0.0002) (0.0004) (0.0003)
Family size 0.0090*** 0.0119*** 0.0173*** 0.0160*** 0.0080*** 0.0107*** 0.0152*** 0.0144***
(0.0023) (0.0029) (0.0036) (0.0039) (0.0018) (0.0024) (0.0027) (0.0031)
Wealth (Log) 0.0220*** 0.0253*** 0.0258*** 0.0318*** 0.0176*** 0.0220*** 0.0189*** 0.0268***
(0.0028) (0.0037) (0.0035) (0.0044) (0.0022) (0.0028) (0.0026) (0.0036)
Income (Log) 0.0115*** 0.0128*** 0.0094*** 0.0140***
(0.0030) (0.0032) (0.0027) (0.0031)
2nd Quintile of income 0.0085 0.0095 0.0097 0.0104
distribution (0.0088) (0.0182) (0.0067) (0.0146)
3rd Quintile of income 0.0166* 0.0113 0.0123 0.0206*
distribution (0.0098) (0.0149) (0.0081) (0.0120)
4th Quintile of income 0.0191 0.0204 0.0132 0.0309**
distribution (0.0124) (0.0151) (0.0119) (0.0132)
5th Quintile of income 0.0455*** 0.0441*** 0.0474*** 0.0519***
distribution (0.0162) (0.0152) (0.0150) (0.0130)
Seatbelt 0.0000 0.0013 0.0005 0.0013
(0.0052) (0.0048) (0.0041) (0.0037)
Traffic rule 0.0024 0.0000 0.0015 0.0018
(0.0051) (0.0052) (0.0040) (0.0045)
Expectation 0.0193*** 0.0119*** 0.0166*** 0.0093***
(0.0050) (0.0043) (0.0042) (0.0035)
Donation 0.0221*** 0.0235*** 0.0175*** 0.0177***
(0.0080) (0.0082) (0.0067) (0.0066)
Children 0.0239*** 0.0177** 0.0208*** 0.0154***
(0.0063) (0.0073) (0.0049) (0.0057)
Old 0.0226*** 0.0229*** 0.0171** 0.0184**
(0.0086) (0.0088) (0.0071) (0.0073)
Unemployed 0.0066 0.0051 0.0004 0.0007
(0.0091) (0.0100) (0.0073) (0.0080)
Travel 0.0135** 0.0058 0.0159*** 0.0102*
(0.0067) (0.0069) (0.0059) (0.0058)
House 0.0365** 0.0549*** 0.0111 0.0388**
(0.0167) (0.0169) (0.0130) (0.0151)
Public security 0.0062 0.0098** 0.0066** 0.0088***
(0.0039) (0.0039) (0.0033) (0.0032)
Finance 0.0903** 0.0241 0.0898** 0.0194
(0.0373) (0.0172) (0.0352) (0.0150)
Community income 0.0578*** 0.0459*** 0.0497*** 0.0351*** 0.0495*** 0.0388*** 0.0454*** 0.0320*** 0.0041
(Log) (0.0154) (0.0104) (0.0153) (0.0105) (0.0130) (0.0086) (0.0132) (0.0088) (0.0084)
Table 3 (continued)
Cash-gift Communication Response

(1) (2) (3) (4) (5)
OLS Probit OLS Probit OLS Probit OLS Probit OLS
Community 0.1106 0.1161 0.0891 0.0785 0.1384 0.0841 0.1351 0.0446 0.1158
unemployment (0.1077) (0.0917) (0.1115) (0.0956) (0.0917) (0.0727) (0.0950) (0.0765) (0.0820)
Rural 0.0021 0.0439** 0.0062 0.0262 0.0041 0.0434*** 0.0036 0.0260 0.0049
(0.0187) (0.0196) (0.0197) (0.0203) (0.0155) (0.0166) (0.0166) (0.0172) (0.0106)
Eastern 0.0025 0.0045 0.0006 0.0014 0.0063 0.0033 0.0021 0.0010 0.0023
(0.0111) (0.0124) (0.0107) (0.0116) (0.0095) (0.0106) (0.0093) (0.0102) (0.0077)
Central 0.0022 0.0079 0.0007 0.0070 0.0018 0.0087 0.0016 0.0075 0.0012
(0.0090) (0.0121) (0.0088) (0.0118) (0.0075) (0.0101) (0.0074) (0.0101) (0.0072)
Community wealth (Log) 0.0230***
(0.0069)
Community education 0.0186***
(0.0068)
Community age 0.0011
(0.0008)
Job categories No No Yes Yes No No Yes Yes No
Constant 0.8951*** 0.8405*** 0.7439*** 0.7168*** 0.5835***
(0.1715) (0.1673) (0.1455) (0.1453) (0.1026)
Observations 5902 5902 5642 5642 7850 7850 7439 7439 320
R-squared 0.1904 0.2092 0.1949 0.2119 0.7036
Notes: All the standard errors are adjusted as the cluster-robust standard errors clustering on the community level and the t-values are reported in the
parentheses.
*
Represent the significance level of 0.1.
**
***
rate. For communities with no access to internet (97 communities), a 10% increase in Community Response rate is associated
with a 1.3% increase (10% ⁄ (0.13 0.53 ⁄ 0) = 1.3%) in average community participation; but the same magnitude of increase
in response rate is associated with only 0.13% (10% ⁄ (0.13 0.53 ⁄ 22%) = 0.13%) for communities with the average Internet
availability.
Other variables, including risk attitude, education, income and wealth, all significantly influence stock market partic-
ipation. Take the Probit results in column (2), for example. Other things being equal, compared with a risk lover, the proba-
bility to participate drops by 4.5% (5.5%) if the household is risk neutral (risk averse). Apart from the findings that stock
market participation increases in income and education, we find evidence for income and education ‘‘thresholds’’ of partic-
ipation. Compared with primary school education households, Middle Education does not promote participation significantly,
but households with College Education are 5.6% more likely to invest in stocks at 0.01 significance level. Households whose
incomes lie in the middle three quintiles are not significantly different with the lowest quintile in participation, but those in
the highest quintile have a 4.4 percent higher probability of participating significantly.
Table 3 also suggests that household structure affects stock market participation. Generally speaking, the probability of
participation increases by 0.2% for an additional year in the household’s highest age and decreases by 1.6% for an additional
household member. Family burden also matters. Owning a house and having an additional old member lower the partic-
ipation by 5.5% and 2.2%, while an additional child is associated with a 1.8-percentage-points-higher probability of partic-
ipation. Among the five proxies for personality traits, only Expectation and Donations have significant influences. The
coefficients of regional variables imply a considerable difference of stock market participation rates across China, especially
between rural and urban areas, and among the communities with different average income levels.
Though we attempt to alleviate the endogeneity problem caused by omitted variables by controlling for various personal
traits and household characteristics, it is still possible that a household decided to invest in the stock market in the first
place, and then bought computers and acquired information via the Internet. Therefore, we replace Interneti in Eq. (1) with
the community average Internet access. The results are reported in Table 4, with the only difference from Table 3 that the
Internet and its interaction terms are replaced by Community Internet and the corresponding interaction terms.
Table 4 shows that the Internet access measured by the community average still influences the stock market participation
significantly in all eight specifications. The substitution relationship between Internet access and social interactions remains,
and the coefficient of the interaction term are still significant in OLS regressions. Take the OLS results in column (2) for exam-
ple: a 10% increase in the community average Internet access—a move from 90th to 95th percentile of its distribution—is
associated with a 3-percentage-pints increase in the probability of participation for non-sociable households, but only with
Table 4
Results using community average Internet access as proxy.
Cash-gift Communication
(1) (2) (3) (4)
OLS Probit OLS Probit OLS Probit OLS Probit
*** *** *** *
Sociability 0.0371 0.0165 0.0278 0.0154 0.0213 0.0162 0.0116 0.0160
(0.0081) (0.0134) (0.0082) (0.0141) (0.0061) (0.0120) (0.0066) (0.0129)
Sociability Internet 0.1556*** 0.0358 0.1284*** 0.0414 0.1004*** 0.0090 0.0583* 0.0110
(0.0361) (0.0298) (0.0330) (0.0297) (0.0350) (0.0274) (0.0320) (0.0262)
Internet 0.3677*** 0.1515*** 0.3042*** 0.1140*** 0.3289*** 0.1175*** 0.2605*** 0.0833***
(0.0507) (0.0314) (0.0495) (0.0302) (0.0482) (0.0294) (0.0460) (0.0277)
Middle education 0.0107 0.0431* 0.0016 0.0285 0.0155*** 0.0466** 0.0011 0.0330*
(0.0068) (0.0238) (0.0067) (0.0222) (0.0055) (0.0204) (0.0049) (0.0195)
College education 0.0601*** 0.0910*** 0.0629*** 0.0656*** 0.0518*** 0.0862*** 0.0580*** 0.0652***
(0.0110) (0.0232) (0.0116) (0.0218) (0.0088) (0.0198) (0.0093) (0.0189)
Risk neutral 0.0794*** 0.0499*** 0.0739*** 0.0466*** 0.0711*** 0.0405*** 0.0661*** 0.0391***
(0.0145) (0.0095) (0.0144) (0.0095) (0.0127) (0.0079) (0.0124) (0.0078)
Risk averse 0.1016*** 0.0737*** 0.0908*** 0.0654*** 0.0938*** 0.0632*** 0.0848*** 0.0583***
(0.0137) (0.0087) (0.0133) (0.0087) (0.0120) (0.0069) (0.0116) (0.0072)
Age 0.0003 0.0002 0.0018*** 0.0014*** 0.0004* 0.0004* 0.0016*** 0.0013***
(0.0003) (0.0003) (0.0005) (0.0004) (0.0002) (0.0002) (0.0004) (0.0004)
Family size 0.0057** 0.0090*** 0.0153*** 0.0144*** 0.0053*** 0.0084*** 0.0138*** 0.0134***
(0.0023) (0.0028) (0.0035) (0.0039) (0.0018) (0.0023) (0.0026) (0.0031)
Wealth (Log) 0.0230*** 0.0264*** 0.0263*** 0.0334*** 0.0189*** 0.0233*** 0.0197*** 0.0288***
(0.0028) (0.0037) (0.0035) (0.0045) (0.0022) (0.0029) (0.0027) (0.0038)
Income (Log) 0.0146*** 0.0150*** 0.0136*** 0.0176***
(0.0030) (0.0033) (0.0027) (0.0032)
2nd Quintile of income distribution 0.0110 0.0065 0.0132* 0.0083
(0.0088) (0.0185) (0.0069) (0.0150)
3rd Quintile of income distribution 0.0157 0.0090 0.0139* 0.0198
(0.0097) (0.0151) (0.0080) (0.0123)
4th Quintile of income distribution 0.0125 0.0212 0.0078 0.0330**
(0.0119) (0.0152) (0.0114) (0.0133)
5th Quintile of income distribution 0.0514*** 0.0446*** 0.0617*** 0.0567***
(0.0163) (0.0154) (0.0145) (0.0130)
Seatbelt 0.0000 0.0013 0.0013 0.0011
(0.0050) (0.0047) (0.0040) (0.0037)
Traffic rule 0.0046 0.0017 0.0041 0.0004
(0.0050) (0.0052) (0.0039) (0.0045)
Expectation 0.0208*** 0.0139*** 0.0189*** 0.0112***
(0.0051) (0.0043) (0.0043) (0.0035)
Donation 0.0204** 0.0230*** 0.0174*** 0.0186***
(0.0079) (0.0082) (0.0066) (0.0066)
Children 0.0237*** 0.0182** 0.0210*** 0.0161***
(0.0062) (0.0072) (0.0047) (0.0055)
Old 0.0235*** 0.0224** 0.0182** 0.0197***
(0.0088) (0.0090) (0.0073) (0.0075)
Unemployed 0.0062 0.0033 0.0002 0.0019
(0.0094) (0.0102) (0.0073) (0.0081)
Travel 0.0157** 0.0081 0.0201*** 0.0134**
(0.0066) (0.0068) (0.0059) (0.0057)
House 0.0337** 0.0562*** 0.0096 0.0423***
(0.0159) (0.0168) (0.0127) (0.0150)
Public security 0.0050 0.0079** 0.0047 0.0069**
(0.0040) (0.0040) (0.0034) (0.0033)
Finance 0.0929** 0.0259 0.0951** 0.0211
(0.0386) (0.0179) (0.0370) (0.0158)
Community income (Log) 0.0192* 0.0232** 0.0177 0.0185* 0.0171* 0.0201** 0.0176* 0.0186**
(0.0112) (0.0100) (0.0115) (0.0105) (0.0100) (0.0087) (0.0103) (0.0094)
Table 4 (continued)
(1) (2) (3) (4)
OLS Probit OLS Probit OLS Probit OLS Probit
*
Community unemployment 0.0601 0.1626 0.0279 0.1179 0.1065 0.1210 0.0838 0.0758
(0.1152) (0.0931) (0.1173) (0.0970) (0.1001) (0.0748) (0.1012) (0.0786)
Rural 0.0227 0.0360* 0.0242 0.0212 0.0202 0.0386** 0.0212 0.0233
(0.0193) (0.0187) (0.0200) (0.0195) (0.0162) (0.0155) (0.0169) (0.0165)
Eastern 0.0039 0.0085 0.0016 0.0043 0.0049 0.0058 0.0011 0.0005
(0.0111) (0.0121) (0.0109) (0.0114) (0.0097) (0.0106) (0.0096) (0.0103)
Central 0.0011 0.0107 0.0024 0.0094 0.0004 0.0104 0.0002 0.0089
(0.0090) (0.0120) (0.0088) (0.0118) (0.0078) (0.0103) (0.0077) (0.0103)
Job categories No No Yes Yes No No Yes Yes
Constant 0.5654*** 0.5603*** 0.4760*** 0.4697***
(0.1227) (0.1261) (0.1116) (0.1137)
Observations 5906 5906 5646 5646 7858 7858 7446 7446
R-squared 0.1838 0.2038 0.1806 0.2009
parentheses.
*
**
***
a 1.76-percentage-points increase for sociable households. The effect of social interactions remains the same magnitude for
all specifications.
To check the robustness of our results, we explore two variations regarding the definition of stock market participation.
First, we require the investment amount to be greater than 10,000 RMB (about 1660 USD, one third of the sample median
household income level). This might preclude the inactive small individual investors. In this case, 64 additional households
are assigned to the non-participation group. Second, we think perhaps making frequent deals is information-demanded,
instead of just owning stock investment accounts. So we pay attention to the number of stocks holding and take 95 house-
holds who own stock accounts, but do not hold any stocks, out of the participation group. The corresponding results are pro-
vided in Appendix Table 4, and our findings in Table 4 remain regardless of these variations.
4.3. Social multiplier effect of social interactions
The social multiplier effect refers to the idea that social interactions act as a channel through which other community
members influence individual decisions. As our Hypothesis 3 suggests, the marginal effect of social interactions on stock
market participation should be greater in a high-participation community. To examine this effect, we use the Community
Stock (Exclusive) variable, which is calculated as the average participation rate of the other households in the community.
Therefore, similar to the approach in Hong et al. (2004), in Table 5 we add Community Stock (Exclusive) in Columns (1)
and (3) with the same control variables as those in columns (2) and (4) of Table 3, and replace Community Stock
(Exclusive) with the interaction term of Sociability and Community Stock Category in columns (2) and (4).21 The Community
Stock Category variable takes the value of 1 for the households who live in the communities with zero participation rate
(3668 households), and 0 and 1 respectively for those in non-zero participation communities with average participation below
(2287 households) and above (2113 households) the median level. The results about Sociability, Internet and Community stock
participation are shown in Table 5, and the results of other variables are provided in Appendix Table 5.
Table 5 provides support for the community effect suggested by Hong et al. (2004) and Brown et al. (2008). It shows that
the average participation rate of the other households within the community is significantly positively associated with the
household’s stock market participation. As revealed in results of the Probit regression, a 10-percentage-points rise in
Community Stock (Exclusive)—a move from 50th to 75th percentile of its distribution—increases a household’s participation
probability by 1.9–2.2% in columns (1) and (3). Households living in high participation communities are more likely to invest
in stocks.
Now we look at the social multiplier effect of social interactions. The results in columns (2) and (4) of Table 5 suggest this
effect. For instance, the OLS results in column (2) show that when Internet is held constant at its average level, active social
interactions raise the probability to participate in stock market by 5.13 if the household lives in a high participation
21
The Community Stock Category variable is not included to avoid the multicollinearity problem—the correlations between Sociability Community Stock
Category and Community Stock Category are 0.70 and 0.69 respectively using cash-gift and communication expenditures as sociability proxies.
Table 5
The ‘‘Social Multiplier Effect’’ of stock market participation.
(1) (2) (3) (4)
OLS PROBIT OLS PROBIT OLS PROBIT OLS PROBIT
*** * *** * ** ***
Sociability 0.0161 0.0151 0.0300 0.0049 0.0113 0.0213 0.0323 0.0022
(0.0060) (0.0090) (0.0076) (0.0098) (0.0060) (0.0091) (0.0077) (0.0093)
Sociability Internet 0.0681*** 0.0267* 0.1059*** 0.0422*** 0.0798*** 0.0258** 0.1301*** 0.0412***
(0.0216) (0.0138) (0.0228) (0.0144) (0.0191) (0.0106) (0.0204) (0.0107)
Internet 0.1245*** 0.0637*** 0.1494*** 0.0706*** 0.1469*** 0.0608*** 0.1815*** 0.0692***
(0.0172) (0.0101) (0.0181) (0.0101) (0.0183) (0.0090) (0.0197) (0.0091)
Community stock (Exclusive) 0.5370*** 0.2164*** 0.5118*** 0.1882***
(0.0566) (0.0402) (0.0488) (0.0357)
Sociability Community stock 0.0446*** 0.0609*** 0.0576*** 0.0617***
category (0.0077) (0.0077) (0.0071) (0.0068)
parentheses. The other control variables are the same as those in columns (2) and (4) of Table 3.
*
**
***
community; however, the increase is only 0.67% for those living in low participation communities. Active social interactions
in effect are associated with 3.79% lower probability of participation for those living in zero participation communities! This
supports our Hypothesis 3 that the marginal effect of social interactions is higher for those households in high participation
communities.
Our result is in line with Hong et al. (2004), which demonstrates that among American communities with low partic-
ipation rates, social interactions reduce stock market participation by 0.5%. Besides, the aggregate effects of social interac-
tions in these two columns are 0.19%, 1.42%, 0.37% and 1.13%, respectively; also close to the previous result.
4.4. Analyses of Sub-samples
China is a country with more than 1.3 billion residents, thus wide differences across the country exist. In this subsection,
as a robustness check, we investigate the informational effect and social multiplier effect of stock participation in groups
with different demographic characteristics, e.g., urban–rural areas, high-low income, and high-low educational level. We
firstly run the OLS regression for different subsamples with the same controls as the specification reported in columns (2)
and (4) of Table 4, and then test if the coefficients in different subsample pairs are equivalent using ‘‘seemingly unrelated
estimation’’ (Zellner, 1962).22 The main results and the comparisons of coefficients with Chi-square statistics are present in
Table 6, with the results of other variables provided in Appendix Table 6-1, -2 and -3.23
As revealed in Table 6, overall our basic results remain in different subsamples. Social interactions and Internet access
increase stock market participation. The coefficients of their interaction term are always negative, supporting the sub-
stitution relationship between them. The marginal effect of social interaction increases in community participation rates.
However, the informational effect of social interactions varies with subsamples. For the comparison of urban–rural subsam-
ples (Table 6-1), social interactions have a significantly greater effect for rural households.24 Access to the Internet is also
associated with higher probability for stock market participation in rural areas than that in urban, though not statistically sig-
nificant. It implies that social interactions remain an effective channel of information transmission in rural China and that there
is more room for increasing rural stock market participation.
In Table 6-2, households are divided into high and low annual income groups according to the sample median level
29602.5 RMB (about 4852 USD). The influence magnitudes of Sociability and Internet, as well as the two interaction terms
are all smaller for low income households than high income households. It implies the limited roles of information channels
for low income households to participate in the stock market, as well as the possibility that apart from the information chan-
nel, low income families may face other restrictions on participating in the stock market. As to educational level, we divide
households into the high-low education groups by the sample median (high school education) in Table 6-3. The coefficients
22
We also try the dummy variables for subsamples and put its interaction terms with other variables into the regression models, which allows us to estimate
the coefficients of different sample pairs at the same time. The results are similar.
23
The methodology of generating Sociability, Donation and income distribution dummies is that their original values are compared with the subsample
medians before the they are assigned with the binary values. Take the Sociability dummies of the urban subsample for example, we compare the ratio of cash-
gift and communication expenditures to household income with the urban subsample median level, and then divide urban households into high and low
sociability groups.
24
Due to the missing values of variables, the numbers of observations in regressions are less than the numbers of households in different subsamples.
Table 6
Comparisons of coefficients in different subsamples.
Urban Rural Chi-square Urban Rural Chi-square
Table 6-1
Sociability 0.0155 0.0789*** 8.07*** 0.0181* 0.0586*** 5.78**
Sociability Internet 0.0934*** 0.1662* 0.65 0.1092*** 0.1651* 0.41
Internet 0.1329*** 0.2275** 1.06 0.1524*** 0.2650** 1.08
Sociability Community stock category 0.0585*** 0.0784*** 0.85 0.0619*** 0.0683*** 0.14
Number of observations 3641 2001 4596 2814
High income Low income Chi-square High income Low income Chi-square
Table 6-2
Sociability 0.0308*** 0.0289*** 0.02 0.0362*** 0.0323*** 0.07
Sociability Internet 0.1508*** 0.1109*** 0.62 0.1586*** 0.0392 12.97***
Internet 0.1789*** 0.1608*** 0.21 0.1935*** 0.0579** 7.60***
Sociability Community stock category 0.0820*** 0.0428*** 8.33*** 0.0840*** 0.0546*** 3.70*
High education Low education Chi-square High education Low education Chi-square
Table 6-3
Sociability 0.0280 0.0207** 0.19 0.0245 0.0364*** 0.54
Sociability Internet 0.1344*** 0.0360* 4.31** 0.1167*** 0.0238 4.30**
Internet 0.1609*** 0.0967*** 2.92* 0.1549*** 0.0958*** 2.08
Sociability Community Stock Category 0.0682*** 0.0348*** 5.36** 0.0715*** 0.0544*** 1.79
Notes: The results are from OLS regressions for different subsamples with the other control variables the same as those in columns (2) and (4) of Table 4. The
Chi-square statistics are obtained from ‘‘Seemingly unrelated estimation’’ (Zellner, 1962).
*
**
***
of regressions suggest that high education households rely more on Internet access to make stock market participation deci-
sions, and their behavior is rarely correlated with the frequency of social interactions. It might be that households with
higher educational levels are better at processing and analyzing information, which makes them put more attention on
the public and more precise Internet access.
5. Conclusions
We use a national representative sample of Chinese households to investigate the informational effect and the social mul-
tiplier effect of social interactions on stock market participation in China. We find that as two information channels, both
access to the Internet and social interactions increase stock market participation, but they substitute with each other.
This suggests that the usage of modern communication devices (Internet) might crowd out the informational effect of social
interactions. The marginal effect of social interactions falls if the household has access to the Internet. In addition, after con-
trolling for the substitution between social interactions and access to the Internet, the remaining mechanisms of social inter-
actions discourage participation as a whole. It confirms the negative effect of the contextual interaction—the negative
demonstration effect of the widespread loss cases in stock investment—on stock market participation in China. We also show
that the marginal effect of social interactions on stock market participation is greater in high participation communities,
which supports the social multiplier effect of social interactions. The above findings are robust in different subsamples, with
the coefficients of these information channels varied.
A large body of literature has focused on the role of social interactions in promoting stock market participation. (Hong
et al., 2004; Li, 2006a, 2006b; Brown et al., 2008). Our study focuses on the information role of social interactions.
Although social interactions are dominated by other channels (here the access to Internet) in acquiring information, they still
have room to take effect because of the low Internet penetration rate in China, especially in rural areas.
Because the first wave of CHFS only provides the cross-sectional data, even though we use several attempts to control for
the possible omitted variables problem, we still could not completely rule out the endogeneity problem. Establishing a
clearer causal relationship between the informational effect of social interactions and stock market participation is left
for future research.
Internet access almost becomes a necessity in the Internet Age, and many people nowadays seek social interactions
online. In particular, there is an important tendency that online interaction substitutes for face-to-face social interaction.
Therefore, we are aware that the Internet access of households not only facilitates accurate information processing (such
as information from the official news), but also leads to the online social networking which will bring both information from
word-of-mouth and the social multiplier effect of the online social networking. Even though the current paper focuses on the
substitution between Internet access and social interaction in information acquisition, the possible substitution between
Internet access and social interaction in the social multiplier effect is an interesting topic for future research.
The development of the stock market plays an important role in economic development. The access to stock market infor-
mation—as the most important prerequisite for participating in the stock market—relies not only on the technological chan-
nel (Internet), but also on the cultural channel based on social interaction. Both of them require indispensable considerations
for establishing a developed and efficient stock market in China. Actually, the information delivered by the Internet might be
quite different from the information spread by social interactions in precision and scope. Hence, our findings of their sub-
stitution relationship might have practical implications. For example, the information spread by social interactions is usually
biased and inaccurate (Shiller, 1984, 1990; Kaustia and Knupfer, 2012) and hence might even result in heterogeneous beliefs
among investors, as well as asset price bubbles (Hong and Stein, 2007). In comparison, the information transmitted by the
Internet is more accurate and reliable. The influence of information channels on stock market participation also varies with
households groups. So the regulators and policy makers can exert influence on different groups of individual investors
accordingly by deliberately choosing to disclose information through different information channel.
References
Abel, Andrew B., 1990. Asset prices under habit formation and catching up with the joneses. American Economic Review 80, 38–42.
Ai, Chunrong, Norton, Edward, 2003. The interaction term in logit and probit models. Economic Letters 80, 123–129.
Becker, Gary S., 1991. A note on restaurant pricing and other examples of social influences on price. Journal of Political Economy 99, 1109–1116.
Bernheim, B. Douglas, 1994. A theory of conformity. Journal of Political Economy 102, 841–877.
Bernheim, B. Douglas, Garrett, Daniel M., 2003. The effects of financial education in the workplace: evidence from a survey of households. Journal of Public
Economics 87, 1487–1519.
Bogan, Vicki, 2008. Stock market participation and the internet. Journal of Financial and Quantitative Analysis 43, 191–212.
Brown, Jeffery R., Ivkovic, Zoran, Paul, Smith, Weisbenner, Scott, 2008. Neighbors matter: causal community effects and stock market participation. Journal
of Finance 63, 1509–1531.
Bursztyn, Leonardo, Ederer, Florian, Ferman, Bruno, Yuchtman, Noam, 2014. Understanding mechanisms underlying peer effects: evidence from a field
experiment on financial decisions. Econometrica 82, 1273–1301.
Campbell, John Y., Cochrane, John H., 1999. By force of habit: a consumption-based explanation of aggregate stock market behavior. Journal of Political
Economy 107, 205–251.
Cohen, Lauren, Frazzini, Andrea, Malloy, Christopher, 2008. The small world of investing: board connections and mutual fund returns. Journal of Political
Economy 116, 951–979.
Demarzo, Peter M., Kaniel, Ron, Kremer, Ilan, 2004. Diversification as a public good: community effects in portfolio choice. Journal of Finance 59, 1677–
1715.
Durlauf, Steven N., Fafchamps, Marcel, 2004. Social capital. In: Aghion, Philippe, Durlauf, Steven (Eds.), Handbook of Economic Growth, vol. 1. Elsevier, pp.
1639–1699.
El-Attar, Mayssun, Poschke, Markus, 2011. Trust and the choice between housing and financial assets: evidence from Spanish households. Review of Finance
15, 727–756.
Ellison, Glenn, Fudenberg, Drew, 1995. Word-of-mouth communication and social learning. Quarterly Journal of Economics 110, 93–125.
Gan, Li, Yin, Zhchao, Jia, Nan, Shu, Xu, Ma, Shuang, Zheng, Lu, 2014. Data You Need to Know About China: Research Report of China Household Finance
Survey. Springer-Verlag, Berlin.
Georgarakos, Dimitris, Pasini, Giacomo, 2011. Trust, sociability and stock market participation. Review of Finance 15, 693–725.
Glaeser, Edward, Bruce, Sacerdote, Scheinkman, Jose, 1996. Crime and social interactions. Quarterly Journal of Economics 111, 507–548.
Glaser Markus, Klos Alexander, 2013. Causal Evidence on Internet Use and Stock Market Participation. University of Munich Working Paper.
Granovetter, Mark., 1973. The strength of weak ties. American Journal of Sociology 78, 1360–1380.
Grossman, Sanford J., Stiglitz, Joseph E., 1980. On the impossibility of informationally efficient markets. American Economic Review 70, 393–407.
Guiso, Luigi, Sapienza, Paola, Zingales, Luigi, 2003. People’s opium: the economic effects of religion. Journal of Monetary Economics 50, 225–282.
Guiso, Luigi, Sapienza, Paola, Zingales, Luigi, 2004. The role of social capital in financial development. American Economic Review 94, 526–556.
Guiso, Luigi, Sapienza, Paola, Zingales, Luigi, 2006. Does culture affect economic outcomes? Journal of Economic Perspectives 20, 23–48.
Guiso, Luigi, Sapienza, Paola, Zingales, Luigi, 2008. Trusting the stock market. Journal of Finance 63, 2557–2600.
Hong, Harrison, Stein, Jeremy C., 2007. Disagreement and the stock market. Journal of Economic Perspectives 21, 109–128.
Hong, Harrison, Kubik, Jeffrey D., Stein, Jeremy C., 2004. Social interaction and stock-market participation. Journal of Finance 59, 137–163.
Ivkovic, Zoran, Weisbenner, Scott, 2007. Information diffusion effects in individual investors’ common stock purchases: covet thy neighbors’ investment
choices. Review of Financial Studies 20, 1327–1357.
Kahneman, Daniel, Tversky, Amos, 1979. Prospect theory: an analysis of decision under risk. Econometrica 47, 263–292.
Katz, Elihu, Lazarsfeld, Paul, 1955. Personal Influence: The Part Played by People in the Flow of Mass Communication. Free Press, New York.
Kaustia, Markku, Knupfer, Samuli, 2012. Peer performance and stock market entry. Journal of Financial Economics 104, 321–338.
Levine, Ross, 1997. Financial development and economic growth: views and agenda. Journal of Economic Literature 35, 688–726.
Levine, Ross, 2005. Finance and Growth: Theory and Evidence. In: Aghion, Philippe, Durlauf, Steven (Eds.), Handbook of Economic Growth. Elsevier, The
Netherlands.
Li, Tao, 2006a. Social interaction, trust, and stock market participation. Economic Research Journal (Jing Ji Yan Jiu), 1.
Li, Tao, 2006b. Social interaction and investment choice. Economic Research Journal (Jing Ji Yan Jiu), 8.
Li, Geng, 2014. Information sharing and stock market participation: evidence from extended families. Review of Economics and Statistics 96, 151–160.
Li, Tao, Guo, Jie, 2009. Risk attitudes and stock investment. Economic Research Journal (Jing Ji Yan Jiu), 2.
Liu Yu-Jane, Meng Juanjuan You Wei, Zhao Longkai, 2013. Word-of-Mouth Communication, Observational Learning, and Stock Market Participation.
Working Paper, Peking University.
Ma, Guangrong, Yang, Enyan, 2011. Social networks, informal finance and entrepreneurship. Economic Research Journal (Jing Ji Yan Jiu), 11.
Mankiw, Gregory, Zeldes, Stephen P., 1991. The consumption of stockholders and nonstockholders. Journal of Financial Economics 29, 97–112.
Manski, Charles F., 1993. Identification of exogenous social effects: the reflection problem. Review of Economic Studies 60, 531–542.
Manski, Charles F., 2000. Economic analysis of social interactions. Journal of Economic Perspectives 14, 115–136.
Mehra, Rajnish, Prescott, Edward C., 1985. The equity premium: a puzzle. Journal of Monetary Economics 15, 145–161.
Olken, Benjamin, 2009. Do television and radio destroy social capital? Evidence from Indonesian villages. American Economic Journal: Applied Economics 1,
1–33.
Pool Veronika K., Noah Stoffman, Yonker Scott E., The people in your neighborhood: social interactions and mutual fund portfolio. Journal of Finance
(forthcoming).
Rogers, William, 1994. Regression standard errors in clustered samples. Stata Technical Bulletin 13, 19–23.
Scheinkman, Jose, Xiong, Wei, 2004. Heterogeneous beliefs, speculation and trading in financial markets. In: Paris-Princeton Lectures on Mathematical
Finance. Lecture Notes in Mathematics, vol. 47. Springer-Verlag, Berlin.
Shiller, Robert, 1984. Stock prices and social dynamics. Brookings Papers on Economic Activity 2, 457–498.
Shiller, Robert, 1989. Survey evidence on diffusion of interest and information among investors. Journal of Economic Behavior and Organization 12, 47–66.
Shiller, Robert, 1990. Market volatility and investor behavior. American Economic Review 80, 58–62.
Sims, Christopher, 2003. Implications of rational inattention. Journal of Monetary Economics 50, 665–690.
Veldkamp, Laura, 2011. Information Choice in Macroeconomics and Finance. Princeton University Press, Princeton.
Vissing-Jorgensen, Annette, 2002. Limited asset market participation and the elasticity of intertermporal substitution. Journal of Political Economy 110,
825–853.
Vissing-Jorgensen Annette, 2003. Perspectives on behavioral finance: does ‘‘irrationality’’ disappear with wealth? Evidence from expectations and actions.
NBER Macroeconomics Annual.
Yang, Rudai, Binkai, Chen, Shi’e, Zhu, 2011. The credit behavior of rural households from the perspective of social networks. Economic Research Journal (Jing
Ji Yan Jiu), 11.
Zellner, Arnold, 1962. An efficient method of estimating seemingly unrelated regressions and tests for aggregation bias. Journal of the American statistical
Association 57 (298), 348–368.
Zhou, Mingshan, Lei, Sun, Yuzhen, Liu, 2011. Social interaction, attention on relative wealth and stock market participation. Journal of Financial Research
(Jin Rong Yan Jiu), 368.

Social 1-1

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Social 1-1

Uploaded by

Copyright:

Available Formats

Journal of Comparative Economics xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

Journal of Comparative Economics

Social interaction, Internet access and stock market

3. Data and hypotheses

3.1. Data description

Variable N Mean Median St. dev. Min Max Explanation

(continued on next page)

Variable N Mean Median St. dev. Min Max Explanation

3.2. Theoretical hypotheses

P. Liang, S. Guo / Journal of Comparative Economics xxx (2015) xxx–xxx

4.1. Descriptive statistics

Stock Market Participation Rate

across income across education across Internet access

Stock Market Participation Rate

across cashgift expenditures across communication expenditures across respond rates

4.2. Informational effect of social interactions

Cash-gift Communication Response

Cash-gift Communication Response

4.3. Social multiplier effect of social interactions

4.4. Analyses of Sub-samples

You might also like