CLAN CULTURE AND FAMILY OWNERSHIP IN CHINA...clan culture fosters high trust in family members but low trust in outsiders, concentration of family ownership becomes an efficient solution

CLAN CULTURE AND FAMILY OWNERSHIP IN CHINA*

Jiameng Cheng, Yanke Dai, Shu Lin, Haichun Ye†

Abstract

This study explores a cultural determinant of family ownership concentration in China’s

private business. Exploiting regional variations in local clan culture intensity, we find robust

evidence that stronger clan culture influences are associated with higher family ownership

concentration. Relying on the existence of movers, we separate the impact of inherited clan

culture from that of external environmental factors and show that the former remains

significantly positive. Our IV estimates further establish the causal effect of Chinese clan

culture on family ownership concentration. In the end, we also provide additional evidence

for underlying mechanisms.

Keywords: clan culture; family ownership; private firms; China

JEL code: G32, P34, N25, Z1

* The authors thank Ying Bai, Ruixue Jia, and David Reeb and seminar participants at the

Chinese University of Hong Kong, the National University of Singapore, and Nanyang

Technological University for comments and suggestions. † Corresponding author: Haichun Ye Mailing address: School of Management and

Economics, The Chinese University of Hong Kong (Shenzhen), 2001 Longxiang Boulevard,

Longgang District, Shenzhen, China. Telephone: +86 755 2351 8819. Email address:

[email protected].

1

I. Introduction

Privately-owned business in China was prohibited during the planned economy era, yet it

has experienced explosive growth since the legalization of private ownership in the 1980s.

Over the past four decades, the thriving private sector has contributed most to China’s fast

growth (e.g., Allen, Qian, and Qian, 2005; Dollar and Wei, 2007; Song, Storesletten, and

Zilibotti, 2011). It now contributes to more than 50% of China’s total tax revenues and

accounts for over 60% of its gross domestic product (GDP).

An interesting feature of China’s private sector is that a vast majority of this sector

comprises family-controlled business (e.g., Fukuyama, 1995; Cai, Li, Park, and Zhou, 2013).

Concentration of family ownership and family management are quite common in Chinese

privately-owned firms. According to the data from the Chinese Private Enterprise Survey

(CPES) conduced in 2010, total equity held by firm owner and his/her family members

averaged around 82.3%, and over 92% of surveyed firms were managed by their founding

families. Even large listed private firms in China are mostly controlled by families. In 2016,

84.2% of Chinese-listed private firms were family controlled.1

Despite a high degree of family centeredness on average, there are large regional

variations in family ownership in China. For example, according to the CPES2010 data, the

average share of family ownership in Wuwei city of Gansu province is only 49%, whereas

those in Meizhou city of Guangdong province and Shaoxing city of Zhejiang province are as

high as 93%.

Yet another interesting fact is that family centeredness exists not only in Mainland China

1 Controller information for listed nonstate-owned firms is obtained from the China Stock Market and Accounting Research

(CSMAR) database. Following La Porta, Lopez-de-Silanes, and Shleifer (1999), a 20% ownership threshold is employed

when defining family-controlled firms, and state-owned firms were excluded from the calculation.

2

but is also prevalent in other Chinese societies, such as Hong Kong and Taiwan, and even in

firms established by overseas Chinese but located in non-Chinese societies with sharp

variations in legal and financial institutions (e.g., Redding, 1990; Fukuyama, 1995;

Claessens, Djankov, and Lang, 2000). This similarity points to some potential common

underlying factor that has a persistent impact on the ownership structure of Chinese private

business.

A natural candidate for such a determinant is culture. Recently, a growing literature

identifies culture as an important determinant of various economic outcomes (e.g., Guiso,

Sapienza, and Zingales, 2004a, 2006, 2008, 2009; Nunn, 2008; Nunn and Wantchekon, 2011;

Alesina and Giuliano, 2015; Zingales, 2015).2 There has also been an argument that

concentration of family ownership may be attributable to cultural norms (e.g., Weber, 1940;

Banfield, 1958; Bertrand and Schoar, 2006). Some studies in social sciences have also

pointed out that a strong family orientation in Chinese private business is likely to have its

culture roots (e.g., Fukuyama, 1995; Perkins, 2000). Yet, so far, very few studies have been

done to formally examine the causal linkage between culture and family ownership,

especially at the micro level.

The goal of this study is therefore to fill this gap by empirically investigating the cultural

impacts on family ownership in China using detailed firm-level data. We focus on China for

two primary reasons. First, that China is well known for its familial tradition and its clan

culture, makes it an ideal case to study this subject. Second, although existing studies have

investigated firm ownership structures in many other parts of the world, little attention has

2 See Guiso et al. (2006) and Alesina and Giuliano (2015) for reviews of this literature.

3

been paid to China, the second largest economy in the world, where private business accounts

for a large share of its economy and plays an important role in shaping its economic

development. Understanding the ownership structure of Chinese private business remains an

important yet under-researched issue.

In particular, we examine the impacts of a symbolic culture in China – the Chinese clan

culture – on family ownership. Originated from the Song Dynasty, clans are kinship-based

organizations made up of component families which trace their patrilineal descent from a

common ancestor (e.g., Fei, 1946; Freedman, 1958; Liu, 1959, Feng, 1994, 2013; Xu, 1995,

2012; Greif and Tabellini, 2010, 2017).3 Clan organizations used to be the dominant social

structure in pre-modern China. Although formal clan organizations were abolished after the

Communist Party took power in China, the Chinese clan culture continues to persist and exert

influence on individuals’ behavior in today’s China.

To empirically identify the effects of clan culture on family ownership, we construct

prefecture-level measures of local clan culture intensity using hand-collected genealogy data.

Based on the firm-level data from the CPES, we find a positive and significant impact of

local clan culture on family ownership concentration. This finding is robust to alternative

measures of local clan culture intensity and family ownership as well as different subsamples

used in estimation. The estimated marginal effect of local clan culture intensity is also

economically significant. A one-standard-deviation increase in the local clan culture intensity

measure raises the share held by owner’s family by 8.94 percentage points, equivalent to an

increase of about 10.87% relative to the sample average family ownership share.

3 See Section II for more background information on clan organizations and cultural norms in China.

4

To address the potential concern of omitted variables, we conduct three additional sets of

exercises. First, we control for additional factors that can be potentially correlated with both

local clan culture strength and family ownership concentration. Second, we search for

complementary evidence using Chinese listed family firm data which contain owners’

birthplace information at the prefecture level. Specifically, we resort to the presence of

movers (i.e., owners whose birthplaces differ from firm locations) in the data and include

firm location prefecture fixed effects to separate the inherited component of clan culture at

owner’s birthplace (i.e., clan culture of origin) from confounding environmental factors. We

show that clan culture of origin has a positive and significant effect on the concentration of

family ownership even after controlling for firm location prefecture fixed effects.

Our final strategy of causality identification is to employ an instrumental variable (IV)

approach. The IV we use is the minimum distance to the two historical Confucian academies

established in the Southern Song Dynasty, the Kaoting Academy (Kaoting Shuyuan) and the

Xiangshan Academy (Xiangshan Shuyuan). These two academies were respectively founded

by two prominent neo-Confucian scholars, Zhu Xi and Lu Jiuyuan, and were their primary

places of holding lectures on clan-related doctrines. As we shall discuss in more details in

Section 5.3, both Zhu Xi and Lu Jiuyuan played important roles in the historical development

of clan organizations in China. Given that prefectures closer to these two academies were

more exposed to Zhu’s and Lu’s influences historically, and through their persistence, we

expect stronger local clan culture in these prefectures today. Our relevance test for the IV

finds supportive evidence for this link in the data. We also conduct various exclusion tests to

rule out other potential direct channels through which the IV affects family ownership. The

IV estimates from both the CPES and the listed family firm data confirm that local clan

5

culture has a positive causal effect on the concentration of family ownership in China’s

private business.

Why would strong Chinese clan culture lead to high concentration of family ownership?

We propose three possible explanations. The first one is the short-radius trust attitude. As the

clan culture fosters high trust in family members but low trust in outsiders, concentration of

family ownership becomes an efficient solution to agency problems in corporate governance.

The second explanation is related to the tradition of common property ownership cherished

by clan organizations historically. To the extent that this cultural norm facilitates financial

resources pooling among family members, the need for external capital in firm’s first

establishment and subsequent operation is reduced. The third possible reason is amenity

potential. Under the influence of strong local clan culture, firm owners tend to have high non-

pecuniary private benefits of family control. They may either directly derive utility from

keeping firms in family hands or using their firms for private benefits, such as fulfilling the

obligation of providing employment for family members.

While our main objective is not to further distinguish these finer mechanisms, we do

provide suggestive evidence consistent with the above three channels. First, for the short-

radius trust channel, we show that individuals from prefectures with stronger clan culture do

exhibit a higher degree of trust asymmetry (i.e., high trust in family members but low trust in

strangers). We also find that local clan culture has a bigger impact on family ownership in

weaker contracting and legal environments or more volatile business sectors, where trust is

more needed. Moreover, there is also evidence that owners are less likely to invest in other

firms, probably due to lack of trust. Second, to seek for evidence consistent with the financial

resources pooling channel, we decompose the total family share into owner’s share and that

6

held by other family members and show that the relative share held by other family members

increases with the strength of local clan culture. We also explore the role of local financial

institutions and find a more pronounced effect of clan culture in prefectures with weaker

financial institutions where family funding is more crucial to firms. Finally, we also provide

some supportive evidence for the amenity potential channel by showing that, in prefectures

with stronger local clan culture, owners feel more obligated to provide employment for their

family members.

Our work contributes to the relevant literature in the following aspects. First, our study

provides a nice complement to the growing literature on cultural traits and economic

outcomes. Existing contributions have documented culture as an important determinant of

various economic phenomena (e.g., Guiso et al., 2004a, 2008, 2009; Nunn, 2008; Nunn and

Wantchekon, 2011; Alesina and Giuliano, 2015; Zingales, 2015). We add to this literature by

exploring the unique clan culture in China and its impact on firm ownership structure. Our

findings are consistent with the main theme of this recent “cultural revolution” in economics

and finance literature that cultural traits have persistent and significant impacts on economic

outcomes.

Second, our study also contributes to the broad literature on firm ownership structure,

and, in particular, family ownership. Previous studies have analyzed the effects of family

ownership on firm’s performance and information disclosure (Anderson and Reeb, 2003;

Anderson, Duru, and Reeb, 2009; Anderson, Reeb, and Zhao, 2012; Perez-Gonzalez, 2006;

Lins, Volpin, and Wagner, 2013) and explored determinants of family ownership, such as

institutions (Demsetz and Lehn, 1985; Burkart, Panunzi, and Shleifer, 2003; Bertrand and

Schoar, 2006). We add to this literature by exploring a cultural determinant of family

7

ownership and establishing the causal linkage. The evidence from Chinese private business

also lends support to theories (e.g., Demsetz and Lehn, 1985; Ang, Cole, and Lin, 2000;

Burkart et al., 2003; Bertrand and Schoar, 2006) that emphasize cultural norms and agency-

costs related factors as key determinants of family ownership.

Last, our study is also related to the literature on the ownership structure of Chinese

enterprises. Existing work in this literature has examined the effects of ownership structure

on Chinese firms’ efficiency (e.g., Allen et al., 2005; Dollar and Wei, 2007; Song et al., 2011)

and organization designs (e.g., Cai et al., 2013), and investigated institutional determinants of

firm ownership (e.g., Che and Qian, 1998; Bai, Li, Tao, and Wang, 2000; Song and Hsieh,

2015) in China. Previously, Fukuyama (1995) and Perkins (2000) have hypothesized that a

strong family orientation in Chinese business is likely due to cultural reasons. We contribute

to this literature by showing that clan culture plays a crucial role in determining the

ownership structure for Chinese private firms.

The remainder of this paper is organized as follows. In Section II, we provide some

background information on clan organizations and cultural norms in China and also discuss

our conceptual framework. Section III describes the data used in our analyses. Section IV

specifies our empirical models and presents our main results. In Section V, we conduct

additional exercises to address the concern of omitted variables and also report our

instrumental variable regression results. Section VI provides further empirical evidence to

shed light on the underlying mechanisms. Concluding remarks are offered in Section VII.

II. Background Information and Conceptual Framework

II.A. Historical background of Chinese clan organizations

8

Clans (tsung tsu) are distinct epitomizing social structures in imperial China (e.g., Greif

and Tabellini, 2010). A Chinese clan is a consolidating patrilineal group made up of

component families that trace their patrilineal descent from a common ancestor. Publications

of genealogy books and formations of kinship networks in China can at least be traced back

to the Zhou Dynasty (1046 – 256 BC). In ancient China, only a handful of privileged noble

families could have their kinship organizations (shih). During the period from the Eastern

Han Dynasty to the Tang Dynasty, power aristocratic families were given the legal status of

“esteemed clans” (men fa) with special privileges in tax exemptions and civil service

appointments, while common people were not entitled to kinship practices at all.

“Modern” clan organizations originated in the Song Dynasty (960 – 1279 AD). Starting

from the Song Dynasty, government official selection in China switched from hereditary

succession to an imperial examination system, which significantly weakened the power of

aristocrats and gave rise to a gentry/scholar class. To preserve wealth and privileges, the

newly-emerged gentry class advocated the formation of clan groups among ordinary people

and started to form their own clans. Meanwhile, the popularization of “modern” clan

organizations also benefited greatly from the neo-Confucianism ideology of moral and ethical

teaching. Neo-Confucian scholars not only provided theoretical basis for the plebeianization

of clans but also participated actively in designing clan organization structures and rules. (Xu,

1995). For example, in his influential work Jiali (the Family Rituals), the famous neo-

Confucian scholar, Zhu Xi, offered detailed instructions on the designs of ancestral halls and

the establishment of clan ritual land from which the proceeds could be used for ancestral

offerings. Another prominent scholar, Lu, Jiuyuan, formed a large communal family together

with his brothers. They also published an influential book on clan management and

9

governance, LuShi JiaZhi (Management of the Lu Family), based on their clannish practices.

From then on, clan organizations gradually became the dominant social structure in imperial

China and exerted a pervasive and long-lasting influence on Chinese society.

In terms of clans’ geographical distribution, it is well-documented in both the history

and the anthropology literature that (modern) clan institutions were particularly strong in the

south-east part of China, well-developed in the central part, but rather weak in the north and

west parts (e.g., Lang, 1946; Hu, 1948; Freedman, 1958; Liu,1959; Feng, 1994, 2013; Xu,

1995, 2012).

II.B. Characteristics of Chinese clan organizations

While clan organizations in post-Song China varied in sizes and practice details, they

were formed and governed by the common principle of group cohesion and thus shared

several key characteristics. First, the member families of a clan either lived in the same

community or among several nearby communities in the same region. Second, clans held

common properties and organized routine group activities such as ancestor worshiping. Clans

usually pooled resources from members to establish a variety of common properties, mainly

in the form of land, and used the yields to provide different types of public goods. For

instance, charity land was used to provide support for the poor, ritual land was used for

ancestor worshiping and offerings, and education land was used for sponsoring children’s

education.

Third, compilation of genealogies is another key feature of Chinese clans. A genealogy

book typically records all descents from the apical, or common, ancestor and describes a

clan’s honorable descents and events. In addition, it also presents and advocates clan rules

and codes of conduct, which, apart from advocating merits like hardworking and integrity,

10

often value clannish solidarity. Combined with common property ownership and ancestors

worshiping, genealogy compilation is an important means for clans to reinforce

consanguinity ties among clan members and help to unite them (jing zong shou zu) (e.g.,

Feng, 1994, 2013; Xu, 1995, 2012).

Finally, clans had their own internal governance structures. Most clan groups had clan

leaders presiding over clan operation and management. There also existed a set of internal

rules to regulate clan members’ behavior so as to reinforce the principle of internal cohesion

and loyalty among clan members (e.g., Liu, 1959).

II.C. Clan culture in today’s China

After the Chinese Communist party took power in 1949 and especially during the

Cultural Revolution, clan organizations lost their legal status and were largely destroyed in

Mainland China. Nonetheless, the influences of clan culture, such as norms, beliefs, and

values associated with clan organizations, continue to persist. As a matter of fact, clan-related

cultural activities such as compiling genealogy books and building ancestral halls reemerged

in China after its economic reform in 1978. In our empirical analysis below, we will use

hand-collected genealogy data to construct measures of clan culture intensity at the Chinese

prefecture city level and employ these regional variations in clan culture strength to identify

its causal effect on family ownership.

II.D. Conceptual framework

Why and how would the strength of local clan culture affect the concentration of family

ownership in China’s private business? In this section, we propose three possible

explanations based on some special features of clan culture and existing theories of family

ownership (e.g., Demsetz and Lehn, 1985; Ang et al., 2000; Burkart et al., 2003; Bertrand and

11

Schoar, 2006; Franks, Mayer, and Rossi, 2009). The first explanation is a short-radius trust

attitude. According to agency-cost based theories, family ownership can be affected by

individuals’ trust attitudes. In places where cultures foster high trust in family members but

low trust in outsiders, high concentration of family ownership becomes an efficient solution

to agency problems.

A distinct feature of clan culture is the formation of short-radius trust attitude (e.g.,

Redding, 1990; Fukuyama, 1995; Feng, 1994, 2013; Xu, 1995, 2012; Greif and Tabellini,

2017). Chinese clans were built upon the principles of cohesion among group members but

competition with outsiders. To promote group cohesion, clans usually set out rules that

required members to live in harmony with one another and to help members in need to

strengthen mutual trust. To fuel competition with outsiders, clans also imposed stringent

regulations on members’ external social activities. Members were taught and required not to

easily trust outsiders and to be cautious when socializing with strangers (e.g., Liu,1959; Feng,

1994, 2013; Xu, 1995, 2012). These clan rules and regulations predisposed clan members to a

short-radius trust attitude, which amplifies the agency costs of outsider control but reduces

the cost of family control. Thanks to this trust asymmetry embedded in the Chinese clan

culture, family ownership can be an effective second-best solution to reducing agency costs.

The second explanation is closely linked to the clan tradition of resources pooling. As

discussed in Section II.B, common property ownership is an important characteristic of

Chinese clan organizations. Clans had a long tradition of resource pooling and sharing among

members. Moreover, traditional clan rules also discouraged selling properties to people

outside of the clan. Selling common properties to outsides were strictly prohibited in general.

Even for privately-owned properties, owners were required to find buyers first inside the

12

clan. Selling to outsiders was permitted only if no other clan members were interested in

purchasing. To the extent this cultural norm helps facilitate financial resource pooling among

family members, the need for external capital is reduced.

The third explanation is amenity potential. In places where cultures foster familism, firm

owners may have high non-pecuniary private benefits of family control. Owners may directly

derive utility from keeping the firm in family hands or enjoy private benefits of control, such

as providing employment for family members. Because clans were historically organized

along patrilineal descent, the key element of this culture is its emphasis on family values and

obligations and loyalty to the family. Other things constant, firm owners subject to stronger

influences of clan culture may have a higher amenity potential of family control and opt for

more concentrated family ownership.

Taken together, the three channels above all predict a positive effect of local clan culture

on family ownership. Although we have listed these channels separately, we would like to

stress that these mechanisms are likely to be intertwined and are by no means mutually

exclusive. The main objective of this study is not to distinguish further these finer channels.

Rather, our intended contribution is to empirically identify the causal effect of the unique clan

culture on family ownership in China. Nonetheless, we do conduct some exercises in Section

VI to shed light on the underlying mechanisms. Evidence that is consistent with

characteristics of clan culture and theories of family ownership helps us to further establish

causality.

III. Data

III.A. Prefecture-level data on clan culture intensity

13

Given the prominent role of genealogies in the Chinese clan culture, we follow Greif and

Tabellini (2017) and construct measures of local clan culture intensity at the prefecture level

based on the total number of genealogies complied. The genealogy data are hand-collected

from the printed registry “Comprehensive Catalogue of the Chinese Genealogy” (Wang,

2008), which keeps by far the most comprehensive records of all available Chinese

genealogies worldwide.4 In our hand-collected dataset, there are a total of 42,497 entries

spanning 944 years from 1063 to 2007 AD, covering 286 prefecture cities in Mainland China

and 585 Chinese surnames.5 The 585 surnames identified in our genealogy data cover not

only the most commonly used surnames in China but also a substantial amount of rarely used

ones. This surname coverage is quite comprehensive in the sense that over 85% of China’s

1.3 billion population is covered by just 100 of the more popular surnames in use today. Each

entry in our dataset contains information about genealogy records, such as the year and

location of genealogy compilation. We count the total number of genealogies for each

prefecture city, scale it by local population size in 2000 and then use this ratio as our primary

measure of clan culture intensity.

[Figure I about here]

Figure I maps the geographical distribution of this local clan culture intensity measure

across prefecture cities in China. As evident from the figure, clan culture is particularly strong

in southern China but relatively weak in the northern part of China and the western minority

4 The project of compiling the registry “Comprehensive Catalogue of the Chinese Genealogy” was initiated by the Shanghai

Library in 2000. It took nine years to complete and was a joint effort of the National Library of China, the Genealogical

Society of Utah in the United States, 44 provincial and university libraries in China, 614 other genealogy collecting

organizations, and thousands of private genealogy collectors. In a recent study, Greif and Tabellini (2017) also used the

genealogy data from this registry and view it as “the most comprehensive measure regarding the number of clans”. 5 There is a total of 53,944 entries of genealogy records collected in the “Comprehensive Catalogue of the Chinese

Genealogy” (Wang, 2008). After excluding entries with missing information about their locations or compiled in Hong Kong

and Taiwan, there are 42,497 usable entries left in our genealogy dataset.

14

regions. This pattern is in line with the geographical distribution of historical clan

organizations documented in the history and anthropology literature (e.g., Lang, 1946; Hu,

1948; Freedman, 1958; Liu,1959; Feng, 1994, 2013; Xu, 1995, 2012). In addition to this

primary measure, we also consider several alternative measures in Section 4.3.3 to ensure the

robustness of our results. Panel A of Table I contains summary statistics for various measures

of local clan culture intensity. Detailed variable definitions and data sources are provided in

Appendix Table A1.

[Table I about here]

III.B. Data on firm and owner characteristics

III.B.1. Chinese private enterprise survey

Our main firm-level data source is the Chinese Private Enterprise Survey (CPES

hereafter) conducted in 2010. The CPES covers privately-owned firms of various sizes

nationwide and is jointly administered by the All-China Federation of Industry and

Commerce, the State Administration for Industry and Commerce, the China Society of

Private Economy, and the United Front Work Department.

A nice feature of the CPES2010 data is that they contain detailed information about the

shares of firm equity held respectively by owners and their family members6. This allows us

to gain more insights into family ownership structure in these privately-owned firms. Aside

from that, the CPES2010 data also collect a rich set of information about firms’ business

activities and balance sheets, as well as firm owners’ demographic characteristics and family

background. By using each firm’s zip code information, we are able to identify the prefecture

6 In the CPES, family members include parents, spouse, children, siblings and their respective spouses etc.

15

the firm belongs to and match it with our prefecture-level measures of clan culture intensity.

By doing so, there are 2,395 firms for which we have relatively complete information on the

variables of interest. Summary statistics of the main variables from the CPES2010 data are

provided in Panel B of Table I.

A preliminary review of the CPES2010 data shows that the total share of firm equity held

jointly by owner and his/her family members averages around 82.3%, but exhibits substantial

variations (a standard deviation of 27.8%) across regions in China. For example, firm’s

family ownership averages over 93% in both Meizhou city of Guangdong province and

Shaoxing city of Zhejiang province, where the influences of clan culture are particularly

strong. In contrast, average family ownership shares are merely 49% in Wuwei city of Gansu

province and 67% in Qiandong city of Guizhou province, both of which have large ethnic

minority populations and are less exposed to the clan culture embraced by the Han majority.

It is also worth noting that firms in our sample are relatively young with an average age

of 10 years, reflecting the fact that China did not legalize private enterprises until the late

1980s. On average, firms have an employment size of 194 workers. But there is a wide

variation in firm size (a standard deviation of 539 employees), ranging from small-sized

household businesses with less than 10 employees to large-scale companies with more than

2000 employees. With respect to individual-level characteristics, firm owners have an

average age of 47 years old, with the majority (about 87%) being male. Among these

surveyed owners, around 62% have a college or post-graduate degree.

III.B.2. Data on publicly listed family firms

In addition to the CPES data, we also use an unbalanced panel data on China’s publicly

listed family firms obtained from the China Stock Market and Accounting Research

16

(CSMAR) family firm database. While the listed firm data have less detailed information

regarding the composition of family ownership, the key advantage of this dataset is that it

provides owner’s birthplace information at the prefecture level.7 By combining firm location

and owner birthplace information, we are able to identify movers (i.e., owners whose

birthplaces differ from their firm locations) in this sample. We can then resort to the existence

of movers in the sample to further disentangle the effect of the inherited component of clan

culture (i.e., clan culture intensity in owners’ birthplaces, or clan culture of origin) from that

of external environmental factors by controlling for firm location fixed effects in the

regression.

Having cleaned the data following conventional procedures, we obtain an unbalanced

panel of 955 firms over the period 2014 – 2016, for which there is relatively complete

coverage on our interested variables, including firm’s family ownership, balance sheet, and

controlling shareholder’s birthplace as well as other relevant individual-level characteristics.

Panel C of Table I contains the summary statistics for the main variables from this listed

family firm data. Compared to firms included in the CPES2010 data, the listed family firms

are much larger in size, with an average of 3,138 employees, and more mature, with an

average age of 16.2 years. Given the nature of being publicly listed, these family firms tend

to have ownership less concentrated in the controlling shareholder’s family, averaging around

36.2%. Therefore, these supplementary data can also help us shed some light on the impact of

clan culture intensity on large-sized firms with relatively dispersed family ownership.

7 There are some listed family firms with their controlling shareholders’ birthplace information missing in the CSMAR

database. For these firms, we supplement their controlling shareholders’ birthplace information with two additional sources.

One is to extract the first six digits of controlling shareholders’ identification numbers manually from firms’ annual reports

and/or reports of equity changes and use these first six digits to identify their birthplaces at the prefecture level. The other is

to conduct the internet search manually for controlling shareholder s’ birthplaces based on their names and firms’ names.

17

III.C. Other prefecture-level control variables

In this study we also collect data on prefecture’s geographical, economic and political

characteristics from various sources and use them as additional covariates to control for their

confounding effects. First, we obtain each prefecture’s geographical attributes, including

distance to the nearest coastline, longitude and latitude, from the China Historical Geographic

Information System (CHGIS). Second, we include prefecture-level population density and

land taxes per capita in 1820 as proxies for historical economic conditions and GDP per

capita in 2008 as a proxy for current economic condition.8,9 The historical data on prefecture-

level population and land taxes are hand-collected from the Jiaqing chongxiu yitongzhi

(National Gazetteer of Qing Dynasty Recompiled during the Reign of Emperor Jiaqing), a

compendium compiled by government officials between 1820 and 1842 and recording

detailed social and economic data up to 1820. The data on GDP per capita in 2008 is drawn

from the China City Statistical Yearbook. Last, to control for heterogeneity in political

influences across prefectures historically, we also introduce a dummy variable for being a

treaty port as well as a dummy variable for being provincial capital in 1820. The information

on treaty ports is taken from Yan (1955), and that on provincial capitals in 1820 is collected

from the CHGIS data. Summary statistics for the above prefecture-level controls are also

reported in Panel A of Table I.

IV. Empirical Analysis

8 Since the historical variables used in this study were originally recorded at the historical prefecture level, we adjust for

changes in administrative boundaries using the GIS polygon map from the CHGIS. 9 In the CEPS2010 data, survey respondents were asked to provide information about their firms and firm owners in 2009.

Here we use the prefecture-level GDP per capita in 2008 (with one-year lag) as a control for current economic condition.

18

IV.A. Model specification

Given that the dependent variable - the percentage of firm equity held by owner’s family

- is bounded between zero and 100%, we employ a Tobit model to assess the impact of clan

culture intensity on the concentration of family ownership in Chinese privately-owned firms:

𝑦𝑖𝑐∗ = 𝛼 + 𝛽𝐶𝑙𝑎𝑛𝑐 + 𝛾𝑋𝑐 + 𝛿𝑍𝑖 + 𝑢𝑝𝑟𝑜𝑣 + 𝜇𝑖𝑛𝑑 + 𝜖𝑖𝑐,

𝑦𝑖𝑐 = 0 𝑖𝑓 𝑦𝑖𝑐∗ ≤ 0; 𝑦𝑖𝑐 = 𝑦𝑖𝑐

∗ 𝑖𝑓 0 < 𝑦𝑖𝑐∗ < 100; 𝑦𝑖𝑐 = 100 𝑖𝑓 𝑦𝑖𝑐

∗ ≥ 100 (1)

where 𝑦𝑖𝑐∗ is a continuous latent variable for firm 𝑖 in prefecture city 𝑐, 𝑦𝑖𝑐 is the observed

family ownership of this firm, and 𝐶𝑙𝑎𝑛𝑐 is the clan culture intensity measure for prefecture

𝑐. 𝑋𝑐 is an extensive set of prefecture-level controls, and 𝑍𝑖 is a vector of covariates related

to firm and owner characteristics. 𝑢𝑝𝑟𝑜𝑣 and 𝜇𝑖𝑛𝑑 are the province and industry fixed

effects, respectively. The inclusion of the province fixed effects absorbs any variation in clan

culture intensity across provinces, and the remaining variation essentially distinguishes

prefectures within the same province.

Our coefficient of interest here is 𝛽, which captures the impact of prefecture-level clan

culture intensity on family ownership concentration. A positive and statistically significant 𝛽

would be consistent with our conjecture that strong local clan culture influence tends to raise

family ownership in China’s privately-owned firms.

IV.B. Baseline results

Table II reports the results from Equation (1) using the CPES2010 data. Column (1)

estimates a simple bivariate correlation between clan culture intensity and family ownership

controlling for the industry and province fixed effects. Column (2) adds firm size and firm

age as control variables. Column (3) controls for firm owners’ demographic characteristics,

19

including owner age, gender, and a dummy for having a college degree or above. To account

for the impact of household size on family ownership, we also include owner’s household

size as a control in Column (3).

In the next three columns, we further include three sets of prefecture characteristics, one

at a time, to control for their potential confounding effects. Column (4) adds a set of

geographical variables, including prefecture’s log distance to the nearest coastline, longitude,

and latitude. In Column (5), we include log population density in 1820, log land taxes per

capita in 1820, and log per capita GDP in 2008 to control for historical and current economic

factors at the prefecture level. In Column (6), we include a historical treaty port dummy and a

dummy for being a provincial capital in 1820 as proxies for prefecture’s political status

historically. Column (7) combines all above controls for firm, owner and prefecture

characteristics and becomes our baseline model specification.

The results presented in Table II suggest a positive association between local clan culture

strength and family ownership concentration. In all regressions, the estimated marginal

effects of local clan culture intensity on family ownership are positive and statistically

significant at the 1% level. The estimated effects are also economically sizeable. Take the

baseline estimates in Column (7) for example. A one-standard-deviation increase in the clan

culture intensity measure can raise family ownership share by 8.94 percentage points, about

10.87% increase relative to the sample mean of family ownership share.

[Table II about here]

IV.C. Robustness checks

In this subsection, we conduct a battery of robustness checks using alternative measures

of family ownership, different subsamples, and alternative measures of clan culture intensity.

20

IV.C.1. Alternative measures of family ownership

Table III examines whether our main results are sensitive to alternative measures of

family ownership. First, we follow Demsetz and Lehn (1985) to convert the percentage of

shares held by owner’s family, a bounded variable, to an unbounded one by applying a

logistic transformation. We then re-estimate the regression model using ordinary least squares

(OLS). As shown in Column (1), using this log-transformed measure of family ownership

concentration does not affect our main results as the estimated coefficient on the clan culture

intensity variable remains positive and statistically significant.

Although the CPES2010 data are cross-sectional in nature, they also contain information

about the ownership structure when a firm was first established. This allows us to investigate

further how clan culture affects the concentration of family ownership at the start-up stage as

well as and its subsequent evolution. Column (2) uses the family ownership percentage at the

start-up stage as the dependent variable, and Column (3) examines the change in family

ownership concentration between the first year of establishment and the survey year. The clan

culture intensity variable carries a positive sign and is statistically significant in both

columns, indicating that higher local clan culture intensity is associated with not only higher

degree of concentration of family ownership at firm’s start-up stage but also a larger increase

in family ownership afterwards.

[Table III about here]

While the above exercises help us gain some insights into the impact of local clan culture

on firm’s family ownership in the past and present, we are also interested in its effect on

future family ownership. To this end, we turn to the CPES2008 data in which two questions

were asked regarding owner’ s perception of potential ownership diversification in the future:

21

(1) “Do you plan to issue equity to other firms or natural persons?”, and (2) “Do you wish to

transform the firm into a joint-stock enterprise?” For each question, we create a binary

indicator that takes the value of one when a respondent answered yes and zero otherwise. We

then estimate a Logit model using the two binary indicators as the dependent variables in

Columns (4) and (5), respectively. The estimated marginal effects of clan culture intensity are

negative and statistically significant in both columns. That is to say, in places with stronger

local clan culture influences, owners are less likely to reduce the concentration of family

ownership either by issuing equity to outsiders or transforming firms into joint-stock

enterprises.

IV.C.2. Different subsamples

Our second set of robustness checks is to see whether the main results are sensitive to

different subsamples used in estimation. Table IV summarizes the estimation results from five

exercises. In the first two columns, we verify that our results are not driven by potential

outliers in our local clan culture intensity measure. Column (1) removes prefectures with no

genealogy books. Column (2) excludes prefectures with clan culture intensity belonging to

the top 5% or the bottom 5% of the distribution. Since clans were historically more

concentrated in rural areas, we exclude large first-tier cities and provincial capital cities from

our sample in Column (3) to see whether our results still hold. To address the concern of

uneven distribution of firms across prefectures, we include in the sample only prefectures that

have more than 10 firms in the survey and report the results in Column (4).10 Because the

northern part of China has historically experienced more wars and riots, genealogies

10 While not reported, we also estimated a weighted Tobit regression in which the weights are set equal to the inverse of the

square root of the number of surveyed firms for each prefecture. The weighting scheme is meant to reduce the dominance of

large prefectures in the estimation results. The results from this weighting regression are very similar.

22

complied in these places were more likely to have been destroyed, leading to a less precise

measure of local clan culture. As an additional robustness check, we exclude prefectures in

the northern provinces from our estimation sample and report the results in Column (5).

Our results remain intact no matter which subsample is used in estimation. The estimated

coefficients on the clan culture intensity measure are always found to be positive and

statistically significant at the 1% level.

[Table IV about here]

IV.C.3. Alternative measures of clan culture intensity

Finally, we check whether our results are sensitive to different measures of local clan

culture intensity and report the results in Table V. In Column (1), we scale the number of

genealogies by land area.11 To facilitate interpretation and comparison, we normalize this

ratio by subtracting its mean and divided by its standard deviation. Using this alternative

measure of local clan culture intensity does not alter our main finding as we continue to find

a positive and significant effect of local clan culture on family ownership concentration.

A potential concern over our local clan culture measure is that it may be affected by the

number of surnames within a prefecture. Since clans were patrilineal groups, conditional on

population size, areas with more surnames may have more genealogy books. To address this

concern, we make use of China’s inter-census population survey (also known as the 1%

population sample survey) conducted in 2005 to construct the total number of surnames and

an Herfindahl-Hirschman Index of surnames for each prefecture and include them as

additional controls in Columns (2) and (3), respectively. It turns out that controlling for

11 While not reported for the sake of brevity, we also scaled the number of genealogies by prefecture-level population in

1820 and used it as an alternative measure. We obtained very similar results.

23

surname does not affect our results either.

[Table V about here]

Another potential complication associated with our clan culture intensity measure is that

it is constructed using existing stock of genealogies and may be subject to survivorship bias.

In Column (4) of Table IV, we have already partially addressed this issue by confirming the

robustness of our results to excluding prefectures in northern China, where warfare and

rebellions were more frequent historically relative to their southern counterparts and the

survivorship bias could potentially arise. Nonetheless, we are still concerned about the

potential survivorship bias due to “the Campaign to Destroy the Four Olds” (po si jiu), which

occurred during the Cultural Revolution and caused a massive destruction of cultural related

objects, including genealogies, nation-wide.12 If the destruction was random across

prefectures as argued by Cao (1991), the survivorship bias will unlikely to be an issue in our

study.

If the variation in genealogy destruction across prefecture was nonrandom, we need

further distinguish between two possible scenarios. One is that prefectures with stronger clan

culture turned out to have more genealogies preserved. This could happen when stronger

local clan culture acted as a larger local resistance force against the Campaign, leading to

better protection of genealogies (i.e., the “resistance” scenario). In this case, our estimates can

be considered as a lower bound for the true effect of clan culture. On the other hand, if the

Campaign had specifically targeted prefectures with stronger clan culture historically and

caused more severe damages to genealogies in those prefectures, the stock of surviving

12 The Four Olds refer to Old Customs, Old Culture, Old Habits, and Old Ideas.

24

genealogies would be a problematic proxy for local clan culture strength. Consequently, our

estimates would overstate the importance of clan culture influence. This “targeting”

argument, however, is not likely to be tenable. The “Campaign to Destroy the Four Olds” was

launched as a nation-wide mass movement, and no historical records have documented that

the Campaign targeted certain particular areas. Especially since the goal of the Campaign was

to eliminate all traces of pre-communism Chinese culture, not just clan culture, it’s unlikely

that areas with strong local clan culture were the specific targets of the Campaign. Moreover,

started in August 1966, the Campaign was initially confined to Beijing and subsequently

spread to Tianjin, Shanghai, and other major cities. Compared to major cities, non-major

cities and rural areas, where clan culture was more prevalent, suffered relatively less in

general.13

That said, we take three approaches to deal with the potential survivorship bias. Our first

strategy is to use distance to Beijing (national center of the Campaign) and distance to

respective provincial capital city (regional center of the Campaign) as proxies for the

potential influence stemming from the Campaign and include them as addition controls in the

baseline regression. As shown in Column (4) of Table V, our main results still hold after

controlling for the potential confounding effect of the Campaign.

The second approach in dealing with the survivorship bias is to use genealogies compiled

after the end of Cultural Revolution. Thanks to no large-scale internal conflicts or violence

targeting local clan culture in China after the end of Cultural Revolution (CR), the stock of

genealogies complied in the post-CR period should not suffer from the survivorship bias. In

13 We have already partially addressed this concern related to the Campaign in Column (3) of Table IV by removing first-tire

and provincial capital cities from the estimation sample. Our main results remain intact in that case.

25

Column (5) of Table V, we use the post-CR genealogy number scaled by population size as an

alternative measure of clan culture intensity and re-estimate the baseline regression with this

alternative measure normalized. Using this alternative measure does not change our results

either. We again find a significantly positive effect of clan culture on family ownership. In

addition, we also compare prefecture’s genealogy stock prior to the end of the Cultural

Revolution with its post-CR stock to assess the importance of CR-related survivorship

problem. In Panel A of Table VI, we tabulate sample correlations between the total stock,

post-CR stock and pre-CR stock of genealogies. Given that the pre-CR stock is strongly

positively correlated with the post-CR stock (with a correlation coefficient of 0.923), it is thus

unlikely that prefectures with stronger clan culture historically experienced more severe

damages during the Cultural Revolution. This further rebuts the “targeting” argument for the

survivorship bias.

[Table VI about here]

Our final strategy is to make use of the genealogy data collected from Taga (1960),

which recorded detailed information about a collection of 1,228 Chinese genealogy books

held by libraries in Japan by the end of 1950s. This genealogy collection covered 120

Chinese prefectures and spanned from the Yuan Dynasty to 1939 in terms of compilation

time. As most genealogies in this collection were compiled during the Qing dynasty, we scale

the number of genealogies by the population size in 1820 for each prefecture. As evident in

Panel B of Table VI, this Japanese-collection-based measure of clan culture intensity is

positively correlated with those constructed using our primary genealogy data. More

importantly, since genealogies in this Japanese collection were all collected long before the

breakout of the Cultural Revolution, their geographical distributions were not subject to the

26

distortions caused by the Cultural Revolution. In Column (6) of Table V, we re-estimate the

baseline regression using this Japanese-collection-based measure of clan culture intensity

(normalized). The estimated marginal effect of clan culture intensity on family ownership

concentration remains positive and statistically significant. In a nutshell, all above exercises

suggest that our main results are not likely to be driven by survivorship bias in our primary

clan culture intensity measure.

V. Omitted Variables and Instrumental Variable Regressions

The results in Section IV suggest that our finding is robust to alternative measures of

family ownership, samples, and measures of local clan culture intensity. A remaining concern

is that our main results could be driven by some omitted variables. In this section, we carry

out three sets of exercises to tackle this issue. First, we control in our regressions for

additional factors that can potentially be correlated with local clan culture strength and family

ownership concentration. Second, we employ a complementary sample of Chinese listed

family firms in which we can identify owners’ birthplaces and separate the inherited

component of clan culture influence (clan culture of origin) from confounding environmental

factors by controlling for firm location prefecture fixed effects. Finally, we also use an

instrumental variable approach to further establish the causal effect of clan culture intensity

on the concentration of family ownership.

V.A. Additional control variables

Our first way to assess potential omitted variable bias is to further control for an

extensive set of prefecture characteristics. In particular, we consider three sets of variables at

the prefecture-level and report the results in Table VII. The first three columns of Table VII

27

aim to capture the effects of hostile environments. In Column (1), we construct a terrain

ruggedness index following Nunn and Puga (2012), with a larger value indicating more

rugged landscape within a prefecture. In Column (2), we include a measure for a prefecture’s

proneness to climate disasters based on the dryness/wetness index between 1470 and 1979.14

In Column (3), we also control for contemporaneous output volatility, measured by the

standard deviation of real GDP per capita over 1990-2010 (in natural log). The second set

includes prefecture-level human capital indicators. Specifically, we control for prefecture-

level average years of schooling (in natural log) and literacy rate in Columns (4) and (5),

respectively. Finally, in the last column of Table VII, we include the number of dialects for

each prefecture to control for the potential impact of local ethno-linguistic fractionalization.

Adding additional controls does not change our main finding. In all regressions, the

estimated effects of clan culture are still significantly positive. Moreover, the estimated

effects of the local clan culture intensity measure are largely stable when additional controls

are added, suggesting that omitted variable bias is unlikely to drive our results.

[Table VII about here]

V.B. Evidence from listed family firms

While we have controlled for a comprehensive set of prefecture-level covariates, it may

still not be sufficient to fully eliminate omitted variable bias. For instance, in the case that the

concentration of family ownership was related to other unobservable local environmental

factors, which may also correlate with local clan culture intensity, our estimates would be

biased. Towards this end, our second strategy is to turn to the listed family firm dataset, in

14 The dryness/wetness index over 1470-1979 was extracted from Zhongguo jinwubainian hanlao fenbu tuji (Yearly Charts

of Dryness/Wetness in China for the Last 500-Year Period), compiled by the State Meteorological Society in 1981. We

would like to thank Ying Bai for sharing the data with us.

28

which we can resort to the existence of movers and separate the effect of clan culture of

origin from confounding environmental factors by controlling for firm location prefecture

fixed effects.

Given the panel structure of the CSMAR listed family firm data, we estimate the

following Tobit model:

𝑦𝑖𝑗𝑡∗ = 𝛼 + 𝛽𝐶𝑙𝑎𝑛𝑜𝑐 + 𝜃𝑉𝑖𝑡 + 𝜑𝑊𝑗𝑡 + 𝛾𝑋𝑜𝑐 + 𝜇𝑖𝑛𝑑 + 𝜏𝑡 + 𝜔𝑓𝑐 + 𝜖𝑖𝑗𝑡,

𝑦𝑖𝑗𝑡 = 0 𝑖𝑓 𝑦𝑖𝑗𝑡∗ ≤ 0; 𝑦𝑖𝑗𝑡 = 𝑦𝑖𝑗𝑡

∗ 𝑖𝑓 0 < 𝑦𝑖𝑗𝑡∗ < 100; 𝑦𝑖𝑗 = 100 𝑖𝑓 𝑦𝑖𝑗𝑡

∗ ≥ 100 (2)

Here 𝐶𝑙𝑎𝑛𝑜𝑐 is the clan culture intensity at owner’s birth-prefecture, oc. 𝑉𝑖𝑡 is a set of firm-

level controls, including firm i’s size, age, return on equity, sales growth, Tobin’s Q and the

number of listing years. 𝑊𝑗𝑡 is a set of owner j’s demographic characteristics including

gender, age and education background. 𝑋𝑜𝑐 is a set of prefecture characteristics associated

with owner’s birthplace (oc). 𝜇𝑖𝑛𝑑 and 𝜏𝑡 are the industry and year fixed effects. In

addition, we also control for firm location prefecture fixed effects (𝜔𝑓𝑐) to capture all time-

invariant unobserved heterogeneity for prefecture city, fc, where firms are located. In so

doing, we identify the effect of clan culture of origin on family ownership, 𝛽, using

variations across movers whose birthplaces differ from firms’ locations.

Column (1) of Table VIII reports the estimation results from Equation (2) using the listed

family firm data. The estimated marginal effect of clan culture of origin is positive and

statistically significant at the 1% level. This indicates that the inherited component of clan

culture indeed plays a crucial role in determining family ownership concentration in China’s

privately-owned firms.

A potential complication involved in this exercise is that firm owners’ moving decisions

29

may not be totally random and some uncontrolled birthplace characteristics behind the

moving decisions might be related to firm’s family ownership structure. Whereas a full-

fledged analysis of individuals’ moving decisions is beyond the scope of this study, we gauge

the relative importance of nonrandom moving decision following the strategy of Guiso et al.

(2004a). If our results were driven by some uncontrolled factors related to firm owner’s

moving decisions, we should observe (1) significant differences in family ownership

concentration between movers and non-movers and (2) that the impact of clan culture of

origin varies systematically with the direction of moving. To check the first possibility,

Column (2) of Table VIII adds a mover dummy as an additional control in the model. The

estimated effect of clan culture of origin remains positive and statistically significant at the

1% level. Note that the mover dummy is statistically insignificant, indicating no significant

difference in family ownership between movers and non-movers.

To rule out the second possibility, in the rest columns of Table VIII, we further control

for interaction terms of clan culture of origin with factors that are potentially related to both

owners’ relocation patterns and family ownership concentration. The first factor we consider

is the difference in economic development between owner’s birthplace and firm’s location. If

seeking better economic prospects for themselves or their families was correlated with both

owners’ moving decisions and the subsequent family ownership concentration in their firms,

we should expect the impact of clan culture of origin on family ownership concentration be

significantly different for owners moving away from economic backwardness. As coastal

regions in China are typically more developed economically, we include in Column (3) an

interaction term between clan culture of origin and a dummy variable for moving from an

inland prefecture to a coastal one. In Column (4), we also create a dummy variable for

30

moving from a poor prefecture to a rich one, that is, equal to one if the GDP per capita in

owner’ s birthplace is below the sample median while that in firm’s location is above the

sample median, and interact it with clan culture of origin. We find that, in both cases, the

estimated effects of clan culture of origin per se remain positive and significant. Yet neither

interaction term is significant, suggesting that relocating for better economic outcomes seems

unlikely to be a leading factor empirically.

Next, we examine whether relocating across provinces would make a difference on the

relationship between clan culture of origin and family ownership. Given the substantial

amount of heterogeneities in various aspects across provinces in China, an individual’s inter-

provincial moving decision may reflect some unobserved characteristics, such as risk

tolerance, which, in turn, could potentially be intertwined with the impact of owner’s clan

culture of origin. We check this possibility in Column (5) of Table VIII by controlling for an

interaction term between clan culture of origin and a dummy for cross-province moving.

Again, we find a positive and significant marginal effect of clan culture of origin but a

statistically insignificant interaction effect.

In the last column of Table VIII, we further restrict our attention to a subsample of

provinces that experienced both in-migration and out-migration of entrepreneurs (i.e., two-

way moves). While the non-random moving problem is more likely to surface in provinces

with either in-migration or out-migration only (i.e., one-way moves), this would be less of a

concern for those with two-way movements. We re-run the regression using this two-way

move subsample and find very similar results as before. The estimated coefficient on the clan

culture of origin variable remains positive and significant, while that on the mover dummy is

statistically insignificant. Overall, the evidence presented in Table VIII suggests that non-

31

random migration is not a leading concern.

[Table VIII about here]

V.C. Results from instrumental variable regressions

In this subsection we further establish the causal relationship between local clan culture

intensity and family ownership concentration by employing an instrument variable (IV)

approach. Conceptually, a valid instrument in our case should be related to local clan culture

strength but does not affect family ownership directly via other channels than clan culture

influence. Here we use the minimum distance to two academies, the Kaoting Academy

(Kaoting Shuyuan) and the Xiangshan Academy (Xiangshan Shuyuan), as the IV. The two

academies were established respectively in the Southern Song Dynasty by two prominent

neo-Confucian scholars, Zhu Xi and Lu Jiuyuan, and were their primary places preaching

clan-related doctrines.

Zhu Xi and Lu Jiuyuan both have played important roles in the historical development of

clan organizations in China (e.g., Xu, 1995). Being a fervent advocate of clans, Zhu Xi

provided detailed institutional designs for Chinese clan organizations, which became the

standard social practices in the subsequent eight centuries. In his scholarly work Jiali (the

Family Rituals), Zhu Xi not only elaborated on the importance of establishing clan common

properties, such as ancestral halls, ancestral graveyards, and ritual land, but also laid out

detailed instructions on the design of ancestral halls and the rules for family rituals like adult

ceremonies, marriages, funerals, and ancestral offerings etc.

Notwithstanding different philosophical perspectives from Zhu Xi, Lu Jiuyuan is another

ardent supporter of the patriarchal clan organizations. Lu not only taught untiringly in the

academy the rules and rituals that helped to promote the sound functioning of family clans

32

but also, together with his brothers, put those rules and rituals into practice. An honorary title,

Lushi Yimen (the Lu Communal Family of High Moral Standards and Discipline), was

conferred on the Lu family by the emperor of the Southern Song Dynasty in 1242 AD to

honor Lu and his brothers’ efforts in maintaining a large-scale co-resident family clan. One of

Lu’s brothers, Lu Jiushao, also published an influential book, Lushi jiazhi (Management of

the Lu Family), summarizing their family practices regarding clan rules and management,

which later became a classic for the internal governance and management of Chinese clans.

Given the indispensable roles of Zhu and Lu in shaping Chinese clan culture, it is

reasonable to consider the minimum distance to their respective academies as an appropriate

candidate for our IV. Specifically, we expect that prefectures closer to these two academies

were more exposed to their influences historically and, thanks to cultural persistence, are

more likely to preserve strong clan culture heritage today. To test for the relevance of our IV,

we run a prefecture-level cross-section regression in Column (1) of Table IX, using our

primary measure of prefecture-level clan culture intensity as the dependent variable. The

negative and significant coefficient on the minimum distance variable indicates that

prefectures closer to the two academies are indeed characterized by stronger local clan

culture influence, confirming that our IV is relevant.

The location choices of the two academies are largely random. Zhu Xi chose to establish

the Kaoting Academy in the Nanping city of Fujian Province because his father used to be a

local government official there and had a wish to live there after retirement. The Xiangshan

Academy was built by a student of Lu Jiuyuan, and the location was chosen to be close to

Lu’s home. To ensure that our IV does not affect family ownership concentration through

other channels than clan culture, we perform a set of placebo tests. We first consider the

33

human capital channel. A potential threat is that the IV may affect family ownership through

its impact on human capital. There is good reason, however, to believe that this is unlikely.

Historically, there were also many other academies/lecture places of Confucian scholars (with

less relevance to the development of clans though). Human capital thus depended not only on

the minimum distance to these two particular academies but the distances to other academies

as well. For example, according to historical records, China had more than 4,300 Confucian

teaching academies in the early Qing Dynasty (e.g., Bai, 1995). Therefore, we do not expect

our instrument to have a significant effect on modern human capital. This is confirmed in

Columns (2) and (3) of Table IX, where the minimum distance is not significantly correlated

with prefecture’s average years of schooling (in natural log) or literacy rate in 2000.

The second potential channel for our IV to affect family ownership may be related to

some external environmental attributes. To rule out this possibility, we focus on prefecture’s

terrain ruggedness and proneness to climate disasters in history in Columns (4) and (5).

Again, we observe no significant correlation between the minimum distance variable and the

two environmental factors. Third, we check whether the minimum distance variable is

correlated with current economic development. Using the prefecture-level population density

in 2000 as the dependent variable, Column (6) of Table IX finds no significant difference in

economic development for prefectures with different distances to Zhu’s and Lu’s academies.

Last, we also examine whether the minimum distance may affect the degree of ethno-

linguistic fractionalization, which, in turn, could affect the concentration of family

ownership. In Column (7), we regress the number of dialects for each prefecture obtained

from Wurm et al. (1987) on the minimum distance variable and find no statistically

significant relationship between them. We admittedly cannot rule out all possible channels.

34

But we feel that evidence from the above exclusion tests, together with that from the

relevance test, makes a plausible case that the minimum distance variable is a reasonable IV

for our study.

[Table 9 about here]

Panels A and B in Table X present the maximum likelihood estimates from instrumental

variable Tobit regressions using the CPES2010 data and the listed family firm data,

respectively.15 For each panel, we report the second-stage results in the first column and the

first-stage results in the second column. In Columns (1) and (4), the estimated marginal

effects on the clan culture variable remain positive and statistically significant. Note that, in

Columns (2) and (5), the IV is always significantly negative in the first-stage regression. This

finding is consistent with our conjecture that prefectures closer to the two academies are more

heavily influenced by clan culture. Finally, in the last column of each panel, we perform

another placebo test by including the local clan culture measure and the IV simultaneously in

the Tobit regression. Our findings of a significantly positive coefficient on the local clan

culture measure but an insignificant one on the IV further verifies that our instrument does

not exert other independent impacts on family ownership concentration besides working

through local clan culture influence. Overall, the results from the IV regressions confirm

further that local clan culture has a positive causal effect on the concentration of family

ownership in China.

[Table X about here]

15 When the CPES2010 data is used, we apply the IV Tobit to the model specified in Equation (1). When the listed family

firm data is used, we estimate the IV Tobit regression based on the model specification in Equation (2).

35

VI. Understanding the Mechanisms

In this section we also make efforts to explore some plausible mechanisms behind our

main results. Specifically, we turn to an individual-level survey data as well as the firm-level

survey data and provide some suggestive evidence for three possible channels discussed in

Section II.D. In so doing, we are able to shed more light on the causality between local clan

culture influence and family ownership concentration.

VI.A. A short-radius trust attitude

As laid out in Section II.D, one important channel through which clan culture can

potentially influence family ownership is the short-radius trust attitude fostered by the

familism culture. Here we provide some evidence for the linkage between local clan culture

intensity and individuals’ short-radius trust attitudes, using the data from the 2012 wave of

the China Family Panel Studies (CFPS). In this survey, individual respondents were asked to

rate their trust attitudes towards their parents and strangers, respectively, using scores ranging

from 0 to 10 with a larger value indicating a higher degree of trust. We locate each individual

at the prefecture level and match this information with our prefecture-level measure of local

clan culture intensity. We then classify an individual as having a strong trust in parents

(strangers) if his/her self-reported rating belongs to the top tercile (i.e., no less than 8).

Columns (1) and (2) in Table XI report the estimated marginal effects, from logit

models, of clan culture intensity on individual’s trust in their parents and strangers,

respectively. As expected, the clan culture intensity variable is significantly positive in the

trust-in-parents regression but significantly negative in the trust-in-strangers regression. In

Column (3) we also use a measure of trust asymmetry, defined as the differential between the

two trust dummies, as the dependent variable. By construction, this trust asymmetry measure

36

can take the value of -1, 0 or 1, with a larger value indicating more trust in parents but less in

strangers. We estimate this regression using an ordered Logit model. The estimated marginal

effect of clan culture is again found to be significantly positive. That is, individuals in

prefectures with stronger clan culture influences indeed exhibit higher trust in their parents

but lower trust in strangers.

In the next three columns of Table XI, we compute the average scores for trust in parents

and trust in strangers, respectively, as well as their differentials for each prefecture and then

examine the impact of local clan culture intensity on these average trust attitudes in the cross

section of prefectures surveyed in the CFPS2012 data. We again observe a very similar

pattern – strong local clan culture is associated with a short-radius trust attitude.

[Table XI about here]

In Table XII, we rely on variations across provinces and industries to provide additional

evidence consistent with the trust channel. First, we consider the cross-province variation in

local legal environment. Numerous studies have documented that the quality of legal

institutions plays a vital role in determining agency costs (e.g., Rajan and Zingales, 1998;

Guizo et al., 2004b; La Porta, Lopez-de-Silanes, Shleifer, and Vishny, 2002; Burkart et al.,

2003; Bertrand and Schoar, 2006). While a weak legal environment would amplify the impact

of agency costs associated with interpersonal mistrust (e.g., resulted from strong clan

culture), a good legal environment could reduce the impact of the mistrust-related agency

costs. If the short-radius trust channel is indeed at work, we would expect a smaller impact of

clan culture intensity on family ownership concentration in better legal environments.

We test this plausible mechanism in the first three columns of Table XII by introducing

interaction terms between prefecture-level clan culture intensity and province-level measures

37

of legal environments. Column (1) measures the quality of legal institution at the province

level with contract enforcement time, defined as the number of days from the time the

plaintiff files the lawsuit till the time of payment. This data is obtained from the World

Bank’s Doing Business in China 2008. The shorter the contract enforcement time, the better

the legal environment. Columns (2) and (3) use the average share of lawyers in local

population and the legal protection index from Fan et al. (2010) as the proxies for the quality

of legal institutions at the province-level. The higher the lawyer share or the larger the legal

protection index value, the better the quality of legal institution. For each measure, we divide

its sample distribution into terciles and interact the clan culture intensity measure with

dummies for the middle and top terciles, respectively.16 By construction, the top tercile

dummy for contract enforcement time refers to the weakest legal environment, while those

for lawyer share and legal protection index refer to the strongest legal environment. As

evident in the first three columns of Table XII, the positive impact of clan culture intensity on

family ownership concentration is significantly more attenuated in provinces with better legal

institutions characterized by less contract enforcement time involved, higher lawyer share and

larger value of legal protection index. These results thus lend supportive evidence for the

trust channel.

Next, we examine the trust channel through the lens of firm’s control potential, that is,

the wealth gain achievable through more effective monitoring of managerial performance by

firm owners. According to Demsetz and Lehn (1985), disentangling the wealth loss due to

managerial misbehavior from that caused by other largely exogenous factors is more difficult

16 All the province-level legal environment measures are submerged with the inclusion of the province fixed effects.

38

or costly in less predictable environments (e.g., instability in prices, technology or market

shares). Viewed in this light, the noisier a firm’s operating environment, the greater the

payoff to owners in maintaining tighter control (i.e., larger control potential), which, in turn,

could further augment the mistrust-related agency costs. Following this line of reasoning, if

local clan culture indeed works through the short-radius trust attitude, we should expect a

larger effect of clan culture in firms with greater control potential. Here we follow Demsetz

and Lehn (1985) to proxy for firm’s control potential with the industry-level environmental

instability, defined as the standard deviation of annual real value-added growth for each

industry over the period 1980–2005.17 A firm is considered to have a higher control potential

if it operates in a sector with more volatile growth. We then include the interaction term of

clan culture intensity with the control potential variable in our regression in Column (4). The

significantly positive coefficient on the interaction term suggests that clan culture has a more

pronounced effect for firms with larger control potentials.

Finally, we also make use of owner’s self-reported history of investing in other firms

from the CPES2008 data to provide additional evidence for the trust channel. In the survey,

owners were asked whether they took any equity stakes in other firms in the last 3 years. We

then estimate a logit regression for the probability of investing in other firms in Column (5).

The results show that firm owners under stronger clan culture influences are less likely to

invest in other firms, which could be potentially explained by their short-radius trust attitude.

Overall, the results presented in Table XII are consistent with the short-radius trust

17 Industry-level real value-added data for China are obtained from the Groningen Growth and Development Center

(GGDC) ten-sector database, which contains the annual series of real value-added for ten broad industries, including

agriculture, mining, manufacturing, utilities, construction, trade services, transport services, business services, government

services, and personal services. We then use normalized industry volatility in the regression to facilitate interpretation. The

level effect of the industry volatility variable per se is absorbed by the industry fixed effects.

39

mechanism.

[Table XII about here]

VI.B. Financial resources pooling

Another plausible reason why strong clan culture boosts family ownership may be

related to the tradition of common property ownership cherished by the Chinese clan culture.

In prefectures with strong clan culture influences, family members are more likely to pool

together their financial resources to help owners establish and run their businesses, leading to

a larger share of firm equity held by family members. We test this possibility in Table XIII.

In Column (1), we first compute the ratio of other family members’ ownership share to

firm owner’s share using the CPES2010 data and then examine whether this composition of

family ownership would be affected by clan culture intensity.18 In the regression, we also

include the number of family investors (excluding firm owner) to control for its potential

confounding effect. The results are consistent with the financial resources pooling channel:

stronger local clan culture influence is associated with significantly larger shares held by

other family members relative to the firm owner.

In the next three columns, we make use of cross-province variation in financial market

development to verify the financial resources pooling channel. Now that the financial

resources pooling function embedded in Chinese clan culture can effectively relax firms’

financial constraints, we would expect this channel to have a more pronounced impact on the

ownership structure of firms located in prefectures where financial markets are less

developed and obtaining external finance is more difficult. Columns (2) and (3) measure the

18 According to the private enterprise survey, other family members refer to owner’s spouse, children, parents, siblings,

spouses of siblings, siblings of spouses, and other relatives.

40

quality of local financial institutions at the province level with the time and the cost for a firm

to create and register the collateral necessary for obtaining credit. Both measures are obtained

from World Bank’s Doing Business in China 2008, with larger values reflecting more

difficulty in accessing credit and hence poorer financial institutions. Column (4) uses the

province-level financial marketization index from Fan et al. (2010) as a proxy for local

financial environment. For each measure, we also divide it into terciles and interact the clan

culture measure with the middle and top tercile dummies. In all three columns, we find

evidence that is consistent with the financial resources pooling mechanism. The interaction

terms between clan culture and the top tercile dummies are all significant with expected

signs, suggesting that clan culture indeed matters more in financially less developed

provinces.

VI.C. Amenity potential

Finally, it is also possible that clan culture intensity affects firm’s family ownership

concentration via the amenity potential channel. Owners with strong clan culture heritage

could potentially derive, from family ownership and control of firms, more nonpecuniary

private benefits, such as fulfilling the obligations to provide jobs to family members, and

family honor etc. Here we provide some suggestive evidence for this amenity potential

channel by resorting to information about owners’ opinion on providing employment for

family members. In the CPES2010, one question was asked: “Is it a firm owner’s duty to

provide family members with jobs in the firm?” In Column (5) of Table XIII, we examine

whether owners from prefectures with stronger clan culture influence are more likely to agree

on their job-offering duties, using a logit model. The positive and statistically significant

marginal effect of the clan culture intensity measure means that owners indeed feel more

41

obligated to provide employment for family members when they have been exposed to

stronger local clan culture influence. This suggests that the amenity potential could be

another reason why strong clan culture leads to higher concentration of family ownership.

Taken together, the complementary evidence presented in this section is consistent with

the mechanisms we outlined in Section II.D. This thus makes us more confident in suggesting

a causal interpretation of the estimated relationship between local clan culture intensity and

firms’ family ownership concentration.

[Table XIII about here]

VII. Conclusions

This study provides a cultural explanation for the concentration of family ownership in

Chinese privately-owned firms. Using the CPES data, w