Textiles and the Historical Emergence of Gender Equality in China*

Textiles and the Historical Emergence of Gender
Equality in China*
Melanie Meng Xue„
UCLA Anderson School of Management
This Version: November 2015
Abstract
This paper tests the hypothesis that pre-modern textile production leads to greater
gender equality. I exploit variation in pre-modern textiles at the county level to
identify their effects on gender equality since 1300. I find despite the the lack of benefits of working outside home, places where women were more productive in textile
production developed more favorable views towards working women and witnessed
rising gender equality. Pre-modern textile production is negatively correlated with
prenatal sex selection, and positively correlated with wife heading the household
and with share of women with political and managerial roles. My results are robust
to using various subsamples, sex ratios from other cohorts, other gender outcomes,
alternative hypotheses, an instrumental variable analysis, and a micro-level analysis. I also investigate the effects of large political and economic shocks after 1850,
and find the effects of pre-modern textile production to be overall highly resilient. I
document an intermediate effect of textiles on women’s probability of survival upon
widowhood dating back to 1500. In addition, I show that women exposed to preindustrial textile production were more likely to participate in the labor force as early
as at the onset of China’s industrialization in the 20th century. Using survey data, I
show that in contemporary China, individuals in counties with a pre-modern textile
history indeed have more progressive gender moderns and weaker son preference.
Keywords: Culture, historical persistence, gender norms, son preference, relative productivity
JEL Codes:J16 N35 O33 O53
*I
would like to thank Sascher Becker, John Brown, Joyce Burnett, Shuji Cao, Latika Chaudhary, Daniel
Chen, Bill Collins, Lena Edlund, James Fenske, Philip Hoffman, Paola Giuliano, Remi Jedwab, Saumitra Jha,
Noel Johnson, Mark Koyama, Timur Kuran, James Kung, Nan Li, Debin Ma, John Nye, Ömer Ózak, Nancy
Qian, Thomas Rawski, Gary Richardson, Eric Schneider, Yan Se, Nico Voigtländer, Bin Wong and audiences
at ASREC 2013, CES 2014, “Deep Causes of Economic Development” at Utrecht (2014), EHA 2014, EHS
2014, GMU-AU Economic History Workshop (2012), HKEA (2014), the International Workshop on Economic
Analysis of Institutions at Xiamen University (2013), NEUDC(2015), Shanghai University of Economics and
Finance (2014), World Congress of Cliometrics Society (2013), WADES (2014), and WEHC (2015). All the
remaining errors are the fault of the author.
„ [email protected]
I
Introduction
Gender discrimination is extremely common across the world. Going beyond traditional research
on the economic status and political rights of women, recent work has paid increasing attention
to the cultural determination of gender attitudes (Fernández, 2007; Fernández and Fogli, 2009;
Fortin, 2005; Jayachandran, 2015). In particular, Alesina, Giuliano, and Nunn (2013) and
Hansen, Jensen, and Skovsgaard (2015) have focused on the pre-modern roots of gender norms,
highlighting historical agriculture in shaping gender roles and gender norms. This paper builds
on this strand of literature to investigate pre-modern roots of gender equality. It does so by
focusing on textile production—the main economic activity of women in most preindustrial
societies.
Alesina, Giuliano, and Nunn (2013) explore how traditional agricultural practices influenced the
historical gender division of labor and the modern levels of gender equality across countries.1
They find that plough cultivation in pre-industrial periods led to an important belief that
the natural place for women is within the home. In contrast to Alesina, Giuliano, and Nunn
(2013), this paper explores a setting where women works inside home, but the productivity of
women varies. I distinguish between two channels through which pre-modern production could
have shaped gender norms and beliefs: working inside or outside home, versus women’s role in
providing for family. Pre-modern China offers an ideal testing ground because the predominant
ethnic group, Han Chinese, used the plough; however, women’s status varied tremendously
within that ethnic group. The plough was operated in the fields, mainly by men; textiles were
produced at home, mainly by women. Types and quantities of textiles varied greatly across
regions. Women in some places were far more productive than women elsewhere, partly due
to more favorable climatic and geographic conditions. I hypothesize that women doing highly
productive work to provide for her family, even at home, can increase their status.
The extent of variation in women’s status in China provides an ideal setting to study the
historical and modern causes of gender inequality. In contemporary China, a leading indicator
of gender inequality is sex ratio at birth. Across China a tremendous amount of variation is found
in the degree of male bias in this measure. Some counties have maintained close-to-normal sex
ratios at birth, whereas in other parts, sex ratio imbalances have become increasingly more severe
since sex-selective technology became available in the late 1980s. According to 2000 Chinese
Population Census, among all 2876 counties, in Jincheng, 102 boys were born per 100 girls;
whereas in Erzhou, 170 boys were born per 100 girls. The amount of male bias in sex ratio at
birth is not a simple function of economic development. As Jayachandran (2015) points out, the
problem of the male-skewed sex ratio at birth cannot be explained by the process of development:
sex ratio imbalances have been intensifying, not lessening, with economic development. Sex ratio
at birth more reflects long-standing prejudice against girls. Hence I use sex ratio at birth as a
1
For surveys of field see Guiso et al. (2006); Bisin et al. (2011) and Nunn (2012).
1
main outcome variable to test the impact of pre-modern textiles on gender equality.
To identify the causal effect of pre-modern textile production on modern outcomes, including
sex ratio at birth, I employ a variety of empirical strategies. First, I collect information on premodern textile production collected from thousands of local gazetteers. I end up with a sample of
1513 counties in my main analysis. Second, I link this data to contemporary measures of gender
equality. My outcome variables include sex ratio at birth, women’s education, wife heading the
household among married couples, share of women with political roles and managerial roles,
gender-role attitudes and son preference.
Examining variation across counties and individuals, I find a strong positive relationship between
pre-modern textiles and gender equality today. The baseline estimates suggest that pre-modern
textile production is associated with a reduction in sex ratio at birth by a quarter of its standard
deviation. Pre-modern textiles are negatively correlated with sex ratio imbalances at birth, and
positively correlated with wife heading the household among married couples and with share
of women with political roles and managerial roles. Pre-modern textiles predict more favorable
views about women’s natural ability relative to men. People in areas with pre-modern textiles
are more likely to believe that women should be as career-focused as men.
To account for historical and modern factors that might also play a role in modern gender
outcomes, I include controls for a number of historical characteristics of each county, such as
agricultural suitability, proximity to the Grand Canal or the Yangtze River, treaty port status,
as well as a set of geographic controls, such as latitude, longitude, ruggedness and distance
to coast. I also control for current per capita GDP, share of agricultural workforce, share
of non-agricultural household registration, share of ethnic population, and provincial capital.
Region and province fixed effects are included in all specifications. In running various robustness
checks and considering alternative hypothesis, I also analyze the role of tea production, rice
production, level of development prior to cotton textile production, historical postal routes,
Christian missionary interventions, present-day textile manufacturing, educational attainment,
and migration.
My results are robust to an instrumental variable approach. Farnie (1979) points out that
relative humidity played a key role in textile production.2 Humidity makes cotton fibers more
pliable and reduces the chance of breakages in the yarn. Humidity is positively correlated with
the quality of final products until the effect flattens out approximately at 80%. Hardly any cotton
textiles can be produced when relative humidity drops below 60%. This motivates the use of
county-level relative humidity as an instrument for pre-modern textile production. I aggregate
monthly average relative humidity of each county, based on their contribution to a suitable
environment for spinning and weaving, into a humidity-for-weaving index. This rationale behind
this index is to approximate the number of months available for cotton textile production with
2
Fairbank (1978) discusses the relationship between a relative humid climate of Jiangsu and the greater
tensile strength and evenness of yarn.
2
a gradient to quality and efficiency. I obtain IV estimates that are comparable to the OLS
estimates providing further confidence that these coefficients can be causally interpreted.
To explain the mechanisms through which pre-modern textiles affect modern outcomes, I review
the history of widow survival in China, examine a number of outcomes of textile production in
the past, and carefully consider the likely effects of several large political and economic shocks
on the persistence of gender norms. I first show that pre-modern textiles predict higher survival
rates of widows as early as the Ming period, suggesting that textiles were already empowering
women historically. I go on to demonstrate female labor participation pattern in the earliest
stages of China’s industrialization. I find women in regions with pre-modern textiles were far
more likely to be represented in the labor force in the early 20th century. I carefully evaluate the
effect of adopting western institutions, the influence of missionary activities, the formation of the
Chinese communist state, and recent economic reforms in China. Little evidence is found that
any of the large political and economic shocks in modern China have severely confounded the
estimates I obtain on pre-modern textiles. Many shocks could have had a level effect on gender
norms, but they are rarely correlated with pre-modern textiles. Even when they are correlated,
they do not seem to affect my main results. I find the persistence effects pre-modern textiles are
overall highly resilient, despite some suggestive evidence that large political or economic shocks
could temporarily or even permanently weaken the persistence. Before I conclude, I use Chinese
Social General Surveys to provide direct evidence on pre-modern textiles shaping gender-role
attitudes and son preference in contemporary China.
Previous studies have showed increased women’s earnings lead to female empowerment (Anderson and Eswaran, 2009; Ashraf, Karlan, and Yin, 2010; Deininger, Goyal, and Nagarajan,
2010). In studying the determinants of female autonomy, Anderson and Eswaran (2009) find
earned income contributes more to women’s autonomy relative to unearned income, and that
only employment outside their husbands’ farms contributes to women’s autonomy. The relationship between sex-specific income and survival of daughters in East Asia has been documented
by Qian (2008). Qian interprets the finding that excess female mortality is decreasing in female
earnings in terms of women’s increasing bargaining power in the household. My paper shows the
shift in women’s relative earnings can change gender norms in the long run. I provide evidence
that women enjoyed greater autonomy and higher social status in areas with pre-modern textile
production; relative outcomes of women continue to be better even when pre-modern textile
production is long out of the picture. I argue that pre-modern textiles shaped cultural beliefs
about the women’s capability and their proper role in providing for a family.
However, there are also alternative arguments that suggest that part of the long-term impact
arises because pre-modern textile production promoted the development of formal institutions,
gender-specific economic opportunities and overall wealth that may favor women. To rule out
the first channel, I rely on within-country, within-region, and within-province variation, where
3
formal institutions are largely identical and policies are enforced to a similar extent.3 Also,
unlike the west, the Chinese state standardized family practices across regions, classes and
dialect groups by the late imperial period, with far fewer time and space variation in inheritance
practices, marriage rates, naming practices and patrilocality (Ebrey, 1990; Ropp, 1994). After
1949, pervasive reforms in favor of gender equality took place. Formal institutions were created
to guarantee female employment. Due to a high degree of centralization in legislation, official
labor laws or laws on sex discrimination hardly varied from place to place. All of these create
beneficial conditions for me to identify the impact of historical determinants on present-day
gender outcomes. To account for the second channel, I control for sectoral composition today,
including scale of textile production and agricultural workforce. To best account for the third
channel, I control for both current per capita GDP. I show that overall wealth alone hardly
explains the large and systematic differences in sex ratio at birth as documented in this paper.
This paper contributes to the literature on the historical determinants of cultural norms and
beliefs. Many of these document the persistent impact of a negative shock on current cultural
values such as Nunn and Wantchekon (2011)’s work on the effects of the trans-Atlantic slave
trade on corruption and trust today and Voigtländer and Voth (2012)’s study of the persistence
of antisemitic beliefs in Germany. My study is most closely related to those papers that study
how economic factors have shaped contemporary gender norms such Grosjean and Khattar
(2014) who examine conservative gender norms and its origins in historical marriage market
conditions in Australia.4
This paper also contributes to the literature on parental gender bias and sex ratio imbalances
by identifying an important source of differentials in sex ratios. Edlund (1999) explicitly models
sex ratios in relation to son preference, indicating several factors that contribute to unbalanced
sex ratios. Jayachandran (2015) confirms the crucial importance of cultural factors in sex ratio
imbalances. Daul and Moretti (2008) finds evidence for parental gender bias in the U.S. that
parents favor boys over girls. Others have studied son preference, sex-selective abortions, and
changes in sex ratios in non-western countries (Gupta, 2014; Li and Lavely, 2003).5 In particular,
Chung and Gupta (2007) suggests income levels play a key role in unbalanced sex ratios and
that sex ratios can change in nonlinearity through different stages of development.6 Besides,
3
Despite the highly centralized law making process, policies can be implemented by local governments with
greater latitude.
4
Other relevant studies include Jha (2013) who shows that a cities in India that were medieval trading ports
experienced significantly less religious riots between Muslims and Hindus in the period after 1850. Grosjean
(2011) examines the persistence of a culture of honor among Americans of Scots-Irish descent. She finds that
this culture of honor results in higher homicide rates among Scots-Irish in the US South and Mountain West but
not elsewhere and argues that this culture has only persisted where formal institutions are comparatively weak.
5
Contrary to the conclusions of the above studies, Oster (2005) finds evidence that that much of sex ratio
imbalances in developing countries can be attributed to Hepatitis B.
6
Almond et al. (2013) find positive incomes shocks from land reforms increased sex ratios. Finally, the
economic consequences of sex ratio imbalances have also attracted scholarly attention in recent years. Wei and
Zhang (2011) links sex ratio imbalances to differential saving rates across China.
4
in reviewing the history of widow survival in China, this paper also adds to the literature on
“missing unmarried women” (Anderson and Ray, 2015; Miguel, 2005; Oppong, 2006; Sossou,
2002).
The third literature this paper draws on is the economic history literature studying the impact
of textile production on the pre-modern Chinese economy in the context of the Great Divergence
(Huang, 1990; Goldstone, 1996; Li and Li, 1998; Ma, 2005; Pomeranz, 2009; Wong, 2002). Several
scholars have argued that the 17th and 18th centuries were a golden period for the Yangtze Delta,
one of the major textile regions. Pomeranz and Li, in particular, have argued that China’s textile
industry remained highly productive and profitable through the 19th century (Li and Li, 1998;
Pomeranz, 2009), in face of tough competition from British textile manufactures. These studies
substantiate my claim that women’s textile production was highly productive, marketized and
immense in total size.
China is usually considered as having historically conservative gender norms, due to its Confucian
heritage. Severe son preference is often cited as evidence for historically conservative gender
norms and linked to deep cultural determinants such as patrilocality. However, China also
accounts for two-thirds of the richest self-made women in the world.7 Meanwhile, its neighboring
country, Korea, also with a Confucian heritage, has up to 34% CEOs being women, compared
to 9% in Western Europe and 5% in the US. This suggests women in those societies might have
gained status in spite of patrilocality. By exploiting variation in gender gap given the presence
of Confucian heritage, this paper fills in the gap to illustrate the sources of gender equality in
China.
This paper is organized as follows. The second section explains the historical background and
lays out the conceptual framework. Section III discusses data sources and variable constructions.
Section IV summarizes my baseline results, a number of robustness checks, an instrumental
variable analysis and tests of alternative hypotheses. Section V demonstrates that similar effects
can be found using an alternative micro-level datasets. In Section VI, I explore how the effects
of pre-modern textile production first emerged and how they persisted after the decline in
traditional textile production in the late nineteenth century. Section VII concludes the paper.
7
http://qz.com/529508/china-is-home-to-two-thirds-of-the-worlds-self-made-femalebillionaires/http://qz.com/529508/china-is-home-to-two-thirds-of-the-worlds-self-made-femalebillionaires/http://qz.com/529508/china-is-home-to-two-thirds-of-the-worlds-self-made-female-billionaires/
5
II
Historical Background and Conceptual
Framework
A Historical Background
A.1
Women in Pre-modern China
Similar to other patrilineal societies, women in premodern China were assigned different roles
to men and took on different responsibilities in both family and society. Chinese society was
also shaped in important ways by Confucian values.
Confucianism has a twofold impact on attitudes on to women. On the one hand, Confucian
tradition strongly disfavored women. Folk wisdom held that a family would suffer economically
from the birth of a daughter.8 This cultural belief was consistent with economic reality prior to
the emergence of the textile industry: daughters could not work outside home due to concern for
women’s “purity”, and had to rely on family resources to survive.9 And unlike sons, a daughter
would not be able to support her own parents once she became married as they had to move into
the homes of their husband’s family and became an official family member there. As a result of
the cost of dowries, too many daughters could cause serious financial distress to the household
(Harrell, 1995; Watson and Ebrey, 1991). For the reasons described above, parents had an
incentive to control the total number of daughters. Excess female mortality during infancy and
childhood was widely observed.10
On the other hand, Confucianism celebrated the virtues of hard work. It therefore elevated
individuals who worked hard to provide for their family, including hard working women. This
provided a path for women to earn the respect due to them for their contribution to the household. Diligent productive manual labor was seen as the virtue for all women, regardless of class
(Mann, 1997). Qing China (1644–1911) enjoyed a relatively high degree of social mobility, and
it was a society where it was conceivable for individuals to gain social status based on higher
productivity.
A.2
The Cotton Revolution
After agriculture, textile production was the most important economic activity in premodern
societies. In China as in much of the preindustrial world, textile production was carried out
by women “who spent every available moment spinning, weaving, and sewing.” (Barber, 1991).
8
In a play Qujiang Chi from early Yuan, the heroine refers to herself as pei qian huo, which literally
means a money-losing proposition. The term is still used in Mainland China, Singapore, Malaysia, Taiwan,
Macau and Hong Kong today. In 2007, the Yahoo dictionary in Taiwan was caught giving the English-language
translation of the Chinese term pei qian huo as a. “a money-losing proposition” and b. “a girl; a daughter”
(http://news.tvbs.com.tw/entry/305992 ).
9
Chow (1991) regards non-western women’s “purity” or “chastity” as both sexual and nationalistic.
10
Historian James Z. Lee and sociologist Cameron D. Campbell (2007) discovered that girls between ages one
and five had a 20 percent higher mortality than boys.
6
Spinning and weaving were perceived to be a womanly skill. “Men farms and women weave”
as a form of division of labor, became formalized under the state tax system dating back to
300 AD. Under the state tax system, each household was required to pay in-kind taxes in both
grain and textile products. Male labor was occupied by grain production, something women did
not have the comparable physical strength for. Just like other societies, women were far more
productive in making cloth than in agriculture. The fiscal policies of the state solidified this
gender division of labor.
In early times, silk and hemp were two main fabric for clothing. Silk was the most valued of
all fabric, and used in more expensive clothes. Hemp was the predominant fiber for day-to-day
clothes. After 1300 new spinning and weaving technologies for processing the cotton became
available. Cotton textile production was made economically viable for the first time in history
(Bray, 1997; Kang, 1977). The new technologies did not develop endogenously; rather, they
were borrowed from outside of the mainland. Huang Dao Po, a Shanghai native (1245–1330),
learned those technologies from an ethnic group Li residing on the Hainan Island.11 On the
spinning side, new technologies involved a pedal-spinning wheel with three spindles, similar to
the multi-spindle design used in the Spinning Jenny invented in 18th Century England. Rather
than maneuvered by hand only, new spinning machines were maneuvered by both foot and hand.
Prior to the innovation, every person’s workload of weaving has to be matched by three to four
people’s workload of spinning. In comparison, new spinning technologies dramatically increased
the efficiency of spinning.12 Following the technological breakthrough, cotton textiles rapidly
expanded in the following two centuries. Cotton quickly gained popularity over hemp in the
production of day-to-day clothes.
The introduction of new weaving and spinning technology allowed women to increase their
productivity. They sold the surpluses to the regional and national market. By Ming times
(1366–1644), the textile sector was increasingly commercialized and specialized. Cotton textiles
accounted for a large portion of in-kind taxes, second only to grain. Huang (1964a) estimates
that in the early 17th century, at least 1 million bolts of cotton cloth were transported through
the Grand Canal as tax payments to Ming Government.13 The passage of the Single Whip Law
in 1580 further promoted domestic trade and increased the market size for cotton textiles.14
Even after the mid-nineteenth century, as a significant part of spinning began to be displaced
by foreign yarn, weaving remained competitive.15 By 1933, handicraft industry still made up
11
Li people today still use those technologies for textile production. The production scene is an important
part of tourist attractions for the Hainan Island.
12
Other technology improvements include new techniques in cotton fluffing and crushing, and weaving insights
in mixed cotton fabrics, colored fabrics and fabrics with mixed warp and weft fibers.
13
1 bolt of cotton cloth is 33.33 meters in lengh. 1 million bolts of cotton cloth were worth half a million teals
at the time.
14
The Single Whip Law was initiated in the early 16th century, and promoted to the entire empire in 1580
by Zhang Juzheng (Flynn and Giráldez, 1995).
15
Rural China kept using handicraft cloth due to its lower prices and less wear and wear. Foreign merchants
and consular officials in the late nineteenth-century China complained about difficulty of penetrating the Chinese
7
for 61% of the total industry output (Fairbank, 1978, pp.15-28).
Women’s earnings from cotton textile production were substantial. Li (1997) shows a woman’s
textile work around the year was enough to feed 2.7 people. Pomeranz (2002) provides an even
more optimistic estimate:a female’s income could be four times as much as an adult male’s.
Allen’s (2011) wage regressions indicate that textile workers earned a wage premium compared
with workers in construction or agriculture. Women who had the skills to weave artisan cloth
could earn an even higher income.16 Women in textile regions became able to earn enough to
support a household independently for the first time.
B Conceptual Framework
Textile production following the cotton revolution, represented an new opportunity for women to
earn monetary income, and contribute to household income. As the return to producing textiles
was sufficiently high, women were induced to switch away from solely performing non-market
domestic work or producing other fabrics at low quantities mainly for home use.17
Specialization in cotton textile production was possible because of the existence of high developed goods markets. Shiue and Keller (2007) shows the performance of markets in China and
Western Europe overall was comparable in the late 18th century. Textile production at the
time, shared many similarities with work in the proto-industry in other advanced pre-modern
economies such as 18h century England. The following key differences deserve emphasis. (a.)
Chinese households typically owned the machines rather than renting them. Households occasionally owned more than one machine and hired help. But the scale was very limited.(Fairbank,
1978) (b.) Few concurrent technology shocks occurred during the time frame (1300-1840). The
cotton revolution took place in an agrarian economy, and the economy remained largely agrarian for the next sixth centuries. (c.) Though the goods market was both dense and highly
sophisticated, the labor market was far from being a free labor market. Emperors in Ming and
Qing institutes strict laws on labor mobility. The clan system continued to keep individuals tied
to their extended families (as discussed in Greif and Tabellini, 2010, 2015). This constrained
women’s ability to relocate to regions suitable for cotton textile production. (d.) Only a small
number of regions had the geo-climatic conditions suitable for spinning and weaving, and especially, for weaving. Cotton textile prices, for a long time, hovered at a level that generated
enough income for a skilled textile worker to support a family of four.18 (e.) A higher percentage of Chinese households owned land than British households involved in the putting-out
market, especially in the interior provinces.
16
The production of artisan cloth was backed up by popular demand of weddings and funerals in pre-modern
China. Its production requires both higher skills of weavers and longer hours.
17
This can be understood in the context of Pomeranz’s research on economics of respectability. In describing
the role of daughters in a family, he notes that a family’s capacity to survive and to profit from its work relied
upon “an optimal mix of family members of particular ages and sexes” (Pomeranz, 2005).
18
Allen (2009) shows one day’s work by a weaver in the late 17th century produced 7,684 calories, which was
adequate to support a family.
8
system in the 18th century. Despite periodic increases in land concentration ratio, China had
no equivalent of the movement of enclosure Britain experienced (1600-1850). The vast majority
of Chinese households where women were weavers had their own land on which men had to field
work. Together (b.), (c.) and (d.) ensured that prices of cotton textiles stayed reasonably high,
whereas (a.) and (e.) led women to reaping most of the benefits from this revolution.
Chinese women had been doing productive work prior to the cotton revolution. But cotton
textile production enabled woman to take on a new role as their household’s major income
earner. By the late Ming period, both unmarried and married women in many textile regions
became predominantly producing for the market, and in many cases their income became the
main source of family incomes. Pomeranz (2005) argues that women became more respectable
due to their highly productive manual labor in textile production.19 And it is highly plausible
that from the perspective of parents, it became less mentally stressful and financially costly
to accept a daughter into the family as women had a potential role as productive members of
the economy in their own right. In fact, women’s ability to support themselves was frequently
tested in the case of widowhood: remarriage was discouraged in Ming and Qing China. However,
women in textile regions found themselves well capable of remaining solvent, in the absence of
their husband’s incomes.
Together this evidence suggests that the rise of the textile industry since the fourteenth century
constituted a significant shock to the level of women’s engagement in market activities and
greatly increased their economic independence. This shock could lead to the breakdown of
prior cultural beliefs concerning women’s role in providing for a family. (Bertrand et al., 2015)
discusses the importance of a gender identity norm—the view that a husband should earn a
higher income than wife—in marital formation and chances of divorce. They find that in fear of
making more money than her future husband does, wife is more likely to give up on work in order
to start a marriage. Meanwhile, dissolution of marriage is far more likely among couples who
violate this important gender identity norm. Their discussion of this gender identity norm helps
to understand the unique nature of the case of female empowerment driven by advanced cotton
textile technologies. Due to the appearance of a new technology and the presence of both a wellfunctioning market and historically-specific government institutions, women in textile regions
had unparalleled earning opportunities by the standard of past agrarian societies. As a result of
this change, a large number of women in certain regions of China began to earn comparable or
higher incomes than most men did. In those regions it is plausible that the gender identity norm
that husband should always make more money than women broke down under the pressure of
this large relative income shock. This breakdown in traditional gender identity norms could
lead to the emergence of more favorable beliefs about women and a more optimistic assessment
19
Man (2011) provides a summary of depictions of female breadwinners being tough and dependable in
historical accounts. Her sources include: (Chen et al., 1991; Gu, 1995; ?). In ?, Xu’s wife proudly proclaims
that she single-handedly supports the family and is a ‘strong woman’, ‘she-husband’. Apparently, her husband
is just passionate about literary writing and paintings, and clueless to how to make both ends meet.
9
of the fates of prospective daughters.
The evolution of gender norms following a long period of taking a more prominent role in
providing for a family, can play a crucial role in gender gap and women’s well-being in presentday China. Viewing general norms as a complex of nexus of different beliefs and attitudes
concerning the relative status of women, it is sufficient to note that gender norms are both
perpetuated from generation to generation (as shown theoretically by Bisin and Verdier (2001)
and as discussed empirically by numerous studies in sociology and economics (Moen et al.,
1997; Vella and Farré, 2007) and also shaped by the attitudes of others in society (Burda et al.,
2007). Both mechanisms generate cultural persistence and can explain why cultural values, once
established, can be difficult to dislodge.
Inherited gender attitudes shape a wide range of outcomes today. The most important one I
focus on in this paper is sex ratio at birth. When parents today make a decision as to whether
to have a boy or a girl, they do not have complete information on the future prospects of a
boy or a girl in contemporary world. They instead resort to general beliefs about whether boys
or girls are morel likely to thrive in society and to favor the family.20 Sex ratio imbalances
are an important indicator for differential values being assigned to each sex and thus a good
indicator for women’s status in society.21 These cultural beliefs are particularly important under
a one-child policy regime and can be exercised at low costs given the availability of sex-selection
technology.
To sum up, I argue that pre-modern textile production has a strong, persistent impact on
gender roles, gender norms and gender equality today. In particular, I hypothesize a relationship
between pre-modern textile production and gender equality in contemporary China. I theorize
that women’s use of a productive technology increases their social status and the desirability of
daughters. Women’s increasing relative contribution to household income generated new norms
about women’s role in the household as a main breadwinner. In textile regions people developed
the belief that women can protect families from destitution and allowed them to pay their taxes
just as effectively as men. The reformed gender norms and beliefs can persist even though the
economy has moved out of traditional cotton textile production.
20
Altruistic parents who care about whether or not their children have fruitful lives will prefer to have boys
if they live in a society where women are treated less well.
21
President’s Commission for the Study of Ethical Problems in Medicine and Biomedical and Behavior Research of the United States in 1983 states that there is no evidence that amniocentesis is being sought widely to
determine fetal sex. Surveys of parents and prospective parents indicate, however, a preference for sons (especially as the first-born child). If it became an accepted practice, the selection of sons in preference to daughters
would be yet another means of assigning greater social value to one sex over the other and perpetuating the
historical discrimination.
10
III
Data
I construct my main variable, pre-modern cotton textile production, from thousands of local
gazetteers published in historical times. I also construct contemporary measures of gender norms
and gender equality, historical and contemporary county characteristics from a large number of
modern censuses, historical sources and GIS files, and climatic and geographic characteristics
from the Climate Research Unit of University of East Anglia, FAO and NASA. From the digital
world map collection of Harvard University, CHGIS, I obtain shape files that contain historical
characteristics for the counties within China. For modern outcome variables, I use the countylevel National Population Censuses (1990, 2000, 2010) and the 2004 Industrial Census from
the China Geo Explorer, the Chinese City Statistical Yearbooks, individual-level census data
(1982, 1990) from IPUMS-International, and Chinese General Social Surveys (2005, 2010). To
construct large political and economic shock variables as well as past outcome variables, I tap
into local gazetteers and make use of economic censuses and statistics complied by missionaries
in the early twentieth century.
In this section, I mainly focus on data used in my county-level analysis, where a total of 1535
counties, 198 prefectures, 15 provinces, and 8 regions are used. Data sources for other historical,
geographic and contemporary variables can be found in the data appendix.
A Explanatory Variable: Pre-Modern Cotton Textiles
Based on local gazetteers between 1368 and 1840, I construct an indicator variable on pre-modern
cotton textile production at a county level. Local gazetteers were published by prefecture governments and county governments, containing information on local produces and manufactured
products. I go through county-level and prefecture-level gazetteers to extract information on
cotton textiles.22
It is possible a county that started textile production first would see a larger impact of textile
production in shaping values and beliefs. However, the way local gazetteers were organized
does not allow me to pinpoint the starting point of production by county. Due to similar data
limitations, I cannot examine the quantitative dimension of textile production by county. I
also do not have the full knowledge of the quality of cotton textiles at the county level. As
quantity produced and quality can be potential sources of heterogeneity in the treatment effect,
the estimates should be interpreted as average treatment effect of historical textile production.
To obtain an estimate of the distribution of then textile-producing counties and prefectures
across China today, I map historical locations of cotton textile production into a map of China
in 2000. Due to name and boundary changes of historical counties and prefectures, I resort
to time-series maps of Chinese counties and prefectures to first determine historical locations
22
In the data collection process, I mainly refer to the section on local specialties (shi huo zhi ) for evidence of
cotton textile production.
11
of cotton textile production. To be specific, I match county names in the source with county
names in a point shape file comprising times-series counties, prefecture names in the source with
prefecture names in a a polygon shape file comprising time-series prefectures. Finally, I spatially
join both times-series maps with the county shape file corresponding with the 2000 population
census to obtain a county-level estimate of cotton textile production. Figure 1 displays the
location of pre-modern textile production.
Figure 1: Explanatory Variable: Pre-Modern Cotton Textiles
B Main Outcome Variable: Sex Ratio at Birth
I use sex ratio at birth from the 2000 Census.23 Data on sex ratios at birth are available at
the county level. My main outcome variable is sex ratio at birth.24 As the distribution of raw
sex ratios is entirely skewed towards the left, I use a common data transformation technique to
create z cores of raw sex ratios.
There is considerable variation in the extent of sex ratios. In 2000, at the county level, sex ratios
at birth range from 81:100 to 196:100. With the exclusion of five autonomous regions, I still
23
The reason for mainly using sex ratio at birth in the 2000 census is that (a.) sex selection technologies were
widely available (Chen et al., 2013; Ebenstein, 2010). (b.) regional variation in one-child policy was limited to
urban-rural differences. After 2000, some counties began to be experiment a two-child policy for parents that
were both only-child. (c.) marriage rates were still high. Voluntary infertility was relatively rare. (d.) data
quality of the 2000 census is reportedly higher.
24
Alternatively, I can derive a measure of logged form of the deviation of sex ratio at birth from the normal
sex ratio, and I get very similar results using the alternative measure.
12
find a wide range of sex ratios (92:100 to 193:100) across counties. Figure 2 shows sex ratio at
birth in seven quantiles.
Figure 2: Main Outcome Variable: Sex Ratio at Birth
A male-biased sex ratio is a crucial indicator of gender bias. China has had the most unbalanced
sex ratios in East Asia for the past decade. In the 2000 Census, the national average sex ratio at
birth is 118:100, i.e. every 118 boys were born to every 100 girls. Prior to the one-child policy,
people resorted to higher-parity births to ensure male offspring. A major problem for identifying
the magnitude of gender bias in this setting is that the characteristics associated with low fertility
are often correlated with characteristics associated with gender equality. Though stopping rules
can distort sex ratios, the distortions become smaller as number of children increases. As a
result, places with high fertility and high levels of gender inequality may not necessarily have a
more distorted sex ratio than a place with lower fertility but low levels of gender inequality.
When levels of fertility are imposed rather than chosen, the relationship between gender bias
and sex ratio not only becomes more pronounced, but also more comparable across China. In
the 1980s, the state first initiated its one-child policy. Families have since lost much of their
autonomy to ensure male offspring through the channel of higher-parity births.25 As sex-selective
technology became widely available after late 1980s, families started to rely on ultrasound and
other technology to secure a son in their first birth.26 The strategy of sex selection at a lower25
One-child policy was finally phased out on October 29, 2015. http://www.bbc.com/news/world-asia34665539
26
Depending on the household registration status, urban Chinese were allowed to have one child only, whereas
13
parity birth causes sex ratios within a family to be artificially chosen, contributing to sex ratio
imbalances on a much larger scale at the aggregate level (Ebenstein, 2010).
C Baseline Controls
In the baseline regression, I control for contemporary variables including current per capita
GDP, share of agricultural workforce, share of non-agricultural household registration, share of
ethnic population and provincial capital, historical variables including agricultural suitability,
proximity to the Grand Canal or the Yangtze River and treaty port, as well as a set of geographic
variables, such as latitude, longitude, ruggedness and distance to coast.
I obtain most of contemporary variables from the 2000 Census. The current per capita GDP
is based on the map “2000 county GDP” at the digital map collection of Harvard University.
Historical controls include agricultural suitability, proximity to the Grand Canal or the Yangtze
River and treaty port status. Agricultural suitability data are downloaded from the FAO website, used to proxy a county’s agricultural productivity. Rice and tea suitability used to test
the competing hypotheses are also from FAO. Proximity to the Grand Canal or Yangtze is constructed from CHGIS files.27 Pre-1300 commercial tax quota and historical courier routes used
to test competing hypotheses are similarly from the digital map collection of Harvard University.
Distance to the nearest coast and ruggedness are constructed from NASA data. To account for
cultural and endowment differences across regions, I use Skinner socioeconomic macroregions
(Skinner and Berman, Skinner and Berman) as region fixed effects.
D Descriptive Statistics
I construct my data set as follows. I exclude five autonomous regions, as well as autonomous prefectures and counties in other provinces, that historically comprise ethnic minorities. Descriptive
statistics for the county-level analysis can be found in Table 1.
Table 1 gives an overview of the key variables in the main sample. A total of 1513 counties
are included. Sex ratio at birth is for Year 2000. Most modern variables are for 2000 as well,
unless otherwise noted. About 40% of the counties had some form of cotton textile production
before 1840. Average sex ratio at at birth for 2000 is 118.9 boys per 100 girls, with a standard
deviation of 14.2. Roughly 10% of the counties are on a major trade network (Grand Canal or
Yangtze). An average county has a value of 4.398 on the humidity-for-weaving index, with a
standard deviation of 2.277.
rural Chinese were allowed to have a second child if the first-born was a girl
27
Due to lower transportation costs, a good number of counties located near the Yangtze River and the
Grand Canal produced textiles in pre-modern China. Huang (1964b) emphasizes the importance of the Grand
Canal in Ming China, confirming that many counties famous for textile production were located in the Grand
Canal area, and the size of trade was considerable. This could pose a challenge to my identification strategy,
which I will discuss in the next section.
14
Table 1: Summary Statistics: County-Level Analysis
Variable
Sex ratio at birth
Pre-modern textiles
Mean
119
0.42
Std. Dev.
14.273
0.494
Min.
91.622
0
Max.
193.16
1
N
1535
1535
Logged share of ethnic minorities
Logged per capita GDP
Logged per capita GDP2
Logged share of agriculture workforce
Logged share of non-agriculture registration
Provincial capital
Treaty port
Agricultural suitability
On the Grand Canal or Yangtze
Log (ruggedness+1)
Logged distance to coast
Latitude
Longitude
-0.88
13.246
176.704
3.922
2.845
0.102
0.123
-4.309
0.089
1.193
5.395
31.805
114.224
1.608
1.117
27.184
0.985
0.701
0.302
0.329
1.915
0.285
0.792
1.319
4.801
4.285
-4.605
3.472
12.056
-2.408
1.105
0
0
-8
0
0.078
0.083
20.42
101.584
4.544
18.154
329.583
4.57
4.531
1
1
-1
1
3.083
7.07
41.773
122.391
1533
1530
1530
1535
1535
1535
1535
1535
1535
1535
1535
1535
1535
4.423
2.283
1.562
8.333
1535
Yangtze Delta
Net in migration
Sex ratio, aged 5-9
Sex ratio, aged 1-4
Sex ratio, aged 1-4, 2010
Sex ratio at birth, 2010
Women’s years of schooling
Men’s years of schooling
Years of schooling
Women’s years of schooling, 2010
Men’s years of schooling, 2010
Women’s illiteracy rate, 1990
Men’s illiteracy rate, 1990
0.173
0.012
115.15
121.9
119.046
117.65
7.072
8.164
7.632
8.427
9.326
0.343
0.138
0.378
0.296
10.103
16.352
11.472
10.813
1.023
0.856
0.92
1.141
0.951
0.108
0.062
0
-0.293
95.035
91.655
92.004
86.400
4.48
6.09
5.520
5.560
6.07
0.095
0.018
1
4.733
165.582
204.068
161.656
176.744
11.28
11.91
11.48
12.85
13.39
0.748
0.422
1535
1535
1535
1535
1461
1461
1535
1535
1535
1461
1461
1126
1126
Rice suitability
Tea suitability
Log (#textile companies+1)
#courier routes
Logged (Commercial tax+1), 1077
Log (communicants per 10,000+1)
5.937
88.131
2.79
0.434
2.631
1.53
1.739
132.13
1.626
0.841
3.746
1.095
1
0
0
0
0
0
9
287
7.87
6
10.837
5.254
1503
1503
1535
1535
1535
1038
Humidity-for-weaving index
15
IV
County-Level OLS estimates
Having constructed county-level measures of pre-modern textile production, I can examine the
relationship between pre-modern textiles and gender equality in present-day China. I begin by
examining variation at the county level. My outcome variable is sex ratio at birth. I test my
hypothesis by estimating the following equation:
G
C
Sex ratio at birth = α + βPre-modern textilesc + XH
c Ω + Xc Λ + Xc Π + p ,
(1)
where c denotes a county. Pre-modern textilesc is my measure of pre-modern textile production
G
C
at a county level. XH
c is a vector of historical controls, and Xc and Xc are vectors of geographical
and contemporary controls respectively, each measured at the county level.
H
XG
c and Xc are intended to capture geographic and historical characteristics that may have
been correlated with pre-modern textiles and may still affect present-day outcomes. I control
for whether the county is on the Grand Canal or the Yangtze River, the major trade networks
at the time, as how pre-modern textiles were located was likely influenced by access to market.
To account for geographic differences across counties that may be correlated with openness to
trade, I include in XG
c logged distance to coast and logged (ruggedness+1). As historical China
was an agrarian economy, I include in XG
c agricultural suitability as a proxy for agricultural
productivity as well as latitude and longitude. To account for an external intervention in recent history–the establishment of treaty ports by western powers—I also include “treaty port
status” as a control. To deal with norms such as patrilocality and concern for women’s purity
(Jayachandran, 2015) and other differences across regions, I include regional fixed effects corresponding to Skinner’s Socioeconomic Macroregions. Skinner’s Socioeconomic Macroregions
capture deep-rooted differences across regions, and bisect provincial boundaries in many cases.
The contemporary control variables XC
c include the natural log of a county’s per capita GDP
measured in 2000 and its squared term, share of agriculture workforce, share of non-agricultural
household registration, share of ethnic population, governance status of the prefecture containing the county, governance status of the county, and provincial capital status.28 I use share
of non-agricultural household registration to capture an important source of variation in the
one-child policy.29 Governance status of the prefecture containing the county, governance status
28
The logic behind the inclusion of the squared term of log per capita GDP can be found in Chung and Gupta
(2007). Chinese prefectures and counties underwent institutional reforms after 1982. Level of political centralization started to vary across prefectures and counties. I use two categorical variables to characterize governance
status and degree of centralization of a prefecture or a county. Governance status of the prefecture takes the
value of one when the prefecture has no centralized government; zero when it has a centralized government.
Governance status of the county takes one when it is governed by the prefecture-level government, two when it
is self-governed, and three when it is governed by the province-level government.
29
While the one-child policy was strictly enforced among Chinese citizens on non-agricultural registration
status, a more relaxed version of the one-child policy was enforced among those on agricultural household
registration status.
16
Table 2: Pre-Modern Textiles and Sex Ratio Imbalances: OLS Results
Sex ratio at birth
(2)
(3)
(1)
Pre-modern textiles
Log per capita GDP
Log per capita GDP2
Logged share of agriculture workforce
Logged share of non-agriculture registration
Provincial capital
Historical controls
Geographic controls
Region FE
Province FE
Observations
Adjusted R2
Standard errors in parentheses
∗
p < 0.10,
∗∗
(4)
-3.311∗
(1.754)
2.064∗
(0.988)
-0.144∗∗
(0.0487)
0.736
(0.685)
-5.595∗∗∗
(1.494)
3.060
(2.058)
Yes
Yes
Yes
No
-4.117∗∗
(1.706)
-3.404∗
(1.691)
-3.555∗
(1.686)
Yes
Yes
Yes
Yes
0.508
(0.686)
-5.653∗∗∗
(1.510)
2.009
(2.090)
Yes
Yes
Yes
Yes
0.764
(1.986)
Yes
Yes
Yes
Yes
1528
0.360
1535
0.284
1533
0.353
1535
0.310
p < 0.05,
∗∗∗
p < 0.01
Notes: The table reports the impact of pre-modern textiles on sex ratio imbalances. The
unit of observation is a county in 2000 Census. The dependent variable is sex ratio at
birth. Column 1 reports estimates with all controls along with region and province effects.
“Historical controls” are treaty port status, agriculture suitability, and whether a county
was on the Grand Canal or the Yangtze River (major trade network). “Geographic controls” are log of ruggedness plus 1, log of distance to coast, latitude, longitude and their
interaction. Column 2 shows estimates when all contemporary controls are dropped. Column 3 excludes log per capita GDP on the grounds of potential endogeniety. Column 4
drops most contemporary variables such as share of ethnic population, share of agriculture
workforce and share of non-agriculture registration but keeps provincial capital. Robust
standard errors are clustered at the province level.
17
of the county, provincial capital is intended to capture differences in enforcement of one-child
policy, political control and government-led growth. Both share of non-agricultural household
registration and governance status could have an effect on sex ratios through one-child policy (Ebenstein, 2010). In addition, as there is a clear ethnic and cultural component in son
preference, I control for share of ethnic minority population to reduce composition bias.30
OLS estimates of equation (1) including above controls are reported in Table 2. Column 1
reports estimates with all controls along with region and province fixed effects. Column 2, 3
and 4 reports specifications with potentially endogenous variables excluded. Column 2 shows
estimates when all modern controls are excluded, Column 3 shows estimates when per capita
GDP and its squared terms are excluded from the specification, and Column 4 shows estimates
when only provincial capital is included as a contemporary control.
The estimates show that in counties with pre-modern textile production, fewer girls are missing
today. The coefficient estimates are both statistically significant and economically meaningful.
Based on the estimates from Column 1, one unit increase in pre-modern textile production is
associated with a decrease of sex ratio at birth by 1.62 boys per hundred girls (1.62=3.31*0.49).
As I drop all modern controls, the size of the coefficient further increases to -4.117, suggesting
that some of the modern controls are themselves outcomes of pre-modern textile production.
Column 3 suggests the strength of the coefficient for textile production does not depend on
whether I control for per capita GDP.31 In Column 4, I drop most of modern controls but keep
provincial capital status. The coefficient is rather similar to what is in Column 1-3. This will
be the baseline model for the rest of the county-level analysis .
A Robustness checks
A.1
Subsamples
I first check the robustness of my results to the use of alternative samples. Motivated by the
fact that the Yangtze Delta is of special importance to Chinese economy both historically and
contemporarily, I test to see if my results are robust to the control or the omission of three
provinces (Jiangsu, Zhejiang & Shanghai) from the sample.
Next, I look at counties with different rates of migration. Historically, labor mobility was low
due to the control of the clan system. In modern China, the speed of migration has picked
up. Gender norms in the less developed regions of China could have been strengthened if
individuals with more progressive gender norms are more likely to move to developed areas for a
better life. Hence my results could be biased if textile locations are correlated with unobservable
30
Autonomous counties and prefectures, which are predominately resided and governed by ethnic minorities,
are already excluded from the main sample.
31
Though large sex ratio imbalances are a relatively new phenomenon in China, per capita GDP could still
have already been affected that the sex ratio, i.e. it is possible that per capita GDP is partly endogenous to sex
ratio imbalances (Wei and Zhang, 2011).
18
characteristics of counties that attract many modern migrants. For robustness, I control for net
in-migration, or omit counties with positive net in-migration.
Table A.1 in the appendix summarizes the results. The coefficient estimates are relatively stable
across the columns. Coefficient estimates are greater for counties outside of Yangtze Delta or
areas that have no net in-migration.
A.2
Sex Ratio at Birth from More Years
In the main analysis, I focus on sex ratio at birth in the 2000 Census. A natural question is
whether the same pattern holds for years within the close range of Year 2000. Based on 2000,
2010 census data, I construct two additional variables from the 2000 Census: sex ratio for aged
1 to 4, sex ratio for aged 5 to 9; and two variables from the 2010 Census: sex ratio at birth,
sex ratio for aged 1-4. I find coefficients of pre-modern textiles in Table A.2 are fairly close
to coefficients in Column 5 of Table 2. The coefficient of sex ratio at birth for 2010 is much
smaller and not statistically significant. I attribute this to the data reporting procedures used
in the 2010 Census.32 There is no reason to believe that the true sex ratio at birth suddenly
dropped for the 2010 cohort, but not for cohorts 2005-2009. Overall, I find pre-modern textile
production affects the cohorts born after 1990 in a highly consistent way. The relationship
between pre-modern textiles and sex ratio at birth revealed by the 2000 census is unlikely just
a fluke.
A.3
Other Gender Outcomes
Table A.3 examines alternative gender outcomes. I find pre-modern textile production also
predicts women’s educational attainment. In 2000, pre-modern textile production is associated
with an increase of 0.167 years of schooling for women. The coefficient is reduced to 0.106
after controlling for men’s education. Coefficients are very similar for 2010. Pre-modern textile
production is also associated with a reduction of 3.1 percentage points in female illiteracy rates,
or 1.14 percentage points when male illiteracy rate is controlled for with a p-value of 0.11.33
B IV Estimation
A potential concern with the OLS estimates is that the counties that were textile producers
may have a higher likelihood of adopting textile technologies. It is possible that counties that
were economically more developed were more likely to have adopted textile technologies, and
counties that were closer to the market or transportation routes were more likely to sustain its
production and make greater profits. If these counties were more commercial and had “modern”
32
Due to a booming migrant population, the 2010 Census resorted to the recording method of “recording
every individual encountered”. Double counting of the 200-million floating population became highly probable
events. News sources suggest this method might have opened the door to data manipulation by lower-level
governments, due to the political pressure on “keep the sex ratio at birth below 120” emerging after 2000.
http://blog.people.com.cn/article/1354459332295.html
33
For the 1990 Census, though data on women’s education are available by educational level, they are not
aggregated to the “year of schooling” variable.
19
gender norms, this would bias the OLS estimates away from zero. Though a set of variables
(mainly overall agricultural suitability, distance to the major trade network, distance to the
coast, ruggedness and regional fixed effects) have been included in the main specification and
its variants, I am unable to address likely issues caused by unobservable characteristics, such
as attitudes towards women prior to textile production. Besides, due to imperfect data on premodern textile production, some of the coefficient estimates can suffer attenuation bias due to
measurement error.
To derive an exogenous source in determining the location of textile production within China I
use climate data. An important determinant for the location of textile industry is geo-climatic
conditions. Among all contributing factors, scientists, engineers and industry experts highlight
the importance of relative humidity in producing textiles. In a report on the textile industry in
China (1909), the word “humidity” occurs more than 100 times, suggesting the pivotal role of
humidity in the textile industry.
Climatic conditions were a crucial determinant in the geographic variation in spinning and
weaving in pre-modern China. Spinning and weaving far more likely to occur in humid counties.
Even when spinning and weaving were carried on relentlessly, less humid counties often failed
to produce as high quality cotton fabrics. A lack of humidity severely hampered a county’s
performance in the high-ended markets, as top-notch cotton cloth could only be weaved when
relative humidity was close to or greater than 70%.34 Apart from average relatively humidity,
variance in humidity could also affect production decisions. Textiles could be produced much
more efficiently during parts of the day, and parts of the year that were comparatively humid. For
places that experience greater variance in humidity within the day, the number of hours available
for textile production could be quite limited, regardless of the average relative humidity. A textile
machine represented a large fixed cost. For a family the decision to own textile machinery the
total number of hours possible for textile production would be a key consideration.
The Climate Research Unit of University of East Anglia provides 30-year monthly average
relatively humidity data across 10 arc-minute by 10 arc-minute grid cells globally. I extract
relative humidity values on the basis of x, y coordinates. I construct a relative humidity variable
at the county level by averaging over all relative humidity values within a polygon that represents
a county. I then construct a humidity-for-weaving index as follows. First, every county receives a
score ranging from 1 to 5 for each month, based on the distance between actual relative humidity
for the month and suitable humidity for weaving.35 Second, I add up monthly scores. I get a
34
Gazetteer data suggest that a relatively small number of counties were able to produce cotton textiles in
pre-modern China. By 1840, up to 40% of the counties participated in textile production, varying in quantity
and quality.
35
In the absence of modern humidification facilities, hardly any textiles can be produced when relative humidity drops below 60%, and that the benefit of moist air starts to wear off once relative humidity exceeds 80%, i.e.
there is a non-linear relationship between relative humidity and suitability for textile production. To account for
non-linearity in the impact of relative humidity on historical textile production, I set the lower-bound relative
humidity for feasible production to be 60%, and make it take a value of ”5” if actual relative humidity is below
20
number ranging from 12 to 60 for each county, with 12 being the most suitable, and 60 being the
least suitable. Third, I take the inverse of the total score to build an humidity-for-weaving index
where suitability increases in its value. This index can be seen as approximating the number of
months available for production with a gradient to quality and efficiency. Figure 4 shows the
distribution of humidity-for-weaving index at a county level. Darker shades represent higher
relative humidity and hence, higher weaving suitability. Missing values are shaded white.
Figure 3: IV: Humidity-for-Weaving Index
Ideally, this variable would represent the hours available for textile production. In practice,
data do not exist on the relative humidity for any particular day, let alone variance within a
day. Hence I restrict my focus on the number of months humid enough for textile production as
demonstrated above. A benefit of having a relative humidity index specific to weaving is that
I can extract useful information from relative humidity while avoiding additional biases. Table
3 shows the constructed instrument is correlated with pre-modern textile production, and yet,
uncorrelated with factors normally associated with relative humidity, such as overall agricultural
suitability, tea suitability and tea suitability.
I begin my IV estimation by testing the relationship between my humidity-for-weaving index
and pre-modern textile production. As my textile variable is a binary treatment, I opt for a
Probit-2sls that uses a Probit model for the first stage. Panel A of Table 4 shows the estimates
from the first stage: humidity-for-weaving index is positively correlated with pre-modern textile
the cutoff. Once above 60%, a county will be scored on a lower number as its relative humidity increases, i.e.
”4” for 61%-65%, ”3” for 66%-70%, ”2” for 71%-75%, ”1” for 76% or above.
21
Table 3: The Impact of Humidity-for-Weaving on Other Outcomes
Humidity-for-weaving
index
Baseline controls
Observations
Adjusted R2
Pseudo R2
(1)
Logit
Pre-modern
textiles
(2)
Logit
Provincial
capital
(3)
OLS
Agricultural
suitability
(4)
Logit
On the major
trade network
(5)
OLS
Tea
suitability
(6)
OLS
Rice
suitability
0.152∗∗
(0.0595)
Yes
-0.118
(0.0728)
Yes
0.00328
(0.0676)
Yes
0.0989
(0.0869)
Yes
3.275
(4.540)
Yes
-0.00851
(0.0755)
Yes
1524
1524
1535
0.650
1045
1503
0.800
1503
0.660
0.420
Standard errors in parentheses
∗
0.246
p < 0.10,
∗∗
0.296
p < 0.05,
∗∗∗
p < 0.01
Notes: The table reports falsification tests of humidity-for-weaving index. The unit of
observation is a county in the 2000 Census. The dependent variables are pre modern
textiles, provincial capital status, agricultural suitability, distance to the grand Canal or
Yangtze (on the major trade network), tea suitability and rice suitability. All controls in
Column 4 of Table 2 are included, with the exception of the one that happens to be the
dependent variable in the very specification. Region and province fixed effects are included
in all specifications. Robust standard errors are clustered at the provincial level.
production. Second-stage results are reported in Panel B. Column 1 contains my OLS estimates.
Column 2 report my IV estimate with humidity-for-weaving index being the instrument. My IV
estimate is that a one-standard deviation increase in pre-modern textiles leads to a reduction of
sex ratio at birth by 3.8 boys per hundred girls (3.8=7.61*0.49). This is slightly greater than the
OLS estimate. The increase in coefficient estimates can be partly explained by the IV estimate
by the removal of attenuation bias due to the use of better measured data.
C Competing Hypotheses
C.1
Tea and Rice
Tea and rice production were two other economic activities in which women often participated.
Locations of tea and rice production, in some occasions, overlapped with textile production.
Qian (2008) finds a short-term increase in tea prices can enhance women’s household bargaining
power and increase the share of surviving girls. Perhaps either tea or rice production could
affect gender norms and shape attitudes towards women in the long run and also happen to be
correlated with textile production? I include tea and rice suitability in Column 1 and 2 of Table
A.4 and find no significant change in the coefficient of pre-modern textiles. Rice suitability, in
fact, is positively correlated with sex ratio at birth, confirming the negative role of agricultural
production in women’s status (Hansen et al., 2015).
22
Table 4: Pre-Modern Textiles and Sex Ratio at Birth: Instrumental Variable Analysis
First Stage
Dependent variable: pre-modern textiles
(1)
(2)
Humidity-for-weaving index
N/A
Baseline controls
Region FE
Province FE
Yes
Yes
Yes
0.081 ∗∗
(0.036)
Yes
Yes
Yes
Second Stage
Dependent variable: sex ratio at birth
(1)
(2)
Pre-modern textiles
Same controls as in the first stage
Observations
Standard errors in parentheses
-3.555∗
(1.686)
Yes
-7.617∗∗
(3.993)
Yes
1535
∗
p < 0.10,
∗∗
p < 0.05,
1524
∗∗∗
p < 0.01
Notes: The table reports IV estimates. The unit of observation is a county in the 2000
Census. Humidity-for-weaving index is used as an instrument. The dependent variable
is sex ratio at birth. Baseline controls are the same as in Column 1 and 2 of Table 2.
Column 1 contains OLS estimates. Column 2 report IV estimates. Region and province
fixed effects are included in all specifications. Robust standard errors are clustered at the
province level.
23
C.2
Pre-1300 Commerce
Overall economic development and commercialization, prior to the cotton revolution, could have
both promoted cotton textile production and given rise to progressive gender norms. I include
the pre-1300 commercial tax quota in Column 3 of Table A.4 and find only a small increase in
the size of coefficient of pre-modern textiles. Coefficients of pre-1300 commercial tax and its
square term are insignificant. Coefficient signs indicate a likely non-linear relationship between
past commerce and women’s status.36 In contrast to Bertocchi and Bozzano (2013), I do not
find past commerce per se to be consistently conducive to the status of women.
C.3
State Presence
The history of pre-modern textiles suggests that the state played a key role in the relationship
between cotton textile production and women. As I have noted the default norm that women
worked in textile production was partly due to long-standing state policies that required in-kind
taxes and which were most consistent with the paradigm of “men farm and women weave”. In
addition to being used in lieu of other goods for taxation purposes, cotton textiles were important
for military purposes, and were widely used to clothe the imperial army. Also, the state could
have also influenced the local adoption of new textile technologies by state promotion. Despite
the ubiquitous nature of today’s Chinese state, one might still suspect there is a degree of
persistence in state presence. A strong state presence could mean better enforcements of major
state policies, such as the one-child policy or the compulsory education reform. Therefore,
past state presence can simultaneously influence locations of pre-modern textile production and
gender outcomes in contemporary China. Historically, courier routes in China were used and
managed by the central government for conveyors and messengers to fast deliver high-profile
news (Yang et al., 2006). I include historical courier routes in Column 4 of Table A.4 as a proxy
for state presence. I find state presence is negatively correlated with sex ratio at birth. Greater
state presence does improve relative outcomes of women. However, the coefficient of pre-modern
textiles barely changes when historical courier routes are controlled for, suggesting that the two
effects are most likely independent of each other.
C.4
Industrial Persistence
An obvious hypothesis is that pre-modern textiles could shape gender outcomes through the
persistence in sectoral composition. With the availability of modern humidification technologies,
there is little reason to think textile companies continue to locate only in humid areas. However,
being in a naturally humid area might still attract modern textile companies due to cost-saving
considerations. In addition, human capital accumulated in pre-modern production might find its
36
This is not overly surprising given the well-documented U-shaped female labor force function (Goldin and
Schultz, 1995). Rich merchant families were more able to sacrifice labor incomes of daughters’ or wives’. Footbinding, for example, started from the very well-off families as female members of those families had no need to
work.
24
ways into modern textile production, attracting companies to locate close to where talents are.37
In Column 5 of Table A.4 , I include number of textile companies as a control. I find the scale
of modern textile production indeed reduces sex ratio imbalances. The size of the coefficient
of interest falls by a quarter, but remains highly significant. This suggests pre-modern textile
production does not affect women’s status in contemporary China solely through the channel of
modern textile production. Other explanations would be required.
V
Micro-Level Analysis
I turn to a micro-level analysis that examines variation in women’s representation in highpowered professions and women’s position in the family across individuals, using the 1990 Population Census available at the IPUMS - International. The time frame of the 1990 Census
allows an investigate gender norms just before the planned economy era reached an end. An
additional feature of this census is that they are particularly suitable for studying gender outcomes prior to mass sex selection.38 The 1990 Census also has quite high data quality in the
sense that the size of the floating population was negligible at the time.
I construct the following two outcomes of interest: holding political or managerial positions
and head of the household.39 I construct a binary variable for “holding political or managerial
positions”. The variable takes the value of 1 for “legislators, senior officials and management”, 0
for other occupations. Missing values are assigned in the case that an individual is listed as “NIU
(not in universe)”. The “head of the household variable” takes the value of 1 for individuals
listed as “head of the household”, 0 individuals listed as “spouse”.
Table A.5 in the appendix describes my sample based on the 1990 Census.40 For an average
prefecture, 41% of its population live in places with pre-modern textile production. 1.7% of
the individuals hold political or managerial positions. 56% of the individuals are head of the
household; the rest 44% are listed as a spouse.
37
In this case, current textile production is an intermediate outcome of pre-modern textile production, a.k.a
an endogenous variable.
38
In the mass sex selection era, as part of the gender bias has already been reflected in sex selection, outcomes
of the survivors—a pre-selected group— might not reflect the full extent of gender inequalities in society. Lin
et al. (2014); Hu and Schlosser (2015)
39
The US Census used to have the “Head of the household” and an accordant variable “Relationship to head
of the household”. But it has now switched to “Person 1” and “Relationship to first person listed on the questionnaire”. https://www.census.gov/history/www/throught hed ecades/indexo fq uestions/1980p opulation.html I interpret wife heading the household for a currently married couple to be an indicator of her power in the family.
40
Geographic coverage of my 1990 Census sample is comparable to that of the 2000 Census sample used in my
county-level analysis. More than 98% of the individuals within the geographic coverage are Han Chinese, higher
than the national average of 93%. This is to be expected as five autonomous regions and autonomous counties
are not included in my analysis. In my micro analysis, I further restrict the sample to Han Chinese individuals.
25
My estimation equation is
yi,p =α + βPre-modern textilesp + θFemalei + ζPre-modern textilesp × Femalei
I
C
G
+ XH
p Ω + Xp Λ + Xp Π + Xi Γ + i,p ,
(2)
where p denotes a prefecture.41 My outcome variables are “holding political or managerial
positions” and “head of the household”. Pre-modern textilesp is my prefecture-level measure of
pre-modern textile production.42 My variable of interest is the interaction term between premodern textiles and female. If pre-modern textiles were effective in shaping gender norms in
C
G
favor of women, I should see the interaction term being significant. XH
p , Xp and Xp are the
same controls as in the county-level analysis.43 XIi Γ denotes current individual-level controls:
age group, marital status, employment status and literacy. Robust standard errors are clustered
at the prefecture level for all specifications.
Estimation results based on logit regressions are reported in Table 5. Coefficient estimates of
pre-modern textiles interacted with female are statistically significant for all columns. This is
consistent with the hypothesis that pre-modern textile production enhances women’s position
in society and at home. Coefficients of pre-modern textiles suggest pre-modern textile production has little effect on men’s probability to take political or managerial positions, suggesting
that there is no systematic difference between pre-modern textile production and availability
of political or managerial positions in general. In all specifications, women are far less likely
to either take political or managerial positions, or be the head of the household. This suggests
that despite the socialist laws in favor of gender equality and the rich set of political and economic tools available to the state during the planned economy era, women’s position in society
and at home was still not fully equitable with men’s. In Column 1 and 2, I find pre-modern
textile increases women’s probability of holding political or managerial positions. As I restrict
the sample to individuals living in a households with at least one married couple and who are
currently married in Column 3 and 4, the finding should be interpreted as a wife heading the
household rather than denoting female-headed households comprising women who have never
married or divorced women.
41
In the IPUMS 1990 census data, individual residence is only recorded at the prefecture level.
I aggregate the county-level indicator of pre-modern textile production to the prefecture level weighted
by county population. A prefecture unit is constructed from counties belonging to the sample used in the
county-level analysis.
43 C
Xp are county-level census data aggregated to the prefecture level weighted by county population. For XC
p
most controls from the census year 2000 are replaced with controls from the census year 1990.GDP per capita
2000 is replaced by GDP per capita 1989.
42
26
Table 5: Pre-Modern Textiles and Status of Women : 1990 Census
Political or Managerial Position
(1)
(2)
Pre-modern textiles
Head of the Household
(3)
(4)
-0.00767
(0.0913)
Female
-1.991∗∗∗
(0.0776)
Pre-modern textiles
0.218∗
× Female
(0.118)
Individual controls
No
Contemporary controls
Yes
Historical controls
Yes
Geographic controls
Yes
Region FE
Yes
Province FE
Yes
-0.0130
(0.0798)
-2.018∗∗∗
(0.0795)
0.232∗
(0.122)
Yes
Yes
Yes
Yes
Yes
Yes
-0.313
(0.195)
-5.890∗∗∗
(0.232)
0.629∗
(0.375)
No
Yes
Yes
Yes
Yes
Yes
-0.314
(0.194)
-5.794∗∗∗
(0.233)
0.648∗
(0.372)
Yes
Yes
Yes
Yes
Yes
Yes
Observations
Pseudo R2
2666125
0.195
1815655
0.685
1815655
0.686
2666125
0.172
Standard errors in parentheses
∗
p < 0.10,
∗∗
p < 0.05,
∗∗∗
p < 0.01
Notes: The table reports the impact of pre-modern textiles on women’s position in society
at home. The unit of observation is a individual in the 1990 Census. The dependent
variable is binary. All estimates are based on logit regressions. Contemporary, historical
and geographic controls are the same as in Table 2, but contempoary controls here are
from the census year 1990 instead. Individual controls include age group, marital status,
employment status and literacy. Only married individuals are included in the sample for
Column 3 and 4. Robust standard errors are clustered at the prefecture level.
27
VI
The Emergence and Persistence of Gender
Equality
A The Emergence of Gender Equality
A.1
Widow Survival: Changing Notions of Women’s Role in Society
Unlike the Europe Marriage Pattern (De Moor and Van Zanden, 2010; Voigtländer and Voth,
2013), pre-modern China featured universal marriage and early marriage. Unmarried and married women alike had limited opportunities to participate in society on their own. However,
widows were given a certain amount of autonomy in making economic decisions for the household, despite the overall conservative family and property ownership laws in pre-modern China
(Afeng, 2002). However, prior to the cotton revolution, women typically lacked the means to
support themselves after their husband’s death. Remarriage was quite common. Things took
another turn after the 11th century. Influenced by Song-Ming Neo-Confucianism first developed
in the Song Dynasty (960–1279), on one hand, inheritance laws became more unfriendly to
women, creating barriers for women to inherit wealth from their deceased husband; on the other
hand, remarriage became stigmatized. Both changes greatly limited options available to widows.
The difficult circumstances faced by widows are not unique to pre-modern China. Widows in
developing countries today continue to face more or less similar problems: widows not only lose
the main breadwinner of the household, but also are restricted access to economic resources
due to property ownership laws and employment norms. Many studies document the role of
widowhood in excess mortality for unmarried adult women (Anderson and Ray, 2015; Miguel,
2005; Oppong, 2006; Sossou, 2002).
The cotton revolution in the fourteenth century greatly improved the prospect of widows. Ming
and Qing China witnessed an unprecedented number of widows who participated in a wide range
of economic and social activities. The precondition to widows’ participation in economic and
social activities is survival. Stable incomes derived from textile production played a conducive
role in widows’ survival. Relying on those incomes, widows not only survived, but had the
financial latitude to support their children and in-laws (Zurndorfer, 1998; Sommer, 2000; Elvin,
1984). Another aspect of textile incomes was economic independence. A strong financial position
critically shaped a widow’s status in the family of her deceased husband’s, and sometimes, in
her natal family. Between 1300 and 1850, the improvement of widow well-being, which was
closely related to cotton production, contributed to the broadening of women’s space in society.
(Pomeranz, 2004; Bray, 1997; Pomeranz, 2005; Zhao, 2015).
Data on the number of widows and their mortality in historical times are hard to come by. But
the records counties and prefectures kept on “virtuous women”—a state-sponsored historical
institution to commemorate widows with high morals—can be illuminating on the topic of widow
survival. In the spirit of Song-Ming Neo-Confucianism, women were praised for maintaining
28
female chastity after their husband’s death. Those women were called “virtuous” women and
often documented in local gazetteers for their glorified deeds. Before 1300, among all “virtuous”
women, half of the women were “chaste windows” who provided for her in-laws and children
for an number of decades, the other half were “heroic widows” who committed suicide upon
their husband’s death to demonstrate their exemplary character (Jiazun, 1979). After 1300,
cotton textiles began to financially empower women. The percentage of women who chose
chaste widowhood over suicide likely increased. I hypothesize that cotton textiles tilted women’s
decision towards chaste widowhood from suicide, as availability of financial means was key to
widow survival.44 All else equal, women with no financial means would be at a higher risk to
commit suicide.
To test the relationship between cotton textile production and widow suicide, I search local
gazetteers for records on “virtuous” women. To circumvent the problem of varying local standards of awarding “virtuous women” status, I focus on women awarded imperial testimonials of
merit (jingbiao) by the central government. To have a sense of the timing of transition—from
when cotton textiles began to positively shape women’s life trajectories—I start with records of
jingbiao from the Ming Dynasty (1368-1644). Because I perform the search on prefecture-level
gazetteers available on zhongguo fangzhi ku (China’s Gazetteer Database), Series I, I restrict the
sample to prefectures that had at least one prefecture-level gazetteer from the Ming Dynasty
available in the database. Based on the above criteria, a total of 89 prefectures are included in
the sample. Table 6 shows that 41 out of 89 prefectures have pre-modern textile production.
Table 6: jingbiao: Summary Statistics
Variable
Mean Std. Dev. Min.
Widow suicide
0.438
1.588
0
Chase widowhood
1.815
4.617
0
jingbiao
2.253
5.411
0
Pre-modern textiles
0.466
0.461
0
Pre-modern textiles, binary 0.461
0.501
0
Latitude
30.395
4.73
19.193
Longitude
115.129
4.331
104.048
Max.
10
34
38
1
1
40.015
121.409
N
89
89
89
89
89
89
89
These numbers supply qualitative evidence that suggests that pre-modern textile production
decreased the share of “virtuous” women committing suicide. In other words, women in regions
where cotton textile production was more important had a greater chance of survival, holding
constant their “virtuous women” status. In Column 3, A one-standard-deviation increase in
44
To be awarded “chase widow” status, a long wait is required. According to Qing regulations, to be eligible
to the title of “ chase widow”, a woman either had to remain widowed since before the age of 30 years old to
the age of 50 years old, or had been widowed for ten years or more but died before reaching 50 (Mann, 1987).
The long time frame required to be eligible to the “chase widow” status heightened the importance of having
financial resources at one’s disposal. Here I do not try to argue that having financial means was the single key
factor in widows’ decision making; I acknowledge that many factors could be at play (Theiss, 2005; Ropp et al.,
2001).
29
pre-modern textile production is associated with a decrease of 0.16 (0.32*0.51) search records
on widow suicide, or 37% of the mean of widow suicide records. Cotton textiles enabled women
to maintain a livelihood in the absence of their husband. From the perspective of parents, a
daughter’s ability to support herself under adverse circumstances reduces their mental and financial exposure in a world characterized by uncertainty.45 I argue that the image of a financially
empowered, capable and dependable unmarried adult woman, can lead to a notion that women
can be productive and independent members of society.
B Gender Norms and Female Labor Force Participation: Evidence from the Early Twentieth
Century
From 1840 onwards, China began to industrialize at a slow pace. Manufacturing jobs emerged
and they typically required workers to work outside home. Conservative gender norms and
concern for “purity” of women would predict few women would take manufacturing jobs. In
reality, female labor force participation was extremely uneven across regions.The presence of
women in industrial plants was much more common in Jiangsu, Zhejiang and Shanghai, where
women even outnumbered men. Women working outside the home was extremely rare in Zhili,
Shanxi and Shaanxi. Would pre-modern textiles influence women’s decision to participate in the
labor force? Ideally, I would like having disaggregated data at the very onset of industrialization
to examine how women’s initial responses to industrial job opportunities differed. Unfortunately,
a still largely pre-modern and agrarian Chinese state at the time, did not possess the capacity
to collect detailed labor statistics. One of the earliest censuses available that can help answer
this question is the 1916 Economic Census. The census documents number of male and female
workers working in a factory by province and industry, excluding household production workers.
Table ?? provides summary statistics. As seen in the table, in an average province-industry
pair, roughly 19% of the total workers were female, but with high variance.
Table 7: Summary Statistics: Share of Female Workers in 1916
Variable
Pre-modern textiles
%Female
#workers
Log(#workers+1)
Mean Std. Dev.
0.464
0.33
0.189
0.252
3839.652 11057.267
6.506
2.015
Min. Max.
N
0.007 0.921 181
0
1
181
7
123127 181
2.079 11.721 181
To investigate the role of pre-modern textile production in female labor force participation,
I regress share of female workers on pre-modern textiles. Table 8 suggests that in provinces
with a higher percentage of population exposed to pre-modern textiles, a large share of women
worked in factories. A few possible explanations are (a.) Persistence in specific skills. Women
45
Bossler (2000) finds evidence for a continued relationship between a married woman and her natal family.
While a woman became a member of her husband’s extended family upon marriage, her natal family could still
be involved in times of crisis. This includes cases in which a widowed woman in poverty imposed a financial
burden on her natal family.
30
Table 8: 1916 Economic Census: Pre-modern Textile and Share of Female Workers
(1)
Pre-modern textiles
Share of Female Workers
(2)
(3)
0.156∗∗∗
(0.0525)
0.124∗∗∗
(0.0450)
No
Yes
0.0844∗
(0.0448)
0.0445∗∗∗
(0.0134)
Yes
181
0.0362
181
0.266
181
0.338
Log(#workers+1)
Industry dummies
Observations
Adjusted R2
Standard errors in parentheses
∗
p < 0.10,
∗∗
p < 0.05,
∗∗∗
p < 0.01
Notes:The table reports the impact of pre-modern textiles on share of female workers in the
early 20th century. The unit of observation is an industry within a province. The dependent
variable is share of female workers. #workers referes to the total number of workers in an
industry within a province. Robust standard errors are used in all specifications.
who understood household production of textiles had an advantage in industrial production
of textiles. (b.) Persistence in physical mobility. Pre-modern textiles provided women with
opportunities to engage in market exchange. Women likely enjoyed a higher level of physical
mobility than their counterparts in other places. (c) Persistence in the role of bread-winning
females. Families used to incomes generated by women had to adapt to new economic realities
that manufacturing jobs were better jobs for women to support a family.
I rule out (a.) being the only explanation by showing that a larger share of female workers was
found not only in textile manufacturing plants, but also in other industries. In fact, pre-modern
textile production is positively correlated with the share of female workers in most industries,
except for fur making. Figure 4 illustrates this point with a scatter plot and a fitted regression
line for each of cotton textile manufacturing, knitting, dyeing, and match manufacturing. To
further distinguish between (b.) and (c.) I would need higher quality data. Either (b.) or
(c.) would be consistent with the hypothesis that pre-modern textile production generated
gender norms in favor of women and influenced a range of later outcomes through the channel
of reformed gender norms.
These results could suggest an alternative mechanism for pre-modern textile production to affect
modern-day outcomes: places that had more equal gender norms may have had more rapid
industrialization aided by abundant female labor.46 Once the economy began to industrialize,
growth and development can become proximate causes of better relative outcomes for women. To
deal with, in the next section, I explicitly control for early industrialization as well as investigate
its impact on the persistent effects of pre-modern textiles.
46
Previous studies have shown the effect of proto-industry on the locations of modern industries (Farnie, 1979)
31
Figure 4: Share of Female Workers in 1916
C The Persistence of Gender Norms after 1840
C.1
Resilience to Large Political and Economic Shocks
Early Industrialization I try to address potentially uneven effects of the industrialization
and modernization processes across counties. China began to industrialize from the 19th century
onwards, first in treaty ports. Jia (2014) shows that treaty ports had a long-lasting impact on
local economies. This is a potential source of bias if pre-modern textile locations overlapped
with areas that experienced early industrialization, as gender norms might be affected by such
drastic economic and social change. Historical evidence suggests that this should not be a
major concern as industrialization in China was extremely limited and highly isolated (Fairbank,
1978).47 Column 1 and 2 of Table 9 shows coefficient estimates of textile production are robust to
controlling for or omitting treaty ports. Treaty port status has an independent effect on reducing
sex ratio imbalances. The interaction term between treaty port and pre-modern textiles is close
to zero and insignificant, suggesting pre-modern textile production has no differential impact on
modern sex ratio imbalances by treaty port status. The coefficient size of pre-modern textiles
increases slightly when all treaty ports are excluded in Column 2.
47
During the late Qing and Republican China era, much of the rural and hinterland China continued to
perform household production and their traditional lifestyles.
32
Missionary Influence Missionaries came to China to spread Christian religions in the 19th
Century. They built churches, schools and hospitals. While only a small percentage of Chinese
populations were converted, some of them might have had a disproportional influence on the
rest of society. Since Christianity emphasizes the value of life, and specifies different gender
norms from traditional Chinese religions, missionary activities might have changed local gender
norms. To check this, I first include logged number of communicants per 10,000 plus 1 in the
regression in Column 5. As expected, the percentage of believers in the population is positively
correlated with a more normal sex ratio. The coefficient size of pre-modern textiles increases
by roughly one-fifth, suggesting some of the effects of pre-modern textile production might
have been previously masked by missionary activities in counties with no history of cotton
textiles. I then include both logged number of believers per 10,000 plus 1 and its interaction
with pre-modern textiles in Column 6. The positive sign of the interaction term hints at the
possibility that the effect of pre-modern textile production is less persistent where the percentage
of believers is high, but its coefficient is insignificant.
33
34
∗∗
p < 0.05,
∗∗∗
1038
0.305
Yes
Yes
Yes
-0.886
(0.553)
0.128
(2.755)
-3.748∗∗
(1.600)
-3.300
(3.298)
p < 0.01
1038
0.306
Yes
Yes
Yes
-1.333∗∗
(0.586)
1.143
(0.996)
0.151
(2.705)
-5.583∗∗
(1.897)
-3.536
(3.091)
(3)
(4)
Missionary
influence
1535
0.310
Yes
Yes
Yes
-0.473
(1.721)
2.031
(2.643)
-3.743∗∗
(1.716)
-5.129
(3.154)
1379
0.307
Yes
Yes
Yes
-3.513∗
(1.832)
-5.580
(3.330)
1535
0.312
-4.097
(2.436)
4.330∗
(2.187)
Yes
Yes
Yes
0.833
(1.901)
-4.207∗∗
(1.723)
-5.087
(3.036)
1308
0.301
Yes
Yes
Yes
0.401
(1.909)
-4.704∗∗
(1.758)
-8.428∗∗
(3.245)
(7)
(8)
Economic
reforms
1038
0.306
-6.123∗∗∗
(1.876)
-3.378
(2.252)
-0.406
(4.390)
-2.107
(3.025)
3.679
(4.400)
-1.330∗∗
(0.577)
0.982
(0.938)
-4.364
(2.992)
4.264
(3.495)
Yes
Yes
Yes
(9)
All
shocks
1038
0.303
Yes
Yes
Yes
0.0645
(2.731)
-3.842∗∗
(1.623)
-3.613
(3.389)
(10)
No
shock
Notes: The table reports the impact of intermediate shocks on sex ratio imbalances as well as on the persistent effects of pre-modern
textiles. The unit of observation is a county in 2000 Census. The dependent variable is sex ratio at birth. Baseline controls are
those used in Column 4 of baseresults. Christianity is measured by log (communicants per 10,000+1). “on the coast” refers to a
county within 50 kilometers of the coast. Column 2 drops all treaty ports. Column 6 drops all provincial capitals. Column 8 drops
all coastal counties. Column 9 include all interaction terms. Column 10 includes only baseline controls and uses the same sample as
Column 9. As part of the baseline controls, treaty port status and provincial capital are controlled for in all specifications. Robust
standard errors are clustered at the province level.
p < 0.10,
∗
Standard errors in parentheses
1346
0.331
1535
0.310
Observations
Adjusted R2
Yes
Yes
Yes
-0.725
(1.682)
-3.779∗∗
(1.519)
Yes
Yes
Yes
-3.517∗∗
(1.621)
-4.796∗∗∗
(1.443)
-0.371
(4.590)
0.785
(2.113)
Pre-modern textiles
× On the coast
Baseline controls
Region FE
Province FE
Pre-modern textiles
× Christianity
On the coast
Pre-modern textiles
× Provincial capital
Christianity)
Pre-modern textiles
× Treaty port
Provincial capital
Treaty port
Pre-modern textiles
(2)
Early
industrialization
(1)
Sex ratio at birth
(5)
(6)
State
socialism
Table 9: Persistence since 1840: Intermediate Shocks
Post-1949 State Socialism China has been promoting gender equality through laws, policies
and institutions for over half a century (Johnson, 2009). As gender equality has an important
role in the communist ideal, China passed the marriage law in 1950 to grant women the right
to free marriage and divorce, inherit property, and control of their children. The role of Chinese
women changed from a “family private person” of traditional society to a “social person”, and
Chinese women gained the same legal status as men. The Constitution of the People’s Republic
of China enacted in 1954 expressly stated that women and men enjoy equal rights. China also
mandated equal entry to the labor market and instituted equal pay for equal work for men and
women (Entwisle and Henderson, 2000; Hannum and Xie, 1994; Johnson, 2009; Yang, 1999),
and those equalization policies worked relatively well during the planned economy era. Despite
uneven economic growth, contemporary China has kept most formal institutions that guarantee
gender equality.
Post-1949 socialist policies have undoubtedly shrunk the absolute size of gender gap and transformed gender norms to a large extent. However, it is less clear why post-1949 socialist polices
should be more intense in places with pre-modern textile production. There is little local variation in formal institutions of China because of its high degree of political centralization.48 A
potential source of variation in the enforcement of socialist policies is whether a county is in a
provincial capital. If this is true, the effects of pre-modern textiles could be less persistent in
provincial capitals where political shocks were larger. I drop provincial capitals in Column 4 and
find the coefficient stays almost the same. In column 3, I include the interaction term between
pre-modern textiles and provincial capital and find it to be slightly positive but not significant.
Recent Economic Reforms In the midst of post-1979 economic reforms, the state has relaxed part of its control on the economy and society. Previously hidden gender inequality has
since surfaced (Li and Lavely, 2003). On one hand, there remains very little variation in either labor laws or maternity leave law at a local level. On the other hand, the recent thirty
years of growth in China could have led some regions to develop temporary rules or measures
that increase or decrease gender equality as a byproduct of economic growth. I am not overly
concerned with recent manifestations of already existing gender bias, as this could just be endogenous to pre-modern textiles. To address the effect of heterogeneous economic growth and
economic institutions in recent years, I control for per capita income in county-level regressions
(Column 1 and 2 of Table 2), along with other contemporary controls. To see if economic reforms are associated with differential persistent effects of pre-modern textiles, I interact coastal
counties with pre-modern textiles in Column 7.49 I find that large economic shocks appearing
in the coast area might indeed weaken the effect of pre-modern textiles. Yet, the insignificant
48
Urban authorities in China have little or no ability to shape labors laws and policies at a local level. Rural
China does have more policies generated through democratic processes at the local level (O’brien and Li, 2000).
49
China’s export-led economy in the past 30 years has rendered coastal regions a tremendous growth advantage. Yangtze Delta and Pearl River Delta are home to million of exporters.On the side of policy interventions,
the central government created the earliest special economic zones all on the coast too.
35
coefficient means that no reliable interpretation can be inferred from these results.50 In Column
8 I simply drop all coastal regions, and find the coefficient size of interest increases by roughly a
third. This suggests that pre-modern textile production has a larger effect of reducing sex ratio
imbalances in the non-coastal region than in the coastal region.
I include controls for large political and economic shocks and interacted them with pre-modern
textiles in Column 9. This results in a sample of 1038 counties. I find pre-modern textiles
reduces sex ratio imbalances by 6.1 boys per hundred girls in places with no exposure to any
of the large political and economic shocks, when those shocks are taken as exogenous. Within
the same sample, in Column 10, pre-modern textile production reduces sex ratio imbalanced by
3.8 boys per hundred girls. The analysis of those shocks suggests that the persistent effects of
pre-modern textile production are highly resilient.
D Gender-Role Attitudes and Son Preference in Contemporary China: Evidence from CGSS
Thus far I have reviewed the history of pre-modern textile production advancing women’s position and shaping gender norms, and examined a number of outcomes closely related to gender
norms but not studied gender attitudes themselves. Now I directly turn to direct evidence on
gender norms. CGSS 2010 (Chinese General Social Surveys) provides an unprecedented opportunity to examine gender-role attitudes among the Chinese. The survey contains questions
regarding beliefs and attitudes associated with women. In addition, the CGSS includes information on age group, gender, urban/rural site, marital status, education attainment, party member
status and household registration status.
The first measure of beliefs about women is constructed from each respondent’s view of the
following question: “Do you agree with the following statement: men are naturally more capable
than women?” The second measure comes from the question: “Do you agree with the following
statement: men should focus on career; women should focus on family?” The respondent can
choose from a scale of 1 to 5 ranging from “completely disagree” to “completely agree”. In
addition, I create a measure from two questions on the subjective assessment of how many sons
and daughters one wants to have. For those who answer they want more sons than daughters, son
preference takes on the value of 1. For those who are indifferent between sons and daughters, or
want the same number of sons and daughters, or want more daughters than sons, son preference
takes on the value of 0. I take a subsample of CGSS to match the geographic coverage of the
main sample used in Section IV. Summary statistics are available in Table A.6.
50
Besides, heterogeneity in the treatment itself can not ruled out as an alternative interpretation here. While
emperors between 1300 and 1850 banned ocean trade periodically between 1300 and 1850, the treatment of
pre-modern textiles differ in magnitude and intensity for coastal regions.
36
37
∗
Standard errors in parentheses
∗∗
p < 0.05,
5585
0.093
∗∗∗
p < 0.01
5578
0.098
-0.190∗∗∗
(0.0677)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
5594
0.035
-0.0883
(0.0734)
No
No
No
No
No
Yes
Yes
Yes
Yes
5591
0.109
-0.125∗
(0.0675)
Yes
Yes
Yes
Yes
No
Yes
Yes
Yes
Yes
5584
0.116
-0.117∗
(0.0664)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
5533
0.042
0.034
-0.321∗
(0.184)
Yes
Yes
Yes
Yes
No
Yes
Yes
Yes
Yes
5535
-0.307∗
(0.176)
No
No
No
No
No
Yes
Yes
Yes
Yes
0.045
5525
-0.324∗
(0.185)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Son preference
(7)
(8)
(9)
Logit
Notes: The table reports the impact of pre-modern textiles on gender-role attitudes and son preference. The unit of observation
is a survey respondent in CGSS 2010 (Chinese General Social Surveys). Contemporary, historical and geographic controls are the
same as in Table 2. Robust standard errors are clustered at the county level.
p < 0.10,
5587
0.017
Observations
Adjusted R2
Pseudo R2
Pre-modern textiles
-0.169∗∗ -0.199∗∗∗
(0.0663) (0.0646)
Age group
No
Yes
Female
No
Yes
Education
No
Yes
Female ×Education
No
Yes
Individual controls
No
No
Contemporary controls
Yes
Yes
Historical controls
Yes
Yes
Geographic controls
Yes
Yes
Region FE
Yes
Yes
Men naturally more capable
Women focus on family
(1)
(2)
(3)
(4)
(5)
(6)
OLS
Table 10: Gender-role Attitudes and Son Preference in Contemporary China:
Evidence from CGSS
I regress pre-modern textile production on beliefs about women and son preference with the
same controls used in Column 1 and 2 Table 2 as well as in Table 5. Robust standard errors
are clustered at the county level for all specifications. Table 10 summarizes the results. Column
1 through 6 report OLS results for beliefs about women. Column 7 through 9 focus on son
preference. Columns 1, 4, 7 contain no individual controls; Columns 2, 5, 8 include basic
individual controls such as age group, gender and education attainment; Columns 3, 5, 9 contain
a full set of individual controls. Individuals in counties with pre-modern textile production are
more likely to disagree with the statement that men are naturally more capable, or the statement
that women should focus on family, and less likely to favor sons over daughters. Pre-modern
textile production is systematically correlated with more progressive gender norms and weaker
son preference.
VII
Conclusion
Women’s productivity in textile production varied from society to society historically, and it
often depended on the technology and local geo-climatic conditions. Anthropologists posit that
the shift from flax to wool in the ancient Middle East led to a decline in the status of women
as linen was cultivated on a small scale by women and children while wool production relied
on male management of sheep herds (McCorriston, 1997). In historical China, the switch from
linen to cotton empowered women (Bray, 1997).
This paper provides evidence that a portion of the variation in gender norms and gender inequality in modern day China can be accounted for by pre-modern textile production. It suggests
that gender norms can be shaped by large and long-lasting relative productivity shocks.
I use both OLS and IV to estimate the impact of pre-modern textile production on today’s sex
ratio imbalances. The results are robust to the exclusion of Yangtze Delta, a region famous
for pre-modern textile production and a number of features associated with a highly developed
historical economy. I also extend my analysis to include other variables more commonly discussed
in the context of gender equality, such as literacy and education. My finding from county-level
regressions cannot be explained by past tea or rice production, commerce prior to the period of
cotton textile production, state presence, current textile industry, or modernization hypothesis.
My micro-level analysis lends support to my county-level analysis, and generates additional
insights into how gender bias worked itself into economic and social outcomes during the planned
economy era. I find that pre-modern textile production helps to explain women’s position even
in the socialist period of China. In both 1982 and 1990 census, pre-modern textile production
is positively associated with share of women with political or managerial positions and wife
heading the household.
In addition to analyzing gender norms in contemporary China, I also look at the impact of
pre-modern textiles on the gender norms and gender equality in the past. I find evidence
38
for an adaptation in gender norms to the cotton revolution at latest by the end of the Ming
Dynasty (1366-1644). Pre-modern textile production likely reduced the rate of widow suicide
and increased quality of life, self-esteem and social status of widows at the time. Based on
the fact that widows, compared to married women, were more poised to making independent
decisions and as well as being visible in public space, I conclude the elevated status of widows
might have contributed to a more positive evaluation of women’s ability and their role in society.
Another historical period I examine is the onset of industrialization in modern China. I find premodern textile production is positively associated with share of female workers in manufacturing
jobs. I interpret this as a reflection of more relaxed gender norms in places with pre-modern
textile production that allowed more women to work outside home to take advantage of new
economic opportunities.
I examine a number of large political and economic shocks and their impact on the transmission
of gender norms since China began to modernize in 1850. After acknowledging the role each of
those shocks in gender bias reflected in sex ratio imbalances, I find coefficients of pre-modern
textile production remain significant and similar in magnitude to baseline estimates. The persistence effects of pre-modern textiles are overall highly resilient; coefficients of pre-modern textiles
interacted with those shocks do provide suggestive evidence that those intermediate shocks might
have served to temporarily or permanently weaken the persistence effects of pre-modern textiles on gender norms. Had there been none of those shocks, pre-modern textile production
might have reduced sex ratio imbalances by six boys per hundred girls rather than five boys per
hundred girls.
Finally, I investigate gender-role attitudes Chinese people hold in contemporary China, and I
find pre-modern textile production is systematically correlated with more progressive gender
norms and weaker son preference.
39
References
Afeng (2002). Huizhou wenshu suo jian mingqing shidai funv de diwei yu quanli(Status of Women in
Ming and Qing China from Huizhou Archives). Ph. D. thesis.
Alesina, A., P. Giuliano, and N. Nunn (2013). On the origins of gender roles: Women and the plough.
The Quarterly Journal of Economics 128 (2), 469–530.
Allen, R. C. (2009). Agricultural productivity and rural incomes in England and the Yangtze delta,
c.1620ñc.1820. Economic History Review 62 (3), 525–550.
Allen, R. C., J.-P. Bassino, D. Ma, C. Moll-Murata, and J. L. Van Zanden (2011). Wages, prices, and
living standards in China, 1738–1925: in comparison with Europe, Japan, and India. The Economic
History Review 64 (s1), 8–38.
Almond, D., H. Li, and S. Zhang (2013, June). Land reform and sex selection in China. Working Paper
19153, National Bureau of Economic Research.
Anderson, S. and M. Eswaran (2009). What determines female autonomy? evidence from Bangladesh.
Journal of Development Economics 90 (2), 179–191.
Anderson, S. and D. Ray (2015). Missing unmarried women. Technical report, National Bureau of
Economic Research.
Ashraf, N., D. Karlan, and W. Yin (2010). Female empowerment: Impact of a commitment savings
product in the Philippines. World development 38 (3), 333–344.
Barber, E. J. (1991). Prehistoric textiles. The development of cloth in the Neolithic and Bronze Ages
with special reference to the Aegean, 223–243.
Bertocchi, G. and M. Bozzano (2013). Women, medieval commerce, and the education gender gap.
Bertrand, M., E. Kamenica, and J. Pan (2015). Gender identity and relative income within households.
The Quarterly Journal of Economics, qjv001.
Bisin, A., E. Patacchini, T. Verdier, and Y. Zenou (2011, May). Formation and persistence of oppositional identities. CEPR Discussion Papers 8380, C.E.P.R. Discussion Papers.
Bisin, A. and T. Verdier (2001, April). The economics of cultural transmission and the dynamics of
preferences. Journal of Economic Theory 97 (2), 298–319.
Bossler, B. J. (2000). ” a daughter is a daughter all her life”: Affinal relations and women’s networks
in song and late imperial China. Late Imperial China 21 (1), 77–106.
Bray, F. (1997). Technology and gender: Fabrics of power in late imperial China. University of
California Pr.
Burda, M., D. S. Hamermesh, and P. Weil (2007, March). Total Work, Gender and Social Norms.
NBER Working Papers 13000, National Bureau of Economic Research, Inc.
Chen, J., S. Wu, Y. Zhou, and X. Qian (1991). Songjiang qu ge (Poetry of Songjiang). Zhong hua shu
ju.
Chen, Y., H. Li, and L. Meng (2013). Prenatal sex selection and missing girls in china: Evidence from
the diffusion of diagnostic ultrasound. Journal of Human Resources 48 (1), 36–70.
Chow, R. (1991). Woman and Chinese modernity: The politics of reading between West and East. U
of Minnesota Press.
Chung, W. and M. D. Gupta (2007). The decline of son preference in South Korea: The roles of
development and public policy. Population and Development Review 33 (4), 757–783.
Daul, G. B. and E. Moretti (2008, Oct). The demand for sons. Review of Economic Studies 75 (4),
10851120.
De Moor, T. and J. L. Van Zanden (2010). Girl power: the European marriage pattern and labour
markets in the North Sea region in the late medieval and early modern period. The Economic History
Review 63 (1), 1–33.
Deininger, K., A. Goyal, and H. Nagarajan (2010). Inheritance law reform and women’s access to
capital: evidence from india’s hindu succession act.
Ebenstein, A. (2010). The “missing girls” of China and the unintended consequences of the one child
policy. Journal of Human Resources 45 (1), 87–115.
40
Ebrey, P. (1990). Women, marriage, and the family in chinese history. Heritage of China, 197–223.
Edlund, L. (1999). Son preference, sex ratios, and marriage patterns. Journal of Political Economy 107 (6), 1275–1304.
Elvin, M. (1984). Female virtue and the state in China. Past and Present, 111–152.
Entwisle, B. and G. Henderson (2000). Re-drawing boundaries: work, households, and gender in China,
Volume 25. Univ of California Press.
Fairbank, J. K. (1978). The Cambridge History of China: Late Ching, 1800-1911, pt. 2. Cambridge
University Press.
Farnie, D. (1979). The English Cotton Industry and the World Market: 1815-1896. Oxford University
Press.
Fernández, R. (2007). Alfred marshall lecture women, work, and culture. Journal of the European
Economic Association 5 (2-3), 305–332.
Fernández, R. and A. Fogli (2009). Culture: An empirical investigation of beliefs, work, and fertility.
American Economic Journal. Macroeconomics 1 (1), 146.
Flynn, D. O. and A. Giráldez (1995). Born with a” silver spoon”: The origin of world trade in 1571.
Journal of World History, 201–221.
Fortin, N. M. (2005). Gender role attitudes and the labour-market outcomes of women across OECD
countries. Oxford Review of Economic Policy 21 (3), 416–438.
Goldin, C. and T. Schultz (1995). The u-shaped female labor force function in economic development
and economic history. Investment in Womens Human Capital and Economic Development.
Goldstone, J. A. (1996). Gender, work, and culture: Why the industrial revolution came early to
England but late to China. Sociological Perspectives 39 (1), 1–21.
Greif, A. and G. Tabellini (2010). Cultural and institutional bifurcation: China and europe compared.
The American economic review , 135–140.
Greif, A. and G. Tabellini (2015). The Clan and the City: Sustaining Cooperation in China and
Europe. Working Paper 445, IGIER (Innocenzo Gasparini Institute for Economic Research).
Grosjean, P. (2011, December). A History of Violence: The Culture of Honor as a Determinant of
Homicide in the US South. Discussion Papers 2011-13, School of Economics, The University of New
South Wales.
Grosjean, P. and R. Khattar (2014). It’s raining men! hallelujah?
Gu, Y. (1995). Zhao yu zhi:Jiangnan, Volume 9 (reprint ed.). Xu xiu si ku quan shu.
Guiso, L., P. Sapienza, and L. Zingales (2006). Does culture affect economic outcomes? The Journal
of Economic Perspectives 20 (2), 23–48.
Gupta, B. (2014, Feb). Where have all the brides gone? son preference and marriage in India over the
twentieth century. The Economic History Review 67 (1), 1–24.
Hannum, E. and Y. Xie (1994). Trends in educational gender inequality in China: 1949-1985. University
of Michigan.
Hansen, C., P. Jensen, and C. Skovsgaard (2015). Modern gender roles and agricultural history: the
neolithic inheritance. Journal of Economic Growth, 1–40.
Harrell, S. (1995). Chinese historical microdemography, Volume 20. Univ of California Press.
Hu, L. and A. Schlosser (2015). Prenatal sex selection and girls well-being: Evidence from india. The
Economic Journal .
Huang, P. C. C. (1990). The peasant family and rural development in the Yangzi Delta, 1350-1988.
Stanford: Stanford University Press.
Huang, R. (1964a). The Grand Canal During the Ming Dynasty: 1368-1644. University Microfilms.
Huang, R. (1964b). The Grand Canal During the Ming Dynasty: 1368-1644. University Microfilms.
Jayachandran, S. (2015). The roots of gender inequality in developing countries. Annual Review of
Economics.
Jha, S. (2013). Trade, institutions, and ethnic tolerance: Evidence from South Asia. American Political
Science Review 107 (04), 806–832.
Jia, R. (2014). The legacies of forced freedom: China’s treaty ports. Review of Economics and Statistics.
forthcoming.
41
Jiazun, D. (1979). A Statistical Analysis of Virtuous Women by Dynasty. Ph. D. thesis.
Johnson, K. A. (2009). Women, the family, and peasant revolution in China. University of Chicago
Press.
Kang, C. (1977). The development of cotton textile production in China. Harvard East Asian Monographs 74.
Lee, J. Z. and C. D. Campbell (2007). Fate and Fortune in Rural China: Social Organization and
Population Behavior in Liaoning 1774-1873, Volume 31. Cambridge University Press.
Li, B. (1997). ”men farms and women weaves” and the formation of the role of ”half the sky” of
women in peasant family economy: a study of women’s work in jiangnan during the ming-qing
times. Researches in Chinese Economic History 3, 001.
Li, B. and P.-C. Li (1998). Agricultural development in Jiangnan, 1620-1850. Macmillan.
Li, J. and W. Lavely (2003). Village context, women’s status, and son preference among rural Chinese
women. Rural Sociology 68 (1), 87–106.
Lin, M.-J., J.-T. Liu, and N. Qian (2014). More missing women, fewer dying girls: The impact of sexselective abortion on sex at birth and relative female mortality in taiwan. Journal of the European
Economic Association 12 (4), 899–926.
Ma, D. (2005). Textiles in the Pacific, 1500-1900. Ashgate/Variorum.
Man, C. O. (2011). Socioeconomic history of the Ming and Qing Dynasty from a gender perspective:
the histories of agriculture, textile, commerce, and consumers. Journal of Chinese Studies 52 (01),
95—127.
Mann, S. (1987). Widows in the kinship, class, and community structures of Qing dynasty China.
Journal of Asian Studies 46 (1), 37–56.
Mann, S. (1997). Precious records: Women in China’s long eighteenth century. Stanford University
Press.
McCorriston, J. (1997). Textile extensification, alienation, and social stratification in ancient
mesopotamia. Current Anthropology 38 (4), 517–535.
Miguel, E. (2005, October). Poverty and witch killing. Review of Economic Studies 72 (4), 1153–1172.
Moen, P., M. A. Erickson, and D. Dempster-McClain (1997). Their mother’s daughters? the intergenerational transmission of gender attitudes in a world of changing roles. Journal of Marriage and the
Family, 281–293.
Nunn, N. (2012). Culture and the historical process. Economic History of Developing Regions 27 (sup1),
S108–S126.
Nunn, N. and L. Wantchekon (2011, December). The Slave Trade and the Origins of Mistrust in Africa.
American Economic Review 101 (7), 3221–52.
O’brien, K. J. and L. Li (2000). Accommodating “democracy” in a one-party state: Introducing village
elections in China. The China Quarterly 162, 465–489.
Oppong, C. (2006). Familial roles and social transformations older men and women in sub-saharan
africa. Research on Aging 28 (6), 654–668.
Oster, E. (2005). Hepatitis b and the case of the missing women. Journal of Political Economy 113 (6),
1163–1216.
Pomeranz, K. (2002). Beyond the East-West binary: Resituating development paths in the eighteenthcentury world. The Journal of Asian Studies 61 (02), 539–590.
Pomeranz, K. (2004). Women’s work, family, and economic development in Europe and East Asia:
long-term trajectories and contemporary comparisons. In G. Arrighi, T. Hamashita, and M. Selden
(Eds.), The resurgence of East Asia: 500, 150 and 50 year perspectives. Routledge.
Pomeranz, K. (2005). Women’s work and the economics of respectability. In B. Goodman and W. Larson
(Eds.), Gender in Motion: Divisons of Labor and Cultural Change in Late Imperial and Modern
China, pp. 239–63.
Pomeranz, K. (2009). The great divergence: China, Europe, and the making of the modern world
economy. Princeton University Press.
Qian, N. (2008, August). Missing women and the price of tea in China: The effect of sex-specific
earnings on sex imbalance. The Quarterly Journal of Economics 123 (3), 1251–1285.
42
Ropp, P. S. (1994). Women in late imperial china: a review of recent english-language scholarship [1].
Women’s History Review 3 (3), 347–383.
Ropp, P. S., P. Zamperini, and H. T. Zurndorfer (2001). Passionate women: Female suicide in late
imperial China. Brill.
Shiue, C. H. and W. Keller (2007, September). Markets in China and Europe on the eve of the industrial
revolution. American Economic Review 97 (4), 1189–1216.
Skinner, G.W., H. M. and M. L. Berman.
Socioeconomic macroregions of China.
http://hdl.handle.net/1902.1/21766, Harvard Dataverse, V1 .
Sommer, M. H. (2000). Sex, law, and society in late imperial China. Stanford University Press.
Sossou, M.-A. (2002). Widowhood practices in west africa: the silent victims. International Journal
of Social Welfare 11 (3), 201–209.
(1909). Textile World Record. Number v. 37. Lord & Nagle Company.
Theiss, J. M. (2005). Disgraceful matters: The politics of chastity in eighteenth-century China. Univ
of California Press.
Vella, F. and L. Farré (2007, October). The Intergenerational Transmission Of Gender Role Attitudes
And Its Implications For Female Labor Force Participation. Working Papers. Serie AD 2007-23,
Instituto Valenciano de Investigaciones Económicas, S.A. (Ivie).
Voigtländer, N. and H.-J. Voth (2012). Persecution perpetuated: The medieval origins of Anti-Semitic
Violence in Nazi Germany. The Quarterly Journal of Economics 127 (3), 1339–1392.
Voigtländer, N. and H.-J. Voth (2013). How the west” invented” fertility restriction. American Economic Review 103 (6), 2227–64.
Watson, R. S. and P. B. Ebrey (1991). Marriage and inequality in Chinese society, Volume 12. Univ
of California Press.
Wei, S.-J. and X. Zhang (2011). The competitive saving motive: Evidence from rising sex ratios and
savings rates in China. Journal of Political Economy 119 (3), 511 – 564.
Wong, R. B. c. m. (2002). The search for european differences and domination in the early modern
world: a view from Asia. The American Historical Review 107 (2), 447–469.
Yang, M. M.-h. (1999). Spaces of their own: women’s public sphere in transnational China, Volume 4.
U of Minnesota Press.
Yang, Z., B. Huang, and c. Cheng (2006). Mingdai yizhan kao(Courier stations in the Ming Dynasty).
Shanghai Ancient Books Publishing House.
Zhao, c. (2015). Hidden modernity-social life and romantic experience of Ming and Qing women. Zhi
shi chu ban she.
Zurndorfer, H. T. (1998). Chinese women in the imperial past: new perspectives, Volume 44. Brill.
43
A
Appendix
Table A.1: Robustness Check: Subsamples
Sex ratio at birth
(1)
(2)
(3)
(4)
Yangtze Delta
Net in-migration
Pre-modern textiles
Yangtze Delta
-3.555∗
(1.686)
-11.25∗
(6.252)
-4.497∗∗
(1.703)
-3.968∗∗
(1.726)
-5.238∗
(2.779)
Yes
Yes
Yes
Yes
Yes
Yes
-3.379∗∗∗
(0.926)
Yes
Yes
Yes
Yes
Yes
Yes
1535
0.310
1361
0.311
1535
0.288
888
0.299
Net in migration
Baseline controls
Region FE
Province FE
Observations
Adjusted R2
Standard errors in parentheses
∗
p < 0.10,
∗∗
p < 0.05,
∗∗∗
p < 0.01
Notes: See Table 2. Baseline controls are the same as in Column 4 of Table2. Yangtze
Delta provinces are Jiangsu, Zhejiang & Shanghai. Column 1 directly controls for Yangtze
Delta. Column 2 omits all yangtze Delta provinces. Column 3 directly controls for net
in-migration. Column 4 omits all counties with positive net migration. Robust Standard
errors are clustered at the province level.
44
Table A.2: Robustness: Sex Ratio at Birth after 1990
Sex ratio at birth
(1)
Aged 5 to 9
(2)
Age 1 to 4
2000
(3)
At birth
-2.967∗∗
(1.295)
Yes
Yes
Yes
-5.085∗∗
(1.913)
Yes
Yes
Yes
-3.555∗
(1.686)
Yes
Yes
Yes
-3.132∗
(1.527)
Yes
Yes
Yes
-1.193
(1.610)
Yes
Yes
Yes
1535
0.272
1535
0.364
1535
0.310
1461
0.385
1461
0.245
Pre-modern textiles
Baseline controls
Region FE
Province FE
Observations
Adjusted R2
Standard errors in parentheses
∗
p < 0.10,
∗∗
p < 0.05,
∗∗∗
(4)
(5)
Age 1 to 4 At birth
2010
p < 0.01
Notes:The table reports the impact of pre-modern textiles on sex ratio imbalances among
young cohorts. The unit of observation is a county in the 2000 or 2010 Census. Baseline
controls are the same as in Column 4 of baseresults. Column 4 and 5 only include counties
with no administrative code change between 2000 and 2010. Robust standard errors are
clustered at the province level.
Table A.3: Robustness: Other Gender Outcomes
(1)
Pre-modern textiles
Baseline controls
Control for male outcomes
Region FE
Province FE
(5)
(6)
Women’s Illiteracy
1990
0.167∗∗∗
(0.0488)
Yes
No
Yes
Yes
0.106∗∗
(0.0394)
Yes
Yes
Yes
Yes
0.163∗∗∗
(0.0522)
Yes
No
Yes
Yes
0.130∗∗∗
(0.0304)
Yes
Yes
Yes
Yes
-0.0314∗∗
(0.0116)
Yes
No
Yes
Yes
-0.0113
(0.00674)
Yes
Yes
Yes
Yes
1535
0.592
1535
0.932
1461
0.651
1461
0.946
1126
0.418
1126
0.783
Observations
Adjusted R2
Standard errors in parentheses
(2)
(3)
(4)
Women’s years of education
2000
2010
∗
p < 0.10,
∗∗
p < 0.05,
∗∗∗
p < 0.01
Notes:The table reports the impact of pre-modern textiles on women’s education and
women’s illiteracy. The unit of observation is a county in the 2000, 2010 or 1990 Census. Baseline controls are the same as in Column 4 of baseresults. Column 2 and 4 control
for men’s years of schooling. Column 6 controls for men’s illiteracy rate. Robust standard
errors are clustered at the province level.
45
46
∗∗
p < 0.05,
1503
0.318
Yes
∗∗∗
p < 0.01
1535
0.309
Yes
0.414
(0.825)
-0.0463
(0.104)
-3.562∗
(1.687)
1535
0.311
Yes
-0.766∗
(0.363)
-3.431∗
(1.673)
Sex ratio at birth
(3)
(4)
Pre-1300
State
commerce presence
1535
0.228
-0.821
(0.473)
Yes
-2.919∗
(1.553)
(5)
Industrial
persistence
The table reports the results of testing competing hypotheses. The unit of observation is a county in 2000 Census. The dependent
variable is sex ratio at birth. Baseline controls are those used in Column 4 of baseresults. Pre-1300 commerce is measured by
log (commercial tax quota in 1077+1). #historical courier routes refers to number of courier routes passing a county historically.
Robust standard errors are clustered at the province level.
p < 0.10,
1503
0.317
Observations
Adjusted R2
∗
Yes
Standard errors in parentheses
(2)
Tea
-3.490∗
-3.394∗
(1.682) (1.722)
0.00531
(0.00510)
0.615∗
(0.307)
Baseline controls
#textile companies
#historical courier routes
Pre-1300 commerce2
Pre-1300 commerce
Rice suitability
Tea suitability
Pre-modern textiles
(1)
Rice
Table A.4: Competing Hypotheses
Table A.5: Summary Statistics: 1990 Census
Variable
Pre-modern textiles
Mean
0.446
Std. Dev.
0.444
Min.
0
Max.
1
N
140
Head
Political or Managerial Position
0.560
0.017
0.496
0.13
0
0
1
1
2596827
3415654
0.101
0.085
0.083
0.102
0.111
0.093
0.075
0.079
0.058
0.043
0.04
0.038
0.031
0.025
0.017
0.01
0.007
0.302
0.279
0.276
0.302
0.314
0.291
0.264
0.27
0.233
0.204
0.196
0.191
0.174
0.156
0.13
0.101
0.085
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
5828835
5828835
5828835
5828835
5828835
5828835
5828835
5828835
5828835
5828835
5828835
5828835
5828835
5828835
5828835
5828835
5828835
0.449
0.501
0.004
0.046
0.497
0.5
0.062
0.21
0
0
0
0
1
1
1
1
5828835
5828835
5828835
5828835
0.802
0.006
0.192
0.399
0.077
0.394
0
0
0
1
1
1
4259043
4259043
4259043
0.804
0.982
0.191
4.476
0.873
0.397
0.134
0.491
1.869
0.333
0
0
0
1
0
1
1
5
20
1
5139572
5828835
5828835
5828835
5828835
Age—
0 to 4
5 to 9
10 to 14
15 to 19
20 to 24
25 to 29
30 to 34
35 to 39
40 to 44
45 to 49
50 to 54
55 to 59
60 to 64
65 to 69
70 to 74
75 to 79
80+
Marital status—
Single/never married
Married/in union
Separated/divorced/spouse absent
Widowed
Employment status—
Employed
Unemployed
Inactive
Literate
Han
# own children under age 5
# own family member
At least one married couple
47
Table A.6: Summary Statistics: CGSS
Variable
Pre-Modern Textiles
Mean
0.554
Std. Dev.
0.5
Min.
0
Max.
1
N
74
“Men are naturally more capable”
“Women should focus on family
Son preference
2.982
3.62
0.684
1.265
1.226
0.465
1
1
0
5
5
1
6699
6706
6634
Han ethnicity
Female
Urban site
Communist
Age–
Less than 20
21 to 30
31 to 40
41 to 50
51 to 60
61 to 70
71 to 80
81 to 90
90+
Marital status–
Unmarried
Cohabited
Married
Separated
Divorced
Widowed
Educational attainment–
Less than primary completed
Primary completed
Secondary completed
University completed
Household Registration–
Registered as rural
Registered as urban
Registered as other
0.976
0.518
0.61
0.173
0.154
0.5
0.488
0.379
0
0
0
0
1
1
1
1
6705
6718
6718
6709
0.018
0.116
0.184
0.252
0.197
0.134
0.078
0.02
0.001
0.134
0.321
0.387
0.434
0.398
0.341
0.268
0.14
0.023
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
5498
5498
5498
5498
5498
5498
5498
5498
5498
0.081
0.002
0.815
0.004
0.015
0.083
0.272
0.045
0.389
0.065
0.121
0.277
0
0
0
0
0
0
1
1
1
1
1
1
5498
5498
5498
5498
5498
5498
0.162
0.012
0.539
0.174
0.368
0.109
0.498
0.379
0
0
0
0
1
1
1
1
5496
5496
5496
5496
0.613
0.335
0.052
0.487
0.472
0.222
0
0
0
1
1
1
5498
5498
5498
48