Cotton Revolution and Widow Chastity in Mingand Qing China

RUOBING LIANG, XIAOBING WANG, AND FUTOSHI YAMAUCHI

The historical practice of widow chastity in China is regarded as an outcome of traditional Confucianthought. This study examines how economic factors played key roles in the formation of this culturaltradition by studying the causality between technological progress in the cotton textile sector and theincrease in the number of widow chastity cases between the Ming and Qing Dynasties. We find thatthe incidence of widow chastity increased significantly with rapid technological change in the cottontextile sector and increased demands for female labor. This result remains robust when the issues ofwidow chastity selection and alternative economic and cultural channels are addressed.

Key words: Cotton revolution, economic rationality, technological progress, widow chastity.

JEL codes: D1, N5.

The formation and persistence of culture havebecome increasingly exciting research areas inthe field of economics. Several studies docu-ment the evolution and transmission of genderroles, norms (Alesina, Giuliano, and Nunn2013; Hansen, Jensen, and Skovsgaard 2015;Giuliano and Nunn 2017), and social capital(trust) (Guiso, Sapienza, and Zingales 2008,2016; Durante 2009; Nunn and Wantchekon2011; Lowes et al. 2017). Historical events,such as natural disasters, technologicaladvances, social movements, revolutions, andwars play key roles in cultural evolution. Moststudies focus on cultural traits that appear inmany cultures, such as the preference for boys

over girls, polygamy, religious beliefs, trust,and trustworthiness. The widespread andlong-lasting historical tradition of widow chas-tity has been more popular in China than inother cultures. The practice is stereotypicallyassumed to be a unique outcome of Confu-cianism, but its origin has rarely been dis-cussed in economic studies.

Why is chastity culture popular in China?Does chastity represent a unique feature ofChinese culture or, more specifically, reflecta unique environment of cultural evolution?According to Chang (1999), the invasion andconquest of Mongolia in the thirteenth cen-tury and the technological progress of the cot-ton industry in the late Ming and early QingDynasties represent key events in the prolifer-ation of widow chastity. The Mongolian con-quest laid a foundation for the chastitytradition from an institutional perspective.Specifically, a Han widow could choose onlybetween preserving her chastity and remarry-ing one of her late husband’s brothers tomaintain ownership of the family’s assets. Inother words, the Mongolian invasion intensi-fied aspects of the cultural environment con-ducive to widow chastity. From an economicperspective, the cotton revolution providedwidows with a means of livelihood. Duringthe late Ming and early Qing Dynasties, cot-ton textile work better suited women’s

Ruobing Liang is a professor in the School of Economics, XiamenUniversity. XiaobingWang is an associate professor in China Cen-ter for Agricultural Policy, School of Advanced Agricultural Sci-ences, Peking University. Futoshi Yamauchi is a senior researchfellow in the International Food Policy Research Institute. Wewould like to thank participants in the 2018 Agricultural andApplied Economics Association Meeting in Washington DC, theThird Symposium on Quantitative History at Peking University,the First Young Scholars Forum in Public Economics at XiamenUniversity, AJAE Editor Travis Lybbert, and three anonymousreferees for their constructive comments. This work was supportedby the National Natural Science Foundation of China (GrantNo.71573218, 71673008, 71804058), Major Program of NationalSocial Science Foundation of China (Grant No.18ZDA096), SoftScience Foundation of Fujian Province (Grant No.2019R0003),and the Fundamental Research Funds for the Central Universities,XiamenUniversity (Grant No.20720171023). All remaining errorsare our own.Correspondence to be sent to: xbwang.ccap@pku.edu.cn

Amer. J. Agr. Econ. 103(1): 232–252; doi:10.1002/ajae.12085Published online March 26, 2020

© 2020 Agricultural and Applied Economics Association

physical characteristics than traditional agri-cultural production and could provide apotential means to meet basic livingdemands.1

Since the seminal work by Becker (1973),economic theory has been directly applied tothe analysis of social questions regarding mar-riage, divorce, remarriage, and polygamy(Becker 1991; Becker and Murphy 2000).The rationale underlying family formation isbased on the presumption that the benefits ofmarriage are greater than those of staying sin-gle or cohabiting without formal marriage. Inmarriage, higher incomes among men hastended to support the division of laborbetween men and women, with husbandsworking outside the home and wives workingat home. However, due to improvements inwomen’s educational attainment and earnings,they are more likely to seek employment thanto remain at home as housewives. This trendhas delayed the timing of marriage andincreased the likelihood of remaining single.This hypothesis has been empirically sup-ported by an aggregate analysis performed inthe US (Preston and Richards 1975; Lichter,LeClere, and McLaughlin 1991; Cready, Fos-sett, and Kiecolt 1997), which showed a corre-lation between the lower rate of marriage andthe higher employment rate among women.

Becker conceived of marriage as an invest-ment; thus, women’s marriage decisionsheavily depend on the tradeoff between itscosts and benefits. This conjecture has beenconfirmed by a historical analysis of women’smarriage behaviors in Europe. Hajnal (1965)was the first to discover the late medievalEuropean marriage pattern (EMP): west ofthe line between St. Petersburg and Trieste,women’s marriage rates were low, a significantproportion of women remained celibate, andeven those who married chose to marry lateand had a low fertility rate. The EMP not onlyaccelerated economic development in North-western Europe but also laid a foundation forthe industrial revolution by facilitating humancapital accumulation (De Moor and Van Zan-den 2010; Voigtländer and Voth 2013;Foreman-Peck and Zhou 2014). The institu-tional reason underlying the EMP is that inmost Northwestern European countries,

daughters have inheritance rights that are sim-ilar to those of their brothers regardless oftheir marriage status, whereas in SouthernEurope, women can receive property only ifthey marry. Another important underlyingcause is that after the Black Death, an increasein employment opportunities created suffi-cient economic conditions for women whoremained single (De Moor and Van Zanden2010; Voigtländer and Voth 2013).A similar mechanism also exists in remar-

riage behaviors among widows. Brien,Dickert-Conlin, andWeaver (2004) found thatCanadian widows decided to remarry after theage of sixty because of the punitive provisionon their remarriage before age sixty in theCanadian pension system before 1979. The1979 change that eliminated the marriage pen-alty for those at least age sixty resulted in a sig-nificant increase in remarriage. Baker,Kantarevic, and Hanna (2004) also found thatthese provisions on widows’ remarriage nega-tively affected their remarriage rate. Salisbury(2017) recently analyzed the remarriagebehaviors of the widows of soldiers killed inthe US Civil War with reference to the CivilWar Pension Act enacted in 1862, which statesthat once widows of soldiers remarry, they nolonger qualify to obtain a pension. Followingthe Civil War Pension Act, the probability ofwidows remarrying decreased by 25%, andthe interval between a husband’s death and awidow’s remarriage extended by 3.5 years.In contrast, widows who are more finan-

cially independent are less concerned aboutremarriage. In Medieval Europe, particularlyin England, a woman could request assettransfer from her husband after marriage;after the husband died, the widow’s right tothe assets remained legally secured and evena consuetude. The wife’s assets could accountfor more than one third or even half of thetotal assets, and she could be entitled to theassets under the testament. This institutionrendered widows more financially indepen-dent and being more competitive in the mar-riage market may have caused such widows’remarriage rate to increase (Bennett 1981;Hanawalt 1986, 1993). However, when thewidow remarried, asset ownership was trans-ferred to her new husband; thus, on average,financially well-off widows were less likely toremarry than otherwise (Todd 1985).Other studies have shown that women’s

increased financial capability improves theirsocial status and sex ratio. Qian (2008) foundthat the sex ratio of men to women decreased

1Beginning in the Song Dynasty, foot binding had become verypopular among Han women, and this custom reached an apex inthe Ming and Qing Dynasties. It also bound women to handcraftproduction by depriving them of their ability to engage in foodproduction.

Liang, Wang and Yamauchi Cotton Revolution 233

and the educational attainment of women sig-nificantly increased in tea production areasduring the economic reform that began in1979. Xue (2014) found that in the areas wheretextile industries developed in China duringthe thirteenth century, women had a highersocial status and there was a more balancedsex ratio than in areas where the textile indus-tries were underdeveloped. Alesina, Giuliano,andNunn (2013) found that the traditional useof the plow in agricultural production influ-enced the evolution of gender norms, and inplow-using areas, there is still less female par-ticipation in the workplace, politics, and entre-preneurial activities. Thus, the development ofindustries that demand more female laborersimproves the social status of women, andwomen become less likely to depend on theirhusbands. Particularly in conservative socie-ties, such as those of the Ming and QingDynasties, when Confucianism was dominantin China, whether widows remained unmar-ried (widow chastity) was influenced by theirfinancial capability.Using two-period panel data of prefectures

in the Ming and Qing Dynasties, this paperaims to identify the effect of the technologicalprogress in cotton textile production on thenumber of chaste widows in a society domi-nated by Confucianism. First, we discuss theinstitutional background of the origin ofwidow chastity, including the inheritance ofassets since the Yuan Dynasty and the poten-tial social recognition and gratification onecould obtain through widowhood in the Mingand Qing Dynasties. Second, we theoreticallyand empirically establish that the increase inthe number of widow chastity cases from theMing to Qing Dynasties is explained by the“cotton revolution”, which improved thefinancial status of women. Our results showthat the incidence of widow chastity increasedsignificantly with rapid technological changein the cotton textile sector and increaseddemand for female labor. Third, we introducea series of tests to support the robustness ofour baseline results. With proxy variablesrepresenting Confucianism and norms, thetests show that strict adherence to Confucian-ism promotes the growth of widow chastityand the impact of technological progress inthe cotton textile sector remains robust. Inthe falsification tests, it is explicit that Confu-cianism promotes the growth of widow chas-tity from the Ming to Qing Dynasties andprevents widows’ irrational behavior—mar-tyrdom. We also use the economic index,

other geographical variables, conflicts andwar, and the production of other commodities,such as silk, linen and tea, to address the selec-tion issue in this study. When controlling forthe above factors, the coefficients of interestremain stable. Finally, the mechanism analysisalso provides evidence that technological pro-gress in cotton textile production explains thegrowth in widow chastity from the Ming toQing Dynasties better than traditional cottonplanting areas do. Based on additional dataon labor participation and income in industrialsectors from the 1920s and 1930s, the resultsconfirm that technological progress in cottontextile production increased female labor par-ticipation as well as women’s wages, suggest-ing that the cotton revolution improved thefinancial capability of widows, therebyexplaining the increase in the incidence ofwidow chastity.

The paper is organized into seven sections.Section 2 summarizes the historical backgroundof the institutional shift and technological pro-gress of the cotton textile industry. Section 3presents the datasets and theoretical and empir-ical models. Section 4 presents the baselineresults. Section 5 presents the results of therobustness tests, taking into account the culturalinfluences of Confucianism, issues of selectionbias, and the influence of alternative crops andperforming other sensitivity tests. Section 6 fur-ther discusses the possible mechanism of thecotton revolution’s influence on widowhood.The final section presents the conclusions.

Historical Background

Institutional Background of Widow Chastity

The historical practice of widow chastity inChina is considered a product of Confucianethics. “The three guidelines and five rules”and “the three obediences and four virtues”were followed extensively in China, whichled to the institution of honoring widow chas-tity.2,3 Despite the existence of this institution,

2The three guidelines posit that the monarch guides the minis-ters, the father guides the children, and the husband guides thewife. The five rules are benevolence, righteousness, propriety, wis-dom, and trust.

3The three obediences advocate that before marriage, a girlmust follow her father’s guidance; after marriage, the wife mustfollow her husband’s guidance; and if the husband dies, the widowmust follow her son’s guidance. The four virtues are virtue inbehavior, words, comportment, and honor.

234 January 2021 Amer. J. Agr. Econ.

lifetime widowhood was rare, and remarriagewas common before the Ming Dynasty.According to the New Book of Tang and theTang Huiyao, more than one-fifth of TangDynasty princesses remarried after their hus-bands passed away.4 Thus, most Chinese indi-viduals did not conform to the norm undertraditional Confucianism. However, in theSong and Yuan Dynasties, the number ofwidows who did not remarry increased.

It is traditionally thought that the popularityof widow chastity was highly influenced byNeo-Confucianism during the Song Dynasty.Confucian representatives, including YiCheng andXi Zhu, advocated for the “mainte-nance of justice at the expense of humandesire” and the idea that “starvation is trivialwhile disloyalty is serious,” considering remar-riage a severe act of betrayal. These theoriesoverlook the economic factors that played sig-nificant roles in the formation and evolution ofthis norm. Holmgren (1985) and Chang (1999)explored the levirate tradition in Mongolianethnicities and drew evidence from the YuanDynasty.5 During the Yuan Dynasty, due tothe Mongolian takeover and the dominationof the nomadic people in China, the Yuan gov-ernment stipulated that after marriage, awoman’s rights and assets would transfer fromher parents to her husband’s family and thateven after the husband’s death, the husband’sfamily would retain control over the wife’sassets. Mongolia implemented the levirate sys-tem primarily for economic reasons. Becausenomadic families are highly mobile, the hus-band’s family would suffer severe economiclosses if a widow remarried a man outside thetribe and took her late husband’s assets withher. Hence, if a Mongolian widow remarriedor returned to her parents, she would loseownership of her assets, including the guard-ianship of her children. However, the Hanpopulation intensively resisted this leviratesystem, forcing the government to compro-mise; thus, a Han widow was not required toremarry within her husband’s family or to

marry anyone else (Holmgren 1985, 1986;Birge 1995; Chang 1999).In the Ming and Qing Dynasties, to main-

tain social order, rulers instituted a portfolioof social regulations inspired by Neo-Confucianism and Mongolian customs. Forexample, Emperor Yuanzhang Zhu enacteda government policy honoring women whowere widowed before the age of 30 years anddid not remarry for more than 20 years.6 Healso followed the Yuan inheritance law thatstated that a remarried widowwaived the rightto inherit her husband’s assets. In 1723 (theearly Qing Dynasty), the government reducedthe chastity requirement to 15 years; in 1824,the duration was reduced to one decade, andin 1871, it was reduced to 6 years (Shen 1936;Tien 1988; Chang 1999). The positive attitudetoward widow chastity in mainstream ideologyand propaganda might have led to theobserved increase in the number of widowswho did not remarry in the Ming and QingDynasties. Though the government encour-aged widow chastity, widows were still not for-bidden from remarrying. More importantly,widow chastity was not rewarded in kind orwith cash. Although a chaste widow’s familywas exempt from labor services, such a privi-lege would not materialize until at least10 years after a widow lost her husband, thatis, the time when the government could honorher widow chastity.7

Consequently, widow chastity resulted insocial esteem but not material or economicsupport. Furthermore, a widow continued toface financial burdens and was likely to betrapped in poverty because of obligations tocare for her dependents, including her eldersand her children, during her widowhood.Obviously, the government’s and society’sencouragement of long-term widowhood andthe government-issued honors did not solvethe financial constraints of widows; thus, thesemechanisms are not convincing explanationsof the tremendous growth in the number ofwidows who did not remarry. Based on Li’s

4During the Tang Dynasty, one hundred and thirty-four prin-cesses married, twenty five of whom remarried. Three of thesemarried three times and one married four times.

5The husband’s family took charge of the widow’s remarriage.This act could be categorized as remarriage to the same or to thenext generation. For example, remarriage to the same generationmeans that after a husband has died, his wife (the sister-in-law) isremarried to his elder or younger brother. Remarriage to the nextgeneration means that after a husband has died, his wife is remar-ried to a son of his brother; if the uncle died, his wife is remarried toher nephew. A summary of this remarriage system can be found inGuo (2000).

6In principle, widows who met those requirements were eligibleto be honored with widow chastity by the central government. Inpractice, however, many qualified women were never honoreddue to information asymmetry and the complex application pro-cess. Please see Section 5.2 for detailed information on the applica-tion process.

7In the Qing Dynasty, the prize honoring widow chastity was asfollows: the local government paid thirty liang (tael) silver coins tobuild a memorial gateway for widow chastity; the family wasexempt from servitude and offered grain crops in kind in the eventof a natural disaster. However, these incentives hardly improved awidow’s well-being under normal circumstances.

Liang, Wang and Yamauchi Cotton Revolution 235

(1998) findings from the lower reaches of theYangtze River, Chang (1999) hypothesizedthat the introduction of cotton and textile pro-duction affected widows’ welfare: the cottontextile sector’s rapid development beginningin the middle of the Ming Dynasty grantedwidows greater economic capability, enablingthem to live without resorting to remarriage.

Cotton Revolution and Women’s FinancialCapabilities

In general, widows in the Ming and QingDynasties secured their livelihoods by inherit-ing their husbands’ assets and by earning theirown income. Inheritance appeared onlyamongwealthy families, and widows were ableto take over the assets only on behalf of theirchildren, that is, they did not have the rightto directly inherit. Other relatives in the familycould force the widow to remarry to keep theassets within the husband’s family. If the hus-band’s family was not rich enough, the widowcould not depend on the husband’s assets.Most widows had to work in spinning, agricul-ture, housekeeping, and small businesses. Thefirst panel in Figure A1 shows the change inwidow chastity patterns from theMing toQingDynasties by prefecture. An increase in widowchastity is observed from the Ming to QingDynasties. However, the rapid growth inwidow chastity during the Qing Dynasty can-not be explained solely by Confucianism’sinfluence. Following the historical study byChang (1999), we empirically test the hypoth-esis that widows’ improved financial condi-tions contributed to an increase in thenumber of chaste widows.The cotton revolution during the mid-Ming

Dynasty refers to the rapid introduction of cot-ton production and the development of thetextile handcraft industry (Chang 1999). Bythe early Qing Dynasty, the textile sector hadevolved from sideline production within arural family (cottage) to commercialized oreven mainline production due to the signifi-cant technological progress at the end of theMing Dynasty and the early Qing Dynasty(Li 1998). The second panel in Figure A1shows that approximately one-fifth of China,clustered along the Yangtze River and in theHuang-Huai-Hai (Yellow River, HuaiheRiver, and Haihe River) plain, is suitable forcotton production. According to the Compila-tion on Agriculture published during theMingDynasty (Xu 1930), spinning machines with

four or five axes appeared in Southeast Chinaat the end of the Ming Dynasty. By the QingDynasty, spinning machines that could pro-duce 0.5 kg of cotton textile per day werewidely used (Li 1998).

Even in the late Song and early YuanDynasties, the spinning sector had alreadyexperienced significant technological pro-gress. Daopo Huang in Songjiang Prefectureapplied “defend, flick, spin and weave” to roll-ing carts, arches, spinning machines, andweaving machines as well as techniques suchas mixing different colored threads to produceknitted and patterned textiles.8 The weavingmachine was invented by Daopo Huang400 years before the cotton gin was inventedby Whitney in the US. This invention trans-formed Songjiang Prefecture into the centerof the textile industry in the Yuan, Ming, andQingDynasties. Subsequently, the sea journeyban (1371–567) in theMing Dynasty was liftedto promote the development of the textile sec-tor. Following porcelain and silk, cotton textilewas the third most important productexported to Europe, Japan, and SoutheastAsia. Moreover, as proof of cotton produc-tion’s commercialization, 15 to 20 million piof cotton cloth (1 pi = 13 m) was consumedannually by the government alone during theWanli era (1573–620) of the Ming Dynasty(Yan 1955).

Due to technological progress in the cottontextile sector, the share of textiles in the totalvalue of exchanged products reached 27%,while that of grain crops and silk was approxi-mately 40% and 7%, respectively. More thanhalf of cotton textile products entered themar-ket (Wu andXu 1985). Cotton textile commer-cialization also improved women’s financialstatus. Pomeranz (2005) estimated that for afarmer who cultivated rice and wheat on0.67 ha (10 mu) of land, the agricultural pro-duction profit was equivalent to 14 shi(1 shi = 90 kg) of rice. One rural laborer whoworked 200 days per year would earn less than5 shi of rice. However, one female textileworker could earn more than 7 shi of rice.During that time, many women worked inthe cotton textile sector; thus, the role ofwomen in their families changed fromworkingat home to working outside the home andfinancially supporting their families. Thischange also allowed widows not to remarry.

8Songjiang Prefecture in the Ming and Qing Dynasties waslocated near today’s Shanghai municipality.

236 January 2021 Amer. J. Agr. Econ.

Figure 1 presents the uptrend in the incidenceof widow chastity between the Ming and QingDynasties; on average, the number of widowchastity cases in traditional cotton-plantingareas increased faster than that in noncottonareas. Figure 2 shows the negative relationshipbetween the increase in the number of widowchastity cases and the distance to SongjiangPrefecture, which was the cotton textile tech-nological center in the Ming and QingDynasties.

Before the end of the Ming Dynasty, menworked more efficiently than women becausemen could handle more labor-intensive workand were more skilled than women, for exam-ple, traditionally, only men inherited weavingtechniques (Li 1998). However, due to thetechnological progress at the end of the Ming

Dynasty, women could work more efficiently,and men’s comparative advantage diminished.Men began to play an important role in agri-cultural production and to work as secondarylaborers in the cotton textile sector in the agri-cultural production off season. Although thelinen and silk textile sectors were importantduring the Ming and Qing Dynasties, thesetwo sectors cannot explain the increase in theincidence of widow chastity. Before the Yuanand Ming Dynasties, silk and linen wereimportant materials for cloth, but the price ofsilk was high, and silk production was highlyspecialized. Silk production moved from ruralfamilies to workshops, and widows could noteasily engage in silk production. In linen pro-duction, the added value was low, and theincome from linen production was not enoughto secure the livelihood of a family (Li 1998;Chang 1999).

Theoretical Model

In this subsection, we lay out a simplemodel toclarify our key hypotheses. Widows decidewhether they remarry or stay widowed in eachperiod after seeing a randomly drawn value ofa potential spouse. Here, we do not ask howthey became widows – voluntarily or involun-tarily. VR(v) is the value of remarrying to aspouse of value v. Similarly, VW is the valueof staying widowed. Subscripts R andW stand for the states of remarrying and stay-ing widowed, respectively. Thus, the widowwill compare VR(v) and VW to make thedecision.

ð1Þ V vð Þ=maxR,W VR vð Þ,VWf g

where spouse value v~F(v). Assume that ifremarrying, she will have the value of herspouse v and labor income (1 − α)y(z). Thevalue of α represents the time allocated to fam-ily duties and the responsibilities imposed bythe new family. For example, if they areengaged in farming, she is expected to assistwith fieldwork. In our context, y(z) is wagesfrom textile work at home or a workshop,which is a function of technology z. While y(z) is driven by technological change, α is cul-turally determined. In the current setting,VR(v) is determined as

ð2Þ VR vð Þ= 1−αð Þy zð Þ+ v+ βVR vð Þ

01

00

200

300

400

Me

an

of W

ido

w C

hastity

Song Yuan Ming Qing

Traditional Cotton Area Non-Cotton Area

Figure 1. Traditional cotton areas and thenumber of widow chastity cases [Color figurecan be viewed at wileyonlinelibrary.com]

02

46

8

Log

Incre

ase

in W

ido

w C

hastity

4 5 6 7 8

Log Distance to Songjiang Prefecture

Figure 2. Relationship between distance toSongjiang prefecture and change in thenumber of widow chastity cases from theMingto Qing dynasty [Color figure can be viewed atwileyonlinelibrary.com]

Liang, Wang and Yamauchi Cotton Revolution 237

where β is the discount factor. The value ofstaying widowed VW is determined as

ð3Þ VW = y zð Þ−c+ βEV v0ð Þ

where c is the cost associated with widowhood,and EV(v0) is the expected value in the nextperiod. Note that c can be negative if thewidow is socially rewarded. Solving the Bell-man equation in this search model,

ð4Þ V vð Þ=maxR,W

�1

1−β1−αð Þy zð Þ½

+ v�;y zð Þ−c+ βEV v0ð Þ�

we obtain

ð5Þ v* =m βEV v0ð Þ−c½ �+ α−βð Þy zð Þ

wherem = 1 − β. Here, the widowwill remarryif v > v* (i.e., she stays widowed if v < v*).Assume that α > β. The probability of remain-ing widowed is F(v*). We have three observa-tions from the above model. It is more likelyfor the widow to stay widowed if (a) textilewages increase due to an increase in femalelabor demands driven by technological change,(b) the traditional duties cast upon marriedwomen are heavier, and (c) the cost associatedwith widowhood is smaller (or widowhood issocially rewarded). These changes in (a) to (c)will increase v*; therefore, the widow is morelikely to stay widowed in the above scenario.

Econometric Models and Data Sources

Econometric Models

This paper aims to identify the economicmechanism explaining the variation in thenumber of widow chastity cases in a societydominated by Confucianism. Our main theoryis that a dramatic improvement in women’swork efficiency, driven by the technologicalprogress in the cotton textile sector, providedbetter income opportunities for women tosecure a livelihood; this change enabledwidowchastity. We hypothesize that the large differ-ence in widows’ choice to pursue chastitybetween the Ming and Qing Dynasties wasmainly the result of technological progress inthe textile sector and the specialization of

agricultural and textile production. This paperapplies a fixed effects model to disentangle thefactors that may explain the difference in theincidence of honored widow chastity betweenthe Ming and Qing Dynasties.

The estimation model is described asfollows:

ð6Þ lnwidowic = α+ β0tc + β1tc × lndisti

+ γtc ×Zi + εi + μic

where the dependent variable is lnwidowic,which is defined as the log form of the numberof honored widow chastity cases (per 10,000)during period c in i prefecture; tc is a dummyvariable representing the dynasty, where 0 rep-resents the Ming Dynasty and 1 represents theQing Dynasty; and lndistiis the log form of thedistance of i prefecture to Songjiang Prefec-ture, the center of textile production from theSong to Qing Dynasties and the place wherethe famous Daopo Huang introduced andinvented several new textile technologies atthe end of the SongDynasty and the beginningof Yuan Dynasty. Zi is a vector of the time-invariant variables, including the geographicallocation, topographical, and climate variablesand population density, and tc × Zi controlsfor the possible influence of confounding fac-tors; εi and μic are the prefecture fixed effectsand error terms, respectively. Here, geograph-ical variables include the dummy variable ofprovincial capital, land area, elevation, slope,longitude, and latitude; weather data includeannual mean temperature and precipitation.The interaction terms between the QingDynasty dummy and these variables areincluded as additional controls.

One of our hypotheses regards whethertechnological progress in the cotton textilesector had a determinant impact on the growthin widow chastity between the Ming and QingDynasties. We also test the effects of the cul-tural determinism of traditional norms(Confucianism and popular customs) onwidow chastity. To address the selection issue,we test the influence of those variables,namely, the distance to Beijing and provincialcapitals, economic development, the introduc-tion of new food products (maize), naturalcalamities, and domestic conflict, that are asso-ciated with the number of widow chastitycases. Furthermore, other traditional textilesectors, such as silk and linen textiles, and teaproduction may influence widow chastity,and we also test their influence. Finally,

238 January 2021 Amer. J. Agr. Econ.

quasi-experimental analysis requires a com-mon trend test to confirm that the number ofwidow chastity cases increased (more) rapidlyin the Qing Dynasty. For this, we use datafrom before the Qing Dynasty.

In the mechanism analysis, we adopt twostrategies to verify our logic. First, we use atraditional cotton planting variable to replacethe distance variable and then put the two var-iables into the model together. Second, we usethe interaction term of distance and cottonplanting to capture the interactive effects ofcotton production and technological progress.

The specifications are described as follows:

ð7Þ lnwidowic = α+ β0tc + β1tc × lndisti

+ β2tc × cottoni + γtc ×Zi + εi + μic

ð8Þ lnwidowic = α+ β0tc + β1tc × lndisti

+ β2tc × cottoni + β3tc × lndisti

× cottoni + γtc ×Zi + εi + μic

Specifically, cottoni represents whether cot-ton production occurred in i prefecture tradi-tionally (1 = yes; 0 = no). The interaction termtc × cottoni captures the effect of cotton pro-duction from the Ming to Qing Dynasties onwidow chastity; the triple interaction termtc × lndisti × cottoni checks whether the devel-opment of textile production is necessary forcotton production to impact widow chastity(and vice versa). Although cotton and textileproduction are vertically linked, the variabletc × lndisti also includes types of textiles otherthan cotton and therefore represents the wholeeffect of textile development onwidow chastity(female labor demand). However, lack of cot-ton production would also mean the lack of acritically important input to textile production.

Data Sources

The number of honored widow chastity casesand the change in this number between theMingand Qing Dynasties are the dependent variablesof interest. More than 84,000 records of famouswomen are manually sourced from the “Compi-lation of the Qing Dynasty Revised in Jiaqing’sReign” by Guwargiya et al. (2008).9 Further-more, we use the number of honored widows

who committed suicide in the falsification tests.Distance to the technological center, SongjiangPrefecture, is obtained from the ChinaHistoricalGeographic Information System (CHGIS),which was jointly developed by the FairbankCenter for Chinese Studies, Harvard University,and the Historical Geography Research Centerof Fudan University.10 The cotton productionindicator is constructed based on whether cottonwas produced in the prefecture (or province) ineither the Ming or Qing Dynasty (Guwargiyaet al. 2008).Population data from 1600 and 1820 are

obtained from the History Database of theGlobal Environment (HDGE) at the prefec-tural level to normalize the key dependentand independent variables and the populationdensity.11 We also use geographical dataobtained from CHGIS, including the prefec-tures’ average areas, elevations, slopes, longi-tudes, and latitudes. Weather data, includingannual average temperature and precipitation,are collected from the National Earth SystemScience Data Center (NESSDC).2 To assessthe effects of the selection bias potentially cre-ated by the widow chastity procedure, we alsoinclude distance to provincial capitals, distanceto Beijing, regional economic developmentproxied by agricultural tax (Liang 1980), themaize planting suitability index, and the fre-quency of conflicts and of natural calamities(see Figure A3, Chen et al. 1986).13 Finally,in robustness tests, we select silk, linen, andtea as alternative commodities to test the pos-sibility that the cotton textile technologicalprogress variable captures the effects of othercommercial crops closely related to women.To test theConfucianism effect, we use the fol-

lowing four proxy variables to represent the influ-ence of Confucianism and customs (culturaleffect). We use the quota for the imperial exami-nation assigned at the prefectural level (Bai andJia 2016); this quota consists of two parts: eachcounty in the prefecture received a quota, andthe prefecture as a whole received an additional

9This compilation was written in three series for the Kangxi,Qianlong, and Jiaqing reigns. The final series, which was revisedunder Jiaqing’s reign, includes 560 volumes (Figure A2). The com-pilation began in 1812 and was completed during the 22nd year of

Daoguang’s reign in 1842. All information is updated to theyear 1820.

10The data are available at http://www.fas.harvard.edu/~chgis/,and we use the version published in April 2007.

11TheHistoryDatabase of theGlobal Environment is accessibleat: https://themasites.pbl.nl/tridion/en/themasites/hyde/index.html.

2TheNational Earth System ScienceData Center is accessible athttp://www.geodata.cn/.

13The suitability index of maize planting is calculated based onregional temperature, moisture, soil, terrain, and water deficit con-ditions. Detailed information is available at http://gaez.fao.org/Main.html#.

Liang, Wang and Yamauchi Cotton Revolution 239

quota that could be shared among counties.14 Thedataonquotas come fromGuwargiya et al. (2008)and were determined based on the ImperiallyEstablished Institutes and Laws of the GreatQing Dynasty (Aisingoro et al. 1899). Schoolsin the Qing Dynasty mainly taught candidatesfor the imperial examination, and the curriculaand syllabus in these schools mainly focused ontraditional classics of Confucianism (Yuchtman2017). Thus, the number of schools at the prefec-tural level reported in the study by Guwargiyaet al. (2008) is used to represent the impact ofConfucianism. Ancestral shrines are well docu-mented as representing the cultural/customseffect (Guwargiya et al. 2008). The number ofancestral shrines in the Qing Dynasty is alsoobtained from the above study.15

In the mechanism analysis, we employ twodata sources to investigate the labor participa-tion and income of female workers. Specifi-cally, the employment data come from theindustrial enterprise survey conducted by theResource Committee of the National MilitaryCouncil in 1936 (Liu 1937), and the wage dataare obtained from the worker living surveyconducted by the Ministry of Industry andCommerce (MIC) of the National Govern-ment from 1926–30 (MIC 1930). With thesetwo data sources, we can analyze the effect ofthe distance to Songjiang Prefecture on femaleemployment and wages. The descriptive statis-tics of the variables are presented in Table A1.

Empirical Results

This section presents the empirical results. Themain results are shown inColumns 1–6 of table 1.The results show that the coefficients of the logdistance–Qing Dynasty dummy interaction termare negative and statistically significant, and the

coefficient magnitudes are stable across the spec-ifications.Without the distance variable, the coef-ficients of the dynasty dummy variable (t) arepositive and significant, suggesting that the num-ber of widow chastity cases increased betweenthe Ming and Qing Dynasties. This time trendcaptures thepositive effects of the supportingpol-icy aswell as cultural factors onwidowchastity. Inaddition, the number of widow chastity cases per10,000 people decreased between the Ming andQing Dynasties by 37%–44.2% as the distanceto Songjiang Prefecture increased by 1% point,and the results are statistically significant. There-fore, the growth in widow chastity was high inprefectures where cotton textile technology wascomparatively advanced in the Qing Dynasty.

Columns 4–6 add variables such as geo-graphical variables, interacted with the QingDynasty dummy, and population density. Theinteraction terms between the provincial capi-tal dummy and the Qing Dynasty dummy arepositive and statistically significant, which indi-cates that proximity to local authority led to anincrease in the incidence of widow chastityfrom theMing to Qing Dynasties. We use pop-ulation density at the prefecture level to testthe bandwagon theory (Chang 1999). The sig-nificant and positive effects of population den-sity indicate that chastity behavior in the Mingand Qing Dynasties was related to populationdensity, which is consistent with our conjecturethat women might be forced to act under pres-sure from their peers and the public. The mag-nitude of the parameter estimate of theinteraction term between the distance variableand the Qing Dynasty dummy in Column4 slightly decreased when controlling for popu-lation density. Interestingly, in Columns 5 and6, the Qing Dynasty effect becomes insignifi-cant, while population density remains signifi-cantly positive. This finding can be explainedby the fact that population density is relatedto income level (Chen and Kung 2016). Never-theless, the interaction term of the distance andthe Qing Dynasty dummy remains significantand negative in Columns 4 to 6, which appearsto support our key proposition on technologi-cal change and women’s economic status.

Robustness Tests

Alternative Explanations of Cultural Influence

According to traditional explanations basedon Chinese culture, Confucianism, especially

14Further details about the imperial examination can be foundin Lin (1995), Bai and Jia (2016), and Yuchtman (2017).

15All of the four proxy factors exists in the pre-period (Mingdynasty or before); however, the cultural effects differ significantlybetween Ming and Qing Dynasties. For example, Keju imperialexamination started from Sui Dynasty and formalized in TangDynasty. In each of dynasties, there are some reforms on the impe-rial examinations to select the elites. At the beginning of Mingdynasty, the numbers of elites selected from Southern Chinaexceed those from Northern China. To balance the numbers ofthe elites selected between Southern and Northern China, theimperial examines were designed into two exams specified forthe elites in the two regions. In Qing dynasties, the imperial exam-ination was unified into one national exam. The quotas have beenincreased for the Northern China and the remoted province likeYunnan. Similarly, we also expects the expansion of schools inQing Dynasty.

240 January 2021 Amer. J. Agr. Econ.

Tab

le1.

BaselineResults lnwidow

(1)

lnwidow

(2)

lnwidow

(3)

lnwidow

(4)

lnwidow

(5)

lnwidow

(6)

t×lndist

−0.370***

(0.046)

−0.442***

(0.045)

−0.442***

(0.046)

−0.412***

(0.045)

−0.363***

(0.078)

−0.390***

(0.082)

t3.319***

(0.322)

3.807***

(0.309)

3.810***

(0.316)

3.095***

(0.454)

1.074(1.653)

4.043(3.304)

lnpo

pden

0.352***

(0.125)

0.416***

(0.136)

0.444***

(0.142)

t×capi

0.227***

(0.086)

0.216***

(0.082)

0.218***

(0.082)

t×lnarea

0.007(0.030)

0.013(0.028)

0.011(0.028)

t×slop

e0.032(0.024)

0.050*

(0.029)

t×alti

−0.000(0.000)

−0.000(0.000)

t×lati

0.019***

(0.006)

−0.005(0.029)

t×long

i0.008(0.011)

0.010(0.011)

t×temp

−0.001(0.003)

t×lnpreci

−0.235(0.169)

constant

0.101***

(0.012)

0.101***

(0.013)

0.173***

(0.057)

−0.988**(0.385)

−1.185***

(0.419)

−1.273***

(0.438)

Prefectureeffects

NY

NY

YY

Provinceeffects

NN

YN

NN

Observation

s522

522

522

522

522

522

Prefectures

261

261

261

261

WithinR

20.616

0.812

0.691

0.826

0.835

0.837

Notes:C

oefficien

tsarerepo

rted

withstan

dard

errors,clustered

attheprefecture

leve

l,in

parenthe

ses.

*p<0.1.

**p<0.05

.**

*p<0.01

.

Liang, Wang and Yamauchi Cotton Revolution 241

the Neo-Confucianism developed by YiCheng and Xi Zhu in the Song Dynasty,played a key role in forming the chastity cul-ture. We use additional culture-related vari-ables to test this theory. The traditionalcultural variables, including the imperialexamination quota (keju in Chinese), the num-ber of schools teaching the traditional Chineseclassics (Shuyuan in Chinese), and the numberof ancestral shrines (Zongci and Zumiao inChinese), are normalized by population sizeand converted to log form. These variablesare used to reflect the possible cultural impactsof different social factors: the direct influence,indirect influence, and folk-custom influenceof Confucianism. The log form of the numberof temples (Si and Guan in Chinese) per10,000 people is used to capture the influenceof Buddhism and Taoism. We estimate theinteraction effects of each of these variableswith the Qing Dynasty dummy to determinethe magnitude and direction of these socialeffects.In column 1 of table 2, the average imperial

examination quota, the number of schools, andthe number of temples, but not the number oftraditional shrines, are statistically significant.The coefficients of the quota variable and thenumber of temples are positive and significant,suggesting that official Confucian education,Buddhism, and Taoism had positive influenceson widow chastity. In contrast, the negativecoefficient of the number of schools indicatesthat unofficial Confucian education had a neg-ative and statistically significant effect on theincidence of widow chastity. Surprisingly, thiscontradicts the traditional explanation. Col-umn 1 of table 2 also shows that textile technol-ogy played a significant role in widows’

choices, notwithstanding the cultural factors,which also exerted influences. To test thehypothesis that women from wealthy familieswere more likely to comply with Confucian-ism’s common rules in column 2, we conductfalsification tests to analyze the effect of textiletechnological progress on the number ofwidows who committed suicide.16 The techno-logical effect on martyrdom behavior (thenumber of widow suicides) is not statisticallysignificant (Column 2), and its effect on thetotal number of both widow chastity cases andwidows who committed suicide remains nega-tive and statistically significant (Column 3).These results are consistentwith the hypothesisthat the technological development of the cot-ton textile sector affected widows’ choice topursue chastity instead of committing suicide.However, they failed to support the possibilitythat the development of the cotton textile sec-tor decreased the number of widow suicidesby improving their economic prospects.

Because the influence of traditional culturecannot be ignored in widow chastity behavior,could it also impact widow suicide? To answerthis question, we examine the effect of culturalfactors on martyrdom behavior. Column 2 oftable 2 shows the effects of various culturalvariables on martyrdom behavior. We find asignificant negative effect of the quota, indicat-ing that official Confucianism did not promotemartyrdom behavior. The results partiallyreflect the official ideological differences

Table 2. Alternative Explanation of Cultural Influences

lnwidow (1) lnsuicide (2) lnboth (3)

t × lndist −0.480*** (0.076) 0.023 (0.036) −0.409*** (0.070)t × lnquota 0.496*** (0.144) −0.219*** (0.084) 0.269** (0.130)t × lnschool −1.045** (0.477) 0.581* (0.327) −0.358 (0.455)t × lnshrine 0.510 (0.379) −0.345 (0.270) −0.010 (0.379)t × lntemple 0.570* (0.307) 0.020 (0.160) 0.438 (0.267)Prefecture effects Y Y YObservations 522 522 522Prefectures 261 261 261Within R2 0.856 0.288 0.832

Notes: Coefficients are reported with standard errors, clustered at prefecture level, in parentheses; control variables include t, lnpopden, t*capi, t*lnarea, t*slope,t*alti, t*lalti t*longi, t*temp, and t*lnpreci.*p < 0.1.**p < 0.05.***p < 0.01.

16Women from noble families or the courtier class were moreconformist. Guo (1987) explored fifty genealogies and found onlyone case in courtier families in which the widow remarried. Ju(2001) also examined forty-six genealogies and found that among5277 women, more than a quarter (1263 women) were widows.

242 January 2021 Amer. J. Agr. Econ.

regarding chastity and martyrdom recorded inthe Emperor’s Commission. Column 3 reportsthe estimation results for the number of widowchastity and suicide cases.Except for the quota,all other cultural variables are insignificant.

Overall, our findings are consistent with theopinion expressed by the emperors during theQing Dynasty. For example, both Kangxi(1654–722) and Yongzheng (1678–735) hadreservations regarding widow suicide but hon-ored widow chastity on several official occa-sions. The Yongzheng Emperor oftencommented on the two behaviors, that is,widow chastity andwidow suicide, and encour-aged the former. For example, he said that“the widow commit[ing] suicide…she shouldbe responsible for the family after the deathof the husband…and she should not absolveher responsibility by committing suicide”(Records Commission 1985). He did not honorwidow suicide, which to him was irresponsible.He also maintained a conservative opinionregarding widow chastity for those widows inthe Eight Banner clans (Manchurian families).For example, he said, “If Banner widows areyoung, childless and without clansmen, it isnot beneficial to them tomaintain widow chas-tity” (Aisingoro et al. 1976). The reason for hisstance was that once a widow from any of theEight Banner clans chose to maintain widowchastity, her expenses would be covered bythe government. To reduce the costs incurredby widow chastity, the government decidedwhether to allow the widow to maintain herchastity. In contrast, a Han widow, whoseexpenses were not covered by the govern-ment, was free to maintain her chastity.

Widow Chastity Selection

The Qing government was in fact likely toapprove widow chastity, although not allchaste widows could be honored, as widowchastity selection was subject to the followingprocedure. The nomination for widow chastitywas made by the husband’s clan or the localgentry; then, the central government honoredwidows upon the submission of the applica-tions by the local authorities. Guo (2000)noted that widow chastity was mainly recog-nized in families in either wealthy areas orareas with accessible geographical locations.Widows who resided in remote areas or origi-nated from poor families in poor areas wereless likely to be honored for widow chastity.The variable representing technological

progress in the cotton textile sector could cap-ture the selection effects of widow chastity;therefore, this variable may lead to bias.We use the following two groups of variables

to explore the variation in widow chastity selec-tion: an economic index and a vector of distancevariables. Here, agricultural productivity prox-ied by the agricultural tax revenue per capitais used as an economic index, and two distancevariables represent the distance from the pre-fecture seat to the capital of China (Beijing)and the distance between the prefecture seatto the province’s capital city. The interactionterms between the Qing Dynasty dummy andthese variables are included as additional con-trols for the robustness check. Column 1 intable 3 presents the results controlling for theinteraction terms between the Qing Dynastydummy and the additional control variables ofthe economic index and two distance variables.The coefficients of the interaction termsbetween the agricultural tax revenue per capitaand the Qing Dynasty dummy are positive andstatistically significant. Thus, the results overalllead us to the following two conclusions: first,rural income had a positive impact on the num-ber of widow chastity cases; second, widowsresiding in wealthy regions, likely located ineastern China, had better financial capacity.After controlling for the above factors, the coef-ficient slightly increases for the interaction termbetween technological progress and the QingDynasty dummy. Regional differences in themale mortality rate can cause selectivity bias,because the number of widows itself is affectedby the number of adultmale deaths. Thus, if theobserved regional differences in honoredwidow chastity cases could have resulted fromdifferences in the number of widows causedby unobserved factors or correlated with malemortality and the technology variable is alsopotentially correlatedwith such factors, asmen-tioned above, the parameter of interest couldbe biased.We include the maize suitability index

(SIM) to proxy the extent of maize productionduring the early Qing Dynasty, when rapidpopulation growth occurred due to the spreadof maize cultivation (Chen and Kung 2016).We also use the frequency of conflicts and nat-ural calamities to control for the effects of themale mortality rate on widow chastity.17

17Wars and natural calamities also affect the female mortalityrate, rendering the total effect on the number of widows ambigu-ous when combined with the effect on the male mortality rate.

Liang, Wang and Yamauchi Cotton Revolution 243

Tab

le3.

Widow

Cha

stity

Selectionan

dAlte

rnativeCrops

lnwidow

(1)

lnwidow

(2)

lnwidow

(3)

lnwidow

(4)

lnwidow

(5)

t×lndist

−0.448***

(0.081)

−0.388***

(0.081)

−0.393***

(0.081)

−0.447***

(0.080)

−0.503***

(0.074)

t×lnagrtax

0.064***

(0.023)

0.073***

(0.022)

0.024(0.024)

t×lndis2pc

−0.055*

(0.031)

−0.047(0.032)

−0.054*

(0.031)

t×lndis2bj

−0.066*

(0.035)

−0.069*

(0.038)

−0.032(0.038)

t×lnSIM

0.035(0.034)

0.007(0.038)

−0.037(0.040)

t×lnconf

0.087**(0.040)

0.08

7**(0.040)

0.042(0.041)

t×lndisa

0.018(0.025)

0.003(0.023)

0.016(0.024)

t×silk

−0.073(0.059)

−0.049(0.057)

−0.031(0.058)

t×lin

en0.035(0.051)

0.014(0.051)

0.043(0.049)

t×tea

0.004(0.004)

0.010***

(0.004)

0.013***

(0.004)

t×lnqu

ota

0.412**(0.164)

t×lnscho

ol−1.092**(0.496)

t×lnshrine

0.524(0.415)

t×lntemple

0.770**(0.328)

Prefectureeffects

YY

YY

YObservation

s522

522

522

522

522

Prefectures

261

261

261

261

261

WithinR

20.845

0.841

0.838

0.850

0.863

Notes:C

oefficien

tsarerepo

rted

withstan

dard

errors,clustered

atprefecture

leve

l,in

parenthe

ses;controlv

ariables

aret,lnpo

pden,t*cap

i,t*lnarea,t*slope,t*a

lti,t*laltit*long

i,t*temp,

andt*lnpreci.

*p<0.1.

**p<0.05

.**

*p<0.01

.

244 January 2021 Amer. J. Agr. Econ.

Column 2 in table 3 shows that the numberof conflicts interacted with the Qing dummyhas a significant positive effect, indicating thatthe male mortality rate increased with warsand conflicts; therefore, the number of widowsalso increased. Despite the significant effect ofconflicts on widow chastity, the interactionterm of the Qing Dynasty and technologicalprogress remains significant.

Impact of Alternative Crops

The previous estimates supported the hypoth-esis that technological progress in cotton tex-tile production led to an increase in thenumber of widow chastity cases. Althoughthe case seems compelling, another potentialcause could be that women were capable ofworking in handicraft sectors other than cot-ton textile production even in cotton produc-tion areas, enabling them to earn enough toindependently support their families. We con-duct placebo tests. In the following placebotests, we highlight whether the prefecture alsoproduced silk, linen, and tea. Thus, we analyzethe evolution of the production of these othercrops and their effects on the number ofwidow chastity cases.

Existing studies confirm that linen had a lowvalue, and linen productionwas insufficient fora widow to support a family (Chang 1999).However, silk prices were higher, and silk pro-duction was more specialized. Silk productionshifted from rural areas to the city and city sub-urbs before the Qing Dynasty (Li 1998).Women in rural areas were rarely engaged insilk production. Tea was a cash crop in whichwomen tended to specialize (Qian 2008), butthe technological progress in tea productionwas insignificant between the Ming and QingDynasties.

Our results indicate that the production ofthese three crops did not have a significantimpact on the number of widow chastity casesin the Ming and Qing Dynasties. Column 3 intable 3 presents the results including all threecrops combined. The coefficients of the inter-action terms between the production of thesecrops and the Qing Dynasty dummy are statis-tically insignificant. Although the interactionterm between tea and the Qing Dynastydummy is positive and significant at the 1%level in Columns 4 and 5, the interaction termof the cotton technology variable remains sta-ble. Though women could specialize in theproduction of the three alternative crops, the

evolution of their production was orthogonalto the chastity effect of cotton technologicalprogress from the Ming to Qing Dynasties.

Location Effects

There are three possible sample selectionissues that could damage the robustness ofthe baseline results. First, because SongjiangPrefecture was close to the sea, the prefecturesthat are geographically adjacent to the techno-logical center of cotton production could alsobe close to domestic harbors for internationaltrade reopened during the late Ming Dynastyand the early Qing Dynasty. This could con-found the widow chastity effect of technologi-cal progress with international trade. Second,the prefectures that were far from traditionalcotton production regions or located in a prov-ince with no cotton production should not beincluded in the sample. Thus, it is not appro-priate to compare them with prefectures incotton production regions or near cotton tex-tile technological centers. Third, the lowerYangtze River region, known as the tradi-tional cotton-planting region and textile centerof China, has a history of a high number ofwidow chastity cases. In contrast, otherregions far from the lower Yangtze Riverregion recorded far fewer chastity cases. It ispossible that the treatment effect of technolog-ical progress decreased substantially alongwith the distance to Songjiang Prefecture.To address these possible issues, we use

three sample restriction strategies. The first isto run a regressionwith a sample that randomlydrops one province at a time. Figure 3 showsthe estimation results of seventeen regressionsfollowing this procedure. It is notable that evenwhen some coastal provinces are dropped,such as Jiangsu, Zhejiang, Fujian, and Guang-dong, the interaction terms of theQing dummyand the distance to Songjiang Prefecture are allstatistically significant, and the coefficientsrange between −0.406 and − 0.668.The second strategy is to use provinces con-

taining at least one cotton-producing prefec-ture, and the third is to use prefectures thatproduced cotton and prefectures next to atleast one prefecture that produced cotton.The estimation results in Columns 1 and 2 ofTable A2 show that the coefficients of theinteraction terms of the Qing dummy and thedistance to Songjiang Prefecture are −0.533and − 0.613, respectively, both statistically sig-nificant and greater than the baseline results.

Liang, Wang and Yamauchi Cotton Revolution 245

The final strategy is to use three circles ofprefectures with different distances to Song-jiang Prefecture (Figure A4). The inner circleincludes eight provinces near Songjiang Pre-fecture, namely, Jiangsu, Zhejiang, Anhui,Henan, Shandong, Hubei, Jiangxi, and FujianProvinces; the middle circle, Guizhou, Guang-dong, Guangxi, Hunan, Shannxi, Shanxi, Zhiliand the eastern prefectures of Sichuan; andthe outer circle, Gansu, Yunnan, and the west-ern prefectures of Sichuan. Columns 3–5 ofTable A2 show that the coefficients of the esti-mates with different subsamples are stablebetween −0.503 and − 0.589, showing thatthe baseline result is not sensitive to sampleselection issues.Another major change between the Ming

and Qing dynasties was that for most of theMing dynasty, trade was boycotted. To testwhether international trade influenced theeffect of cotton technology on widow chastity,we include a new variable, the distance to thenearest of the four ports opened for trade dur-ing 1685–757.18 The result in Column 6 of

Table A2 shows no evidence directly support-ing this hypothesis.

Common Trend Test

Quasi-experimental estimations assume acommon trend between the treatment andcontrol groups. Without the treatment, thecommon trend in the treatment and controlgroups should be consistent, allowing for thedetermination of whether the differencesbetween the treatment and control groupsare caused by the treatment, that is, the cottonrevolution. To test the common trend, thetrend in the treatment and control groupsbefore the treatment’s occurrence is widelyused. We need to test the common trendacross the prefectures located at different dis-tances to Songjiang Prefecture before theQing Dynasty.

More specifically, we test the differencesbetween the Ming and Yuan Dynasties andbetween the Yuan and pre-Yuan Dynasties.This strategy is appropriate because the com-mercialization of the cotton textile sectorbegan at the end of the Ming Dynasty and inthe early Qing Dynasty; that is, no significantchanges were observed in the textile sectorduring most of the Ming, Yuan and pre-YuanDynasties. Therefore, if our hypothesis is

-.7

-.6

-.5

-.4

-.3

-.2

-.1

0

Coe

ffic

ien

t of t*

lnd

ist

Anh

ui

Fu

jian

Gan

su

Gua

ngd

on

g

Gua

ng

xi

Guiz

hou

Hen

an

Hub

ei

Hun

an

Jia

ng

su

Jia

ng

xi

Sha

ndo

ng

Sh

an

nxi

Sha

nxi

Sic

hua

n

Yun

nan

Zh

ejia

ng

Zh

ili

Province dropped in estimate

Figure 3. Estimates with one province dropped at a time [Color figure can be viewed atwileyonlinelibrary.com]

18In the reign of Kangxi in the Qing Dynasty (1685), Chinaopened four ports for trade, namely, Guangzhou, Fuzhou, Ningboand Shanghai. Then, in the reign of Qianlong in the Qing Dynasty(1757), China closed three of the four ports and left only Guang-zhou open for trade.

246 January 2021 Amer. J. Agr. Econ.

correct, the number of widow chastity cases isnot expected to change before the end of theMing Dynasty.

Columns 1–3 of table A3 present the resultsof the common trend tests among the Ming,Yuan and pre-Yuan Dynasties. The results intable A3 and figure 4 show that cotton textiletechnological progress did not have a statisti-cally significant impact on the number ofwidow chastity cases before the Qing Dynasty.The common trend before the Qing Dynastyindicates that in the cotton textile sector, theweaving innovation introduced by DaopoHuang was a necessary but not sufficient con-dition for the evolution of textile commerciali-zation. Other external conditions, such ascotton seeds, domestic demand and the inter-national trade of cotton and textiles, are alsonecessary for textile commercialization. Dur-ing the Ming Dynasty (1371–567), interna-tional trade was boycotted. The stagnation ofinternational trade lasted until the early seven-teenth century, when the boycott was lifted,but at that time, theMingDynasty was alreadyin its twilight, and the prototype of capitalism

did not evolve into an industrial revolution(Brandt, Ma, and Rawski 2014).19

Omitted Variable Test

Another threat is omitted variable bias, whichis one of the key issues that leads to bias incause-effect analysis (Wooldridge 2005).Although we control for geographical vari-ables, cultural variables, selection variables,and competing explanatory variables, omittedvariable bias is still an issue worth considering.To address this, we adopt the test strategydeveloped by Oster (2019), which takes intoaccount the R-squared to establish a rangefrom a controlled treatment effect to an unbi-ased treatment effect. This method builds onthe theory of Altonji, Elder, and Taber(2005), which evaluates omitted variable biasby observing coefficient changes after theinclusion of control variables under theassumption that selection on observables isinformative about selection on unobservables.We perform the Oster (2019) sensitivity test

by assuming the maximum of R-squared(Rmax) to be 1, 0.95, and 0.9, and compare theresults with the results of a fully controlledregression. Table A4 presents the sensitivityof technological progress results reported inColumn 5 of table 3. The coefficients under

-.7-.6

-.5-.4

-.3-.2

-.10

.1.2

.3C

oeffi

cien

t of t

*lndi

st

Yuan vs. before Ming vs. Yuan Qing vs. MingDynasty Pairs

Figure 4. Common trend test [Color figure can be viewed at wileyonlinelibrary.com]

19Both emperors Shunzhi and Kangxi imposed “sea bans” in1655 and 1716, respectively, but the restrictions were subsequentlylifted in 1683 by Kangxi and in 1727 by Yongzheng. Even duringthe bans, they were not implemented restrictively.

Liang, Wang and Yamauchi Cotton Revolution 247

different Rmax are between −4.661 and− 21.462, and the treatment effects are muchlarger than the controlled treatment effect.This result is reasonable because both baselineand fully controlled regressions have R-squared greater than 0.8, suggesting that tech-nological progress in the cotton sector canexplain the majority of the increase in thenumber of widow chastity cases in the Mingand Qing Dynasties. Hence, the omitted vari-able bias is not large enough to threaten therobustness of the baseline results.

Mechanism Analysis

Technological Progress or Cotton Planting

In addition to technological progress in thecotton sector, another important factor is tra-ditional cotton planting, which may have apositive effect on widow chastity. Figure A1illustrates the change in the number of widowchastity cases between the Ming and QingDynasties and the distribution of traditionalcotton-planting regions. Widow chastity incotton-planting areas recorded a largerincrease, suggesting that the coefficient of dis-tance variable might have captured the effectof cotton planting together with technologicalprogress. We introduce a regression replacingthe distance variable with a cotton plantingdummy and then employ specifications toinvestigate the relationship between cottonplanting and technological progress.Columns 1 and 2 of Table A5 present the

results with the cotton planting dummy with-out and with control variables, respectively.The magnitude of the parameter estimatedecreases from 0.429 to 0.219 (49%) whenincluding all control variables in the specifica-tion, and within R-squared increases by0.102. The omitted variable bias test based onOster (2019) reveals a sensitivity pattern dif-ferent from that of technological progress.Due to the large difference between the con-trolled effect and the baseline effect, the inter-action term of the Qing dummy and the cottonplanting dummy decreases substantially tozero, and even becomes negative when Rmaxincreases from 0.9 to 1.When we consider the effects of both cotton

planting and technological progress (Column3 of Table A5), the coefficient of technologicalprogress changes little, while the coefficient ofcotton planting drops further to 0.106, a less

significant level. The above result indicatesthat technological progress in cotton textileproduction has played a more important rolethan cotton planting. In column 4 of table A5,furthermore, the interaction effect of cottonplanting and technological progress is insignif-icant and has a negative coefficient. The tech-nological progress in the textile sector isdominant.

Female Labor Participation and Wages in theTextile Sector

To validate the logic that the cotton revolutioninfluenced widow chastity through increasedfemale labor demands, a key question iswhether the technological progress in the cot-ton textile sector increased labor participationand earnings for women. Due to data limita-tions unique to the Ming and Qing Dynasties,we use labor participation [the industrialenterprise survey; cross section in 1936] andthe wages of workers in industrial enterprises[The worker living survey; panel from1926–30] to investigate labor market pathswith samples including all industries and thetextile industry.20,21 In both surveys, the unitof observation is the city. In columns 1 to 4 oftable 4, industrial survey data were used toanalyze the employment structure, whereasfor columns 5 to 8, labor survey data were usedto analyze the wage structure. The results incolumns 1 and 2 suggest that the textile indus-try employed more female workers in terms ofboth number and share, and in columns 5 and6, female workers in the textile industryearned higher wages (relative to males) thanthose engaged in other industries. Columns3, 4, 7, and 8 present the effects of technologi-cal progress on female employment and wagesin the textile industry. The results are consis-tent with those found in the analysis of widowchastity. The technological progress in the cot-ton textile sector increased employmentopportunities for women and their wages,

20The industrial enterprise survey was conducted by theResource Committee of the National Military Council in 1936(Liu 1937). This survey covered 18,694 industrial enterprises in136 cities and counties; 7,740 enterprises were in the textile indus-try. The data we use in this paper are aggregated in one city(county)-sector cell.

21The worker living survey was conducted by the Ministry ofIndustry and Commerce (MIC) of the National Government from1926–30 (MIC 1930) and involved over 1,104,396 workers in 1975industrial enterprises in 34 cities and counties. This surveyrecorded the wages of workers by gender and sector. The datawe use in this paper are aggregated in a city (county)-sector-yearcell.

248 January 2021 Amer. J. Agr. Econ.

giving widows enough economic security tolive without remarriage.

Conclusions

Widow chastity behavior experienced dra-matic growth during the Ming and QingDynasties, and the number of honored widowchastity cases in the Qing Dynasty was tentimes greater than that in the Ming Dynasty.This phenomenon has long been consideredan impact of Confucian culture, which belongsto the domain of social-cultural studies,whereas in our study, we discussed the institu-tional background and economic roots of thisphenomenon from an economic standpoint.In the empirical analysis, we used proximityto Songjiang Prefecture, which emerged asthe center of technological progress in the cot-ton textile sector after the Song Dynasty, as akey variable to explain the increase in the inci-dence of widow chastity observed at the pre-fecture level using a fixed effects panel modeland found a significant positive relationship.Several additional results reinforced the

above conclusion. To isolate the technologyeffect from cultural influences, we consideredofficial, unofficial, and folk-custom influencesof Confucianism on chastity and martyrdombehavior. The impact of technological pro-gress in the cotton textile sector remained sta-ble. Furthermore, in the falsification test, weused martyrdom statistics to investigate theeffect of technological progress and found thateconomic factors had no effect on irrationalbehaviors such as martyrdom.To assess the potential effects of the chastity

award selection process, we used agriculturaltax revenue per capita (income level), distancebetween the prefecture seat and the province’scapital city, distance between the prefectureseat and Beijing (political influence), the suit-ability index of maize planting (innovation),and the frequency of conflicts and of naturalcalamities (mortality). Though these effectsexisted, the cotton revolution continued toexplain the change in the incidence of widowchastity between theMing andQingDynasties.In the placebo test, we used other produc-

tion activities, such as the production of linen,silk, or tea, as placebos and found that theeffects of these activities were all insignificant.In the common trend test, we found thatbefore theMingDynasty, cotton technologicalprogress did not exist, so the change in widowT

able

4.Fem

aleLab

orPartic

ipationan

dW

ages

intheTex

tileIndu

stry,1

930s

Allindu

stries

Textile

indu

stry

Allindu

stries

Textile

indu

stry

lnfemwork(1)

shfemwork(2)

lnfemwork(3)

shfemwork(4)

lnfemwage(5)

wageratio

(6)

lnfemwage(7)

wageratio

(8)

textile

2.816***

(0.650)0.285***

(0.091)

0.697***

(0.227)0.164**(0.057)

lndist

−0.580(0.905)−0.191***

(0.060)

−2.179***

(0.000)−0.468***

(0.000)

Prefectureeffects

YY

NN

YY

NN

Yeareffects

NN

NN

YY

YY

Observation

s918

918

161

161

1860

1860

370

370

Adj.R

20.269

0.225

0.394

0.600

0.217

0.165

0.294

0.181

Notes:C

oefficien

tsarerepo

rted

withstan

dard

errors,colum

n(1)–(4)clusterat

theprov

ince

leve

land

column(5)–(8)at

theprefecture

leve

l,in

parenthe

ses;controlv

ariables

ofestimates

ininclud

elnpo

pden,cap

i,lnarea,slope,alti,lalti,

long

i,temp,

lnqu

ota,

lntemple,lnshrine,lnschoo

l,lnag

rtax

,lnd

is2p

c,lndis2bj,lnS

IM,lncon

f,lndisa,silk

,linen,a

ndtea.

*p<0.1.

**p<0.05

.**

*p<0.01

.

Liang, Wang and Yamauchi Cotton Revolution 249

chastity was parallel across the areas. Todetermine the extent of the omitted variablebias, we adopted the approach proposed byOster (2019) and found that the effect of cot-ton technological progress survived the test.Did cotton technological progress or cotton

planting or both drive the observed increase inthe incidence of widow chastity? It is true thatwithout cotton production, textile productionis seriously constrained, but what matters hereis the magnitude of induced labor demands forwomen. Our analysis indicates that technolog-ical progress in the cotton textile sector is thedominant factor determining the incidence ofwidow chastity. Did the above technologicalprogress truly increase demands for femalelaborers? The evidence from later periodsindicates that female employment and wagesin the textile sector increased as proximity tothe core of technological progress in the sectorincreased. Labor participation and wagesamong women were greater in the textileindustry than in other industries. Thus, weconclude that as women’s economic statuswas enhanced by the cotton revolution, thenumber of widows who chose to not remarryincreased accordingly.

Supplementary Material

Supplementary material are available atAmerican Journal of Agricultural Economicsonline.

References

Aisingoro, Kungang, et al. 1899. Daqing Hui-dian (Imperially Established Institutesand Laws of The Great Qing Dynasty).Shanghai, China: Shanghai BookCompany.

Aisingoro, Yunlu, et al. 1976. Shangyu QiwuYifu (Emperor’s Comments and Replieson Banners Affairs). Taipei, Taiwan: Tai-wan Student Bookstore.

Alesina, Alberto F, Paola Giuliano, andNathan Nunn. 2013. On the Origins ofGender Roles: Women and the Plough.Quarterly Journal of Economics 128(2):469–530.

Altonji, Joseph G, Todd E Elder, andChristopher R Taber. 2005. Selection onObserved and Unobserved Variables:Assessing the Effectiveness of Catholic

Schools. Journal of Political Economy113(1): 151–84.

Bai, Ying, and Ruixue Jia. 2016. Elite Recruit-ment and Political Stability: The Impact ofthe Abolition of China’s Civil ServiceExam. Econometrica 84(2): 677–733.

Baker, Michael, Jasmin Kantarevic, and EmilyHanna. 2004. The Married Widow: Mar-riage Penalties Matters. Journal ofEuropean Economic Association 2: 634–64.

Becker, Gary S. 1973. A Theory of Marriage:Part I. Journal of Political Economy 8:813–46.

. 1991. A Treatise on the Family.Cambridge, UK: Harvard UniversityPress.

Becker, Gary S, and Kevin M Murphy. 2000.Social Economics, Market Behavior in aSocial Environment. Cambridge, UK:Harvard University Press.

Bennett, Judith M. 1981. Medieval PeasantMarriage: An Examination of MarriageLicense Fines in Liber Gersumarum. InPathways to Medieval Peasants, ed. JamesA Raftis, 193–246. Toronto, Canada: Pon-tifical Institute of Mediaeval Studies.

Birge, Bettine. 1995. Levirate Marriage andthe Revival of Widow Chastity in YuanChina. Asia Major 8: 107–46.

Brandt, Loren, Debin Ma, and Thomas GRawski. 2014. From Divergence to Con-vergence: Reevaluating the Historybehind China’s Economic Boom. Journalof Economic Literature 52(1): 45–123.

Brien, Michael J, Stacy Dickert-Conlin, andDavid AWeaver. 2004. Widows Waiting toWed? (Re)Marriage and Economic Incen-tives in Social Security Widow Benefits.Journal of Human Resources 39: 585–623.

Chang, Pin-tsun. 1999. Mingqing Shiqi GuafuShoujie de Fengqi: Lixing Xuanze deWenti (Widow Chasity in Ming-Ch’ingChina: A Problem of Rational Choice). 9Xinshixue (New History) 10(2): 29–76.

Chen, Gaoyong, et al. 1986. Zhongguo LidaiTianzai Renhuo Biao (ChronologicalRecord of Natural andMan-MadeDisastersin China). Shanghai, China: ShanghaiBookstore.

Chen, Shuo, and Kai-sing K Kung. 2016. OfMaize and Men: The Effect of a NewWorld Crop on Population and EconomicGrowth in China. Journal of EconomicGrowth 21(1): 71–99.

Cready, CynthiaM,MarkA Fossett, and Jill KKiecolt. 1997. Mate Availability and Afri-can American Family Structure in the

250 January 2021 Amer. J. Agr. Econ.

U.S. Nonmetropolitan South, 1960–1990.Journal of Marriage and the Family 59:192–203.

De Moor, Tine, and Jan L Van Zanden. 2010.Girlpower: The European Marriage Pat-tern and Labour Markets in the NorthSea Region in the Late Medieval andEarly Modern Period. Economic HistoryReview 63(1): 1–33.

Durante, Ruben. 2009. Risk, Cooperation andthe Economic Origins of Social Trust: AnEmpirical Investigation. Rhode Island,England: Mimeo, Brown University.

Foreman-Peck, James, and Peng Zhou. 2014.The Rise of the English Economy1300–1900: A Lasting Response to Demo-graphic Shocks.” Cardiff Business SchoolWorking paper, No. E2014/3. Cardiff,Wales: Cardiff University.

Giuliano, Paola, and Nathan Nunn. 2017.Understanding Cultural Persistence andChange.” NBER working paper,No. 23617. Cambridge, MA: NBER.

Guiso, Luigi, Paola Sapienza, and LuigiZingales. 2008. Social Capital as GoodCulture. Journal of the European Eco-nomic Association 6(2–3): 295–320.

. 2016. Long-Term Persistence. Journalof the European Economic Association 14(6): 1401–36.

Guo, Songyi. 1987. Qingdai Renkou Wenti yuHunyin Zhuangkuang de Kaocha(Historical Review of Population Issuesand Marital Status in Qing Dynasty).Zhongguoshi Yanjiu (China History Stud-ies) 3: 123–37.

. 2000. Ethics and Life: Marriage inQing Dynasty Lunli yu Shenghuo: Qing-dai de Hunyin Guanxi (Ethics and Life:Marriage in Qing Dynasty). Beijing,China: Commercial Press.

Guwargiya, Muzhanga, et al. 2008. JiaqingChongxiu Daqing Yitongzhi (Revised ver-sion of geographical and biographicalchronicles of Great Qing Dynasty in Jiaq-ing). Shanghai, China: Shanghai AncientBooks Publishers.

Hajnal, John. 1965. European Marriage Pat-tern in Historical Perspective. In Popula-tion in History, ed. David V Glass andDavid Eversley, 101–45. London: Arnold.

Hanawalt, Barbara A. 1986. The Ties thatBound: Peasant Families in MedievalEngland. New York: Oxford UniversityPress.

. 1993. Remarriage as an Option forUrban and Rural Widows in Late

Medieval England. In Wife and Widow inMedieval England, ed. SS Walker. AnnArbor, Michigan: University of MichiganPress.

Hansen, Casper W, Peter S Jensen, andChristian V Skovsgaard. 2015. ModernGender Roles and Agricultural History:The Neolithic Inheritance. Journal ofEconomic Growth 20(4): 365–404.

Holmgren, Jennifer. 1985. The EconomicFoundation of Virtue: Widow-Remarriedin Early and Modern China. AustralianJournal of Chinese Affair 13: 1–27.

. 1986. Observations on the Marriageand Inheritances Practices in Early Mon-gol and Yuan Society, with Particular Ref-erence to the Levirate. Journal of AsianHistory 20: 127–92.

Ju, Deyuan. 2001. Qingdai Huangzu RenkouCeji (Demographic Records of the QingImperial lineage). In Qingdai HuangzuRenkou Xingwei he Shehui Huanjing(Demographic Behavior and Social Envi-ronment of the Qing Imperial LineagePopulation), ed. James Lee and SongyiGuo, 170–90. Beijing, China: Peking Uni-versity Press.

Li, Bozhong. 1998. Agricultural Developmentin Jiangnan, 1620–1850. New York: St.Martin’s Press.

Liang, Fangzhong. 1980. Lidai Hukou, Tudi,Tianfu Tongji (Historical statistics onHukou, land, and land tax of China ).Shanghai, China: People’s Press.

Lichter, Daniel T, Felicia B LeClere, andDiane K McLaughlin. 1991. Local Mar-riage Markets and the Marital Behaviorof Black and White Women. AmericanJournal of Sociology 96: 843–67.

Lin, Justin Y. 1995. The Needham Puzzle:Why the Industrial Revolution Did NotOriginate in China. Economic Develop-ment and Cultural Change 43(2): 269–92.

Liu, Dajun. 1937. Zhongguo Gongye DiaochaBaogao (China industry survey report).Beijing, China: Institute of EconomicStatistics.

Lowes, Sara, Nathan Nunn, James ARobinson, and Jonathan Weigel. 2017.The Evolution of Culture and Institutions:Evidence from the Kuba Kingdom. Econ-ometrica 85(4): 1065–91.

Ministry of Industry and Commerce. 1930.Quanguo Gongren Shenghuo ji GongyeShengchan Diaocha Tongji Zongbaogao(National Survey and Statistics Report onWorkers’ Living and Industrial

Liang, Wang and Yamauchi Cotton Revolution 251

Production), Nanjing, China: Ministry ofIndustry and Commerce.

Nunn, Nathan, and Leonard Wantchekon.2011. The Slave Trade and the Origins ofMistrust in Africa. American EconomicReview 101(7): 3221–52.

Oster, Emily. 2019. Unobservables Selectionand Coefficient Stability: Theory and Evi-dence. Journal of Business and EconomicStatistics 37(2): 187–204.

Pomeranz, Kenneth. 2005. Women’s Workand the Economics of Respectability. InGender in Motion: Divisions of Laborand Cultural Change in Late Imperialand Modern China, ed. Bryna Goodmanand Wendy Larson, 239–63. Lanham,MD: Rowman and Littlefield Publishers.

Preston, Samuel H, and Alan T Richards.1975. The Influence of Women’s WorkOpportunities on Marriage Rates.Demography 12: 209–22.

Qian, Nancy. 2008. Missing Women and thePrice of Tea in China: The Effect of Sex-specific Earnings on Sex Imbalance.Quar-terly Journal of Economics 123: 1251–85.

Records Commission. 1985. Qing Shilu: Shi-zong Xian huangdi Shilu (A Record ofQing Dynasty: Shizong Xian EmperorRecords). Beijing, China: Zhonghua BookCompany.

Salisbury, Laura. 2017. Women’s Incomeand Marriage Markets in the UnitedStates: Evidence from the Civil WarPension. Journal of Economic History77(1): 1–38.

Shen, Shixing. 1936.Ming Huidian (Statutes ofthe Ming Dynasty). Shanghai, China: TheCommercial Press.

Tien, Jukang. 1988. Male Anxiety and FemaleChastity: A Comparative Study of ChineseEthical Values in Ming-Ch’ing Times.Leiden, The Netherlands: Brill.

Todd, Barbara J. 1985. The RemarryingWidow: A Stereotype Reconsidered. InWomen in English Society, ed. Mary Prior,60–68. London: Methuen&Co. Ltd.

Voigtländer, Nico, and Hans-Joachim Voth.2013. How the West ‘Invented’ FertilityRestriction. American Economic Review103(6): 2227–64.

Wooldridge, Jeffrey. 2005. Unobserved Het-erogeneity and Estimation of AveragePartial Effects. In Identification and Infer-ence for Econometric Models: Essays inHonor of Thomas Rothenberg, ed. ADonald and J Stock, 27–55. Cambridge,U.K.: Cambridge University Press.

Wu, Cheng-Ming, and Dixin Xu. 1985.Zhong-guo Zibenzhuyi de Mengya (The Seeds ofCapitalism in China). Beijing, China: ThePeople’s Publishers.

Xu, Guangqi. 1930. Nongzheng Quanshu(Complete works on agriculture ).Shanghai, China: The Commercial Press.

Xue, Melanie M. 2014. Textiles and the Histor-ical Emergence of Gender Equality inChina. Fairfax, Virginia: Mimeo, GeorgeMason University.

Yan, Zhongping. 1955. Zhongguo Mian-fangzhi Shigao (China Cotton Textile His-tory). Beijing, China: Science Press.

Yuchtman, Noam. 2017. Teaching to theTests: An Economic Analysis of Tradi-tional and Modern Education in LateImperial and Republican China. Explora-tions in Economic History 63: 70–90.

252 January 2021 Amer. J. Agr. Econ.