UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Essays on development economics in China Zhou, Weina 2014

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2014_september_zhou_weina.pdf [ 779.73kB ]
Metadata
JSON: 24-1.0167500.json
JSON-LD: 24-1.0167500-ld.json
RDF/XML (Pretty): 24-1.0167500-rdf.xml
RDF/JSON: 24-1.0167500-rdf.json
Turtle: 24-1.0167500-turtle.txt
N-Triples: 24-1.0167500-rdf-ntriples.txt
Original Record: 24-1.0167500-source.json
Full Text
24-1.0167500-fulltext.txt
Citation
24-1.0167500.ris

Full Text

Essays on Development Economics in ChinabyWeina ZhouB.E., Tokyo Institute of Technology, 2007M.A., Tokyo Institute of Technology, 2008A THESIS SUBMITTED IN PARTIAL FULFILLMENTOF THE REQUIREMENTS FOR THE DEGREE OFDoctor of PhilosophyinTHE FACULTY OF GRADUATE AND POSTDOCTORALSTUDIES(Economics)The University of British Columbia(Vancouver)June 2014c©Weina Zhou, 2014AbstractRecent research has stressed the role of historical events on economic develop-ment. This thesis aims at understanding impacts of historical events on China’scurrent economic outcomes. The second chapter analyzes the effect of the num-ber of brothers an individual has on that individual’s household savings rate underthe current underdeveloped household financial market in urban China. I showthat having an additional brother reduces an individual’s household savings rateby at least five percentage points. Brothers help households by (1) sharing risks,providing a source of informal borrowing and (2) sharing the cost of support-ing parents. In the third and fourth chapter I investigate the long-term impact ofthe send-down policy. Under the send-down policy (1968–1978) during the Chi-nese Cultural Revolution, more than 16 million youths were forced to move torural areas and carry out hard manual labor. I find that the sent-down males weresignificantly more likely to have had education upgrading after the Cultural Rev-olution. Conditional on education upgrading, the sent-down males earn higherincome than the non-sent-down males who also received education upgrading.iiPrefaceThis dissertation is original, unpublished, independent work by the author, WeinaZhou.iiiTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiiList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xAcknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 Brothers, Household Financial Markets and Savings Rate in China . 52.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52.2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122.2.1 Financial Markets and Household Borrowing Resources . 12iv2.2.2 Facts: Household Savings Rate by Number of Brothersand Sisters . . . . . . . . . . . . . . . . . . . . . . . . . . 152.2.3 Changes in Demographics and China’s Savings Rate Puzzle 162.3 The Impact of the Number of Brothers on Households’ Savings Rate 182.3.1 Identification . . . . . . . . . . . . . . . . . . . . . . . . 182.3.2 Results: the Impact of the Number of Brothers on House-hold Savings Rate . . . . . . . . . . . . . . . . . . . . . . 242.4 Why Brothers Reduce the Savings Rate: Risk Sharing/ExtendingBorrowing Limits and Supporting Parents . . . . . . . . . . . . . 272.5 Robustness Check . . . . . . . . . . . . . . . . . . . . . . . . . . 372.5.1 Son Preference . . . . . . . . . . . . . . . . . . . . . . . 372.5.2 Brothers Effect of Individuals Born After the One-ChildPolicy . . . . . . . . . . . . . . . . . . . . . . . . . . . . 382.6 How the Decline in the Number of Brothers in Households CouldExplain the Savings Rate Puzzle . . . . . . . . . . . . . . . . . . 402.7 Conclusion and Discussion . . . . . . . . . . . . . . . . . . . . . 423 How Does a Hard Manual Labor Experience during Youth AffectLater Life? The Long-term Impact of the Send-down Program dur-ing the Chinese Cultural Revolution . . . . . . . . . . . . . . . . . . 643.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 643.2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 723.2.1 The Send-down Policy . . . . . . . . . . . . . . . . . . . 72v3.2.2 Send-down Experience and Documentations . . . . . . . . 743.3 Sample Restrictions . . . . . . . . . . . . . . . . . . . . . . . . . 763.3.1 Treatment Group and Comparison Group . . . . . . . . . 763.3.2 Family Background . . . . . . . . . . . . . . . . . . . . . 773.4 Education . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 793.4.1 Education Interruption during the Cultural Revolution . . . 793.4.2 Education Upgrading after the Cultural Revolution . . . . 803.4.3 The Send-down Effect on Education Upgrading . . . . . . 813.5 The Send-down Effect on Income . . . . . . . . . . . . . . . . . . 833.5.1 Identification . . . . . . . . . . . . . . . . . . . . . . . . 843.5.2 Estimation Results . . . . . . . . . . . . . . . . . . . . . 853.6 The Send-down Effect on Computer Ownership . . . . . . . . . . 873.7 Robustness Check . . . . . . . . . . . . . . . . . . . . . . . . . . 893.7.1 IV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 893.7.2 Other Robustness Checks . . . . . . . . . . . . . . . . . . 933.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 964 The Long-term Impact of the Send-down Experience: Happiness inLife, Political Attitudes, and Investment in Children . . . . . . . . . 1124.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1124.2 Happiness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1154.3 Political Attitudes . . . . . . . . . . . . . . . . . . . . . . . . . . 116vi4.4 Intergenerational Effect: Investment in the Next Generation’s Ed-ucation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1194.5 Robustness Check . . . . . . . . . . . . . . . . . . . . . . . . . . 1234.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1235 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134A Appendix for Chapter 2 . . . . . . . . . . . . . . . . . . . . . . . . . 142B Appendix for Chapter 3 . . . . . . . . . . . . . . . . . . . . . . . . . 149viiList of TablesTable 2.1 Fraction of Male Siblings by Total Number of Siblings . . . . . 50Table 2.2 Test of Random Assignment of the Number of Brothers Con-ditional on the Number of Siblings . . . . . . . . . . . . . . . 51Table 2.3 The Impact of Number of Brothers on Household Savings Rate 52Table 2.4 Brother’s Sharing Risks / Extending Borrowing Limits Effect . 54Table 2.5 Gender Differences in Supporting Parents . . . . . . . . . . . . 56Table 2.6 The Impact of Number of Brothers on Household Savings Rates- the Effect of Supporting Parents . . . . . . . . . . . . . . . . 57Table 2.7 The Brother Effect in Different Income Groups and Asset Groups 58Table 2.8 Robustness Check: Son Preference . . . . . . . . . . . . . . . 60Table 2.9 IV Estimation Results for Individuals Born after the One ChildPolicy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62Table 3.1 Individual Characteristics and Family Background during theCultural Revolution . . . . . . . . . . . . . . . . . . . . . . . 99Table 3.2 Probit Estimation of Send-down . . . . . . . . . . . . . . . . . 100viiiTable 3.3 Number of Students (10,000 person) . . . . . . . . . . . . . . 102Table 3.4 Education Upgrading . . . . . . . . . . . . . . . . . . . . . . 103Table 3.5 Probit Estimation: the Impact of Send-down Experience on Ed-ucation Upgrading . . . . . . . . . . . . . . . . . . . . . . . . 104Table 3.6 Descriptive Statistics of Monthly Income by Gender and Edu-cation Upgrading . . . . . . . . . . . . . . . . . . . . . . . . . 105Table 3.7 The Impact of Send-down Experience on Income . . . . . . . . 106Table 3.8 The Impact of Send-down on Having Computers . . . . . . . . 107Table 3.9 The Impact of Send-down Experience on Males’ Income andComputer Ownership (IV) . . . . . . . . . . . . . . . . . . . . 108Table 3.10 Other Robustness Checks . . . . . . . . . . . . . . . . . . . . 110Table 4.1 Descriptive Statistics, by Send-down Experience . . . . . . . . 125Table 4.2 Ordered Logit Regression Results: the Impact of the Send-down Experience on Life Happiness . . . . . . . . . . . . . . . 126Table 4.3 Probit Regression Results: Democracy Means Government ShouldMake Decisions on Behalf of People . . . . . . . . . . . . . . 127Table 4.4 Probit Regression Results: the Impact of the Send-down Expe-rience on Communist Party Membership Status . . . . . . . . . 128Table 4.5 The Impact of the Send-down Experience on Other Attitudes . 129Table 4.6 Intergenerational Impact of Send-down: Investment on Chil-dren’s Education . . . . . . . . . . . . . . . . . . . . . . . . . 130Table 4.7 Robustness Check . . . . . . . . . . . . . . . . . . . . . . . . 131ixList of FiguresFigure 2.1 Sources for Borrowing Money in Urban China: Self-Reportsof Borrowing Resource if One Encounters a Negative Shock(Percentage of Respondents) . . . . . . . . . . . . . . . . . . 44Figure 2.2 Age Profile Household Savings Rate by Number of Brothersand Sisters . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45Figure 2.3 Age Profile Household Savings Rate by Number of Brothersand Sisters - Households with No Living Parents . . . . . . . 46Figure 2.4 Number of Brothers and Sisters by Individuals’ Birth Year . . 47Figure 2.5 Number of Brothers and Household Savings Rate in UrbanAreas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48Figure 2.6 Source of Variation: Average Household Savings Rate by Num-ber of Brothers for a Given Number of Siblings . . . . . . . . 49Figure 3.1 Number of Youth Sent to Rural Areas by Year . . . . . . . . . 97Figure 3.2 Send-down Proportion by Year of Birth . . . . . . . . . . . . 98xAcknowledgmentsI am enormously grateful to my two thesis co-supervisors Thomas Lemieux andSiwan Anderson for their patience, emotional encouragement and continuous sup-port of my Ph.D study and research. Although the topics of my thesis only relateto China and are not directly related to their research topics, they have alwaysshown great interest in my work at various stages of this project. Because of theirefforts I was highly motivated about my research from the beginning to the endof my project. Most importantly, under their supervision I very much enjoyed theprocess of undertaking this research.I would like to express my sincere gratitude to my committee member and mysecond year paper supervisor Ashok Kotwal. Without his encouragement and ad-vice, I could hardly stand by myself after two very stressful years in the beginningof my PhD program.I also benefited greatly from Hiroyuki Kasahara’s suggestions which helpedme sort out several important issues in my paper before I entered the job mar-ket. My sincere thanks also goes to Nicole Fortin, Vadim Marmer, Marit Rehavi,Joshua Gottlieb and Matilde Bombardini for offering me valuable suggestions onthis project.Finally, I am grateful to my parents, my partner and my friends who haveprovided tremendous mental support and made my PhD student life hugely en-joyable.xiChapter 1IntroductionThis thesis consists of three papers on the economics in China. The second chap-ter investigates the consequences of the weak household financial market in Chinaand how it affects households’ savings rate. The third and fourth chapter inves-tigates the long term impact of a historical event in China: the send-down move-ment (1968-1978). The third chapter focuses on the effect of send-down on in-dividuals’ education and incomes. The fourth chapter focuses on the impact ofsend-down on a broader set of outcomes, such as subjective well-being, politicalattitude, and investment in children’s education.The second chapter analyzes the effect of the number of brothers an individualhas on his/her household’s saving rate under the current underdeveloped house-hold financial market in urban China. I look at this question using data fromthe China General Social Survey (CGSS) that randomly samples a respondent byhousehold, and provides information about income, expenditures, the number of1brothers and sisters each respondent has, and other social-economic characteris-tics. The estimates indicate that having an additional brother reduces an individ-ual’s household savings by at least 5 percentage points. Having more brothershelps by (1) sharing risks, providing a source of informal borrowing and (2) shar-ing the cost of supporting parents.In the estimation, I exploit the fact that, under the natural selection of gen-der, the number of brothers an individual has is random, conditional on the totalnumber of siblings. I argue that this identification strategy is valid for individualsborn prior to the introduction of the One Child Policy in 1979 who also hold anurban area residence card. By contrast, the strategy may not be valid for other in-dividuals because of sex selection due to selective abortion (in more recent years)or female infanticide (in rural areas). For individuals born after 1979, I use OneChild policy fines to instrument the number of brothers. For individuals born af-ter 1979, the One Child Policy fines are used as instruments for the number ofbrothers.The fact that having more brothers reduces saving rate helps account for 30%of the increased aggregate saving rate in China between 1990 and 2005, becausethere was a large decrease in the average number of brothers per household dur-ing that time. This paper suggests that sisters play a minor role in affecting ahousehold’s saving rate in China, mainly because of cultural norms. This paperalso suggests that the Chinese government might want to consider developing thehousehold financial market as soon as possible due to the change in demographicstructure.2Under the send-down policy (1968-1978) during the Chinese Cultural Revo-lution, more than 16 million youths were forced to move to rural areas and carryout hard manual labor. This study analyzes the long-term impact of such an ex-perience on income when these youths reached 40-55 years of age. Sent-downmales were significantly more likely to upgrade their education after the CulturalRevolution, which caused education interruption for an entire generation. Thesent-down males who upgraded their education earn a 10% higher income thannon-sent-down males who also upgraded their education. Conditional on educa-tion upgrading, the sent-down males are also more likely to have computers athome. These findings are robust against a variety of controls for family back-ground. The send-down experience has had no significant impact on females.Chapter 4 suggests that people who were sent down are significantly less likelyto be happy in their lives and less likely to be believe in the governments demo-cratic capacity, compared to people who were not sent down. They are also lesslikely to become Communist Party members. However, the send-down experiencedoes not affect individuals’ political attitudes towards voting or in seeing ordinarypeople exercise power through decision making. Further, it has no significant ef-fect on how individuals trust others. The potential reason for this could be thatthe send-down experience was caused by a decision made almost exclusively byone person—it was not caused by being cheated by strangers. Neither those whowere sent down nor those who were not sent down have any experience in votingor seeing ordinary people make decisions. All of this evidence supports the ideaof an experience-based process for formulating attitudes and beliefs.3The hard manual labor experience, however, induced people who were sentdown to invest significantly more in their children’s education in hopes of securinga better life for their children. The send-down experience could have made thembetter understand the hardships associated with doing manual labor and, thereby,facilitated the realization that only education could help their children avoid theexperience of hard manual labor.4Chapter 2Brothers, Household FinancialMarkets and Savings Rate in China2.1 IntroductionIt is well documented that the corporate financial market in China is underdevel-oped despite China’s impressive GDP growth in recent decades (Song et al. 2011;Ayyagari et al. 2010; Guariglia et al. 2011; Chen et al. 2011; Allen et al. 2005).Private entrepreneurs usually find it difficult to borrow from banks and must relylargely on the financial resources from their own networks such as family mem-bers or relatives (Cai et al. 2013; Estrin and Prevezer 2011). To date, however,little attention has been paid to the household financial market even though thedegree of development in the household financial market is no better than that ofthe corporate financial market. According to the 2009 China Family Panel Study,5even in Beijing, Shanghai, and Guangdong, China’s most developed regions, morethan 80% of debtors borrowed from family members or relatives in 2008, whilefewer than 20% borrowed from financial institutions. At the same time, house-holds also encounter large uncertainties. Health care reforms, pension reforms,and rising income uncertainties cause households to have high savings rates (Cha-mon and Prasad 2010; Chamon, Liu, and Prasad 2010). The household savingsrate rose from 16 percent in 1990 to 24 percent in 2005 in urban areas.1In developing countries where household financial markets are underdevel-oped, research has provided evidence that shows how extended family membershelp each other by sending transfers and gifts to households that receive negativeeconomic shocks (Fafchamps and Quisumbing 2008; Fafchamps 2008). However,so far it is unknown to what extent the existence of family members, which couldrepresent potential transfers, affects a household’s savings rate and whether thegender of the family members matters.This paper explores the consequences of a weak household financial marketby studying the effect of brothers, the most important members of a household inthe extended family, on the household savings rate in urban China. This is one ofthe first papers to estimate the siblings’ effect on a household’s savings rate usingmicro level data.Although individuals rely largely on their brothers under the current environ-ment of increasing uncertainties and incomplete financial markets, population1The savings rate is defined as 1−LivingExpenditure/DisposableIncome. Data source: ChinaStatistical Year Book.6control policies such as the One-Child Policy (1979) made the situation evenworse. In contrast to the individuals born during the baby boom period (1945–1978), who on average have more than three siblings, the One-Child Policy gen-eration have fewer or even no siblings.2 They suffer from a lack of a family-based safety net in addition to incomplete financial markets. A simple calculationsuggests that the decline in the average number of brothers can explain at leastone-third of the increased aggregate household savings rate.I estimate the brothers effect on household savings rate by using data fromthe China General Social Survey (CGSS) that randomly samples a respondentby household, and provides information about household income, expenditures,the number of brothers and sisters the respondent has, and other social-economiccharacteristics. See data appendix for detailed information of this data set and therelation of respondents with other household members.In estimating the effect of the number of brothers of an individual on the in-dividual’s household savings rate, endogeneity problems arise: the number ofbrothers of a individual could potentially be correlated with that individual’s unob-served characteristics such as his/her parents’ preferred number of children. Thispaper found that conditional on the number of siblings of individuals, the gen-der of the siblings can be considered as a random assignment by nature for urbanresidents born during the baby boom (1945–1978)3. The gender assignments ofsiblings by nature help us to identify the effect of having a brother instead of a2Although the overall fertility rate was high during the baby boom period, it was low duringthe Chinese famine period (1959–1961).3Urban residents are defined as individuals with urban resident cards.7sister (a relative effect).The identification strategy relies on the assumption that conditional on thenumber of siblings, the gender of the siblings is only determined by nature. It iswell known that China had a growing missing female problem in recent decades(Qian 2008; Anderson and Ray 2010). However, I find that it was unlikely thatparents were able to control the gender of their children for a given family sizeamong urban residents born during the baby boom (1945–1978).4 The main rea-son is that ultrasound technology—a technology that can identify gender beforebirth—was introduced in the 1980s, which is after the baby boom. In addition,female infanticide was much more difficult to practice in urban areas. In thiscase, it was unlikely that urban households would risk criminal prosecution forson preference.As a robustness check, for individuals born after the One-Child Policy (1979),I use within-region across-time variation in the One-Child Policy fine to instru-ment the number of brothers of individuals. As the One-Child Policy had a signif-icant impact on the gender ratio and fertility decision (Ebenstein 2010), the One-Child Policy fine had a direct impact on the number of brothers that an individualhas. The results of the IV estimation are consistent with the main results in thispaper: having more brothers reduces a household’s savings rate in urban China;brothers reduce the savings rate by sharing risks with the individual’s household.I find that having a brother instead of a sister reduces the household savings4The baby boom was induced by family planning policies introduced in the 1950s that carriedon until the early 1970s.8rate by at least five percentage points. If sisters also behave like brothers and affectthe household savings rate, the estimated relative effect would be a lower bound.That is, the absolute effect (i.e., having one more brother rather than not) wouldbe larger than the relative effect (i.e., having a brother instead of a sister). The sta-tistical evidence reveals that sisters have almost no effect on a household’s savingsrate for the baby boom generation. Therefore, the estimated relative effect of abrother is likely to be the same as the absolute effect. The lack of an effect of sis-ters on the savings rate may result from the relatively weak connections betweenfemale and male siblings, and between parents and daughters in Chinese culture.Having said this, interestingly, as the number of siblings declines because of thechange in family planning policy, sisters also affect the savings rate like brothersfor the young generation. Young households may use sisters as a substitute forbrothers when there are too few brothers.I show that brothers can reduce a household’s savings rate through two chan-nels: (1) sharing risks and extending borrowing limits, and (2) sharing the costof supporting parents. In order to examine the effect of risk sharing/extendingborrowing limits, this paper tests the effect that brothers have on households withdifferent levels of (a) wage uncertainties, (b) bonus uncertainties, (c) health risks,(d) regional financial development and (e) income or asset levels. The estimationresults are consistent with the risk-sharing/extending-borrowing-limits hypothe-sis: households that encounter larger wage and bonus uncertainties, have higherhealth risks, live in a financially less developed province, have lower incomes orhave fewer assets have a larger brothers effect. The robust and consistent results9suggest a strong risk-sharing/extending-borrowing-limits effect of having broth-ers.In Chinese culture, the expectation is that parents will be supported by theirmale children (Banerjee et al. 2013).5 A household with several brothers wouldneed to save less for their parents’ risks, in particular risks from medical expendi-ture, which are largely shared among the brothers. To test the parent-supportingaspect, I utilize information on whether parents are deceased. Once parents havepassed away, brothers no longer play a role in sharing parents’ risks. The differ-ence in the number of parents still living helps to identify the parent-supportingeffect of brothers.Recent papers have emphasized that change in the demographic structure couldaffect household savings rates because of the effect of the intergenerational sup-port. Ge, Yang, and Zhang (2012) explore the regional variation in One-ChildPolicy fines to examine the effect of changing demographics on household sav-ings rates. Choukhmane, Coeurdacier, and Jin (2013) estimate an OLG modelincorporating endogenous fertility, intergenerational transfers and human capi-tal accumulation, and find that changes in the demographics explain more thanone-third of the rise in the aggregate savings rate. Banerjee, Meng, Porzio, andQian (2013) suggest that the partial equilibrium model could overstate the effectof changing demographics on the savings rate. Wei and Zhang (2011) suggestthat the rising gender ratio induced parents to save more for their male children,5This is the main reason why we observe a large increase in the male–female gender ratio ofnewborns after the “One-Child Policy.”10helping them to secure a better outcome in the marriage market.This is one of the first papers to emphasize that in addition to the intergener-ational support effect, the risk-sharing effect among brothers could also explainwhy changes in demographics could raise the aggregate household savings rate.Furthermore, the role of risk sharing/extending the borrowing limits among fam-ily members could vary greatly depending on the gender of a family member. Itdiscovers a gender difference in China in a new dimension.This paper also helps to explain why there is mixed evidence regarding whetherthe decreasing dependency ratio could explain the rising savings rate. Modiglianiand Cao (2004) use long-term national-level data and find that the decrease inboth the young and old population contributes to the rising savings rate in China.On the other hand, Horioka and Wan (2007) use more recent data and find thatthe change in the dependency ratio does not explain the increasing savings rateadequately. This paper helps to solve the puzzle by emphasizing that individualsof prime age could save less because they have more brothers. The recent youngergeneration contributes to the high savings rate because they do not have siblings.The paper proceeds as follows. Section 2.2 introduces the background tohousehold financial markets, population policies, and the current savings rate inChina. Section 2.3 introduces the identification strategies and presents the esti-mation results. Section 2.4 explores the reason that having more brothers couldreduce the savings rate. Section 2.5 provides a robustness check for the iden-tification strategy. Section 2.6 shows how much of the savings rate puzzle canbe explained by the brothers effect. Section 2.7 concludes the paper. The Data11Appendix provides information on all the data used in this paper.2.2 Background2.2.1 Financial Markets and Household Borrowing ResourcesIt is a well-known fact that the corporate financial market in China is underdevel-oped; private entrepreneurs have to rely largely on financial resources from theirown networks such as family members or relatives (Ayyagari, Demirguc-Kunt,and Maksimovic 2010; Guariglia, Liu, and Song 2011; Chen, Ma, and Tang 2011;Song, Storesletten, and Zilibotti 2011; Allen, Qian, and Qian 2005). To date, littleattention has been paid to the household financial market, even though the degreeof development of this market is no better than that of the corporate financial mar-ket (Yao, Wang, Weagley, and Liao 2011; Coeurdacier, Guibaud, and Jin 2013).Despite the fact that the real interest rate on domestic bank deposits has oftenbeen negative (Gordon and Li 2003; Lardy 2012), by using the China HouseholdFinance Survey 2011, Gan (2012) suggests that the two main financial assets forhouseholds are bank deposits (58%), and cash holdings (18%). The rate of con-sumer loans issued by all financial institutions in China was nearly zero in 1997(Chamon and Prasad 2010). Although it reached 2.2 trillion RMB at the end of2005, mortgage loans amounted to about 80% of total loans.6Households can also encounter significant uncertainties. Medical reforms,pension reforms and rising income uncertainties cause households to save more6The other major loan categories were auto loans and large durable goods loans.12because of the precautionary motive (Chamon and Prasad 2010; Chamon, Liu,and Prasad 2010). How do households in China finance themselves when theyencounter negative shocks?I use two different data sets to investigate how Chinese households financethemselves in the current underdeveloped household financial markets. The firstdata set comes from the Chinese Household Income Project (CHIP, see Data Ap-pendix). The CHIP 2002 urban area survey asked, “If your household suddenlyencountered difficulty and needed 10,000 RMB immediately, where or to whomwould you turn first?”.7 I report the results in Figure 2.1. More than 60% ofthe individuals chose “family members and relatives,” while fewer than 3% ofthe individuals chose “financial institutions.” It is very clear that family mem-bers and relatives are a household’s primary borrowing source. The results alsosuggest that the potential transfer or quasi-credit amount available among familymembers could also be very large. Note that 10,000 RMB is approximately 1,600USD, which is more than half of the median household’s yearly income in the2002 CHIP data.The China Family Panel Study 2009 asked households if they actually bor-rowed money in 2008, if so the sources they borrowed from, and the reason thatthey borrowed. In total, 14% of the survey respondents had borrowed moneyin 2008. As was the case in the report using the 2002 CHIP data, “Relatives7There were nine answers to choose from: (1) family members and relatives, (2) friend, (3)other individuals, (4) work unit, (5) bank and credit union, (6) other financial institutions, (7) needno help, (8) anywhere I can borrow, (9) other. I aggregated (5) and (6) together and named thiscategory “financial institutions,” and I aggregated (3) (4) (8) and (9) together as “other.”13and friends” was overwhelmingly the dominant borrowing resource for house-holds. Conditional on borrowing money, 82.3% of the households borrowedfrom relatives and friends, while fewer than 20% of the borrowers borrowed frombanks.8 Note that this survey was conducted in Beijing, Shanghai and Guangdongprovinces, China’s most financially developed areas. In other less developed ar-eas, the proportion of households relying on family members could potentially beeven larger.The reasons for borrowing also varied from relatives to banks. In the CFPSdata, housing was the main reason for borrowing from financial institutions, whichaccounts for 85%. In contrast, there were a wide range of reasons for borrowingfrom relatives and friends that were evenly distributed among “education”(18%),“medical treatment”(20%), “housing”(22%), “living expense”(15%), and “other”(26%).It is worth noting that the housing loan market is quite developed in China,perhaps because of the government’s enforcement of housing reforms, which en-courages individuals to buy houses. As the primary reason for people borrowingmoney from banks is housing, and mortgages are not considered to be an unex-pected expense, relatives and friends become the only source of borrowing whena household encounters unexpected shocks.8Households had the following options to choose from in the survey: (1) banks (includingcredit unions), (2) relatives and/or friends, (3) loan from a private institution, and (4) other. Only2% of households had borrowed from (3) or (4).142.2.2 Facts: Household Savings Rate by Number of Brothersand SistersI use the China General Social Survey (CGSS) 2006 to construct household sav-ings rate data. The CGSS 2006 data contain the total income, basic living expendi-ture, medical expenditure, and education expenditure information of individuals’households. Savings are calculated as the household total disposable income mi-nus the sum of these three household expenditures. The savings rate is defined assavings divided by household total disposable income.9 Appendix Table 1 showsdetailed descriptive statistics of disposable incomes and expenditures. The aver-age savings rate in 2006 was 26 percent for urban residents, which is the mainsample used in this paper. It is only one percentage point higher than the savingsrate computed by using the data in China Statistical Year book for urban house-holds in 2006, 25 percent.10Figure 2.2 presents the age profile of the household savings rate by the numberof brothers and sisters of individuals. In the upper panel of Figure 2.2, I divide theindividuals into two groups: individuals with zero or one brother, and individualswith two or more brothers. The figure clearly shows that individuals with zero orone brother have a higher savings rate than individuals with two or more brothers,for all age groups. There is a strong negative correlation between the number of9I compute the income taxes based on the Individual Income Tax Law of the People’s Republicof China introduced in 2005. In an earlier version of this paper, I used income instead of disposableincome. The estimation results using disposable income are almost identical to the results using(non-tax-deducted) income.10When we compute the savings rate by using the data in China Statistical Year Book, thehousehold savings rate is defined as 1−Expenditure/Income, where expenditure is per capitahousehold living expenditure, and income is per capita household disposable income.15brothers and the household savings rate.By contrast, the savings rate is quite similar regardless of the number of sistersof individuals (the lower panel of Figure 2.2), in particular for individuals agedover 35 years. It is interesting to note, however, that the pattern of the savingsrate by number of sisters for young generations is close to that of the numberof brothers: having fewer sisters is associated with a higher savings rate. Asthe number of siblings declines because of the change in family planning policy,young households may use sisters as a substitute for brothers when there is ashortage of brothers. Sisters might also start to play the same role as brothers andaffect the household savings rate.Figure 2.3 repeats the same exercise by using individuals with no living par-ents to avoid the potential concerns of the siblings’ supporting-parents effect. Thefigure only presents savings rates for individuals aged over 40 years because thereare very few individual with no living parents below this age. The figure sug-gests that even for individuals with no living parents, the number of brothers stillhas a strong negative correlation with the household savings rate. For the num-ber of sisters, the correlation with the savings rate is not clear (the lower panel ofFigure 2.3).2.2.3 Changes in Demographics and China’s Savings RatePuzzleThe number of siblings of individuals has changed dramatically during recentdecades. Figure 2.4 presents the number of brothers and sisters of individuals by16individuals’ year of birth for the CGSS 2006 data. The figure shows that individ-uals born in the 1950s and 1960s have on average more than three siblings (with1.5 brothers and 1.5 sisters). In contrast, individuals born during the later 1970sand 1980s have fewer or even no siblings.While the average number of siblings has been decreasing in recent decades,the household savings rate has been increasing. Figure 2.4 shows the averagenumber of brothers of individuals from 1980 to 2005 as well as the trend in thehousehold savings rate. The household savings rate increased dramatically duringthis period. It presents one of the largest puzzles in China’s savings literature,which has attracted a lot of attention among researchers and policy makers: whyhas the savings rate in China increased substantially in recent decades. The figuresuggests that the decline in the average number of brothers may be one of thesolutions to this puzzle.The change in the number of siblings of individuals is induced by the change infamily planning policies in China. The population policies in China can be dividedinto three main stages: population expansion (1949–1972), voluntary birth control(1972–1978), and the One-Child Policy (1979–current).After the founding of the People’s Republic of China in 1949, policy makerspromoted population growth. The Chinese government introduced many policiesto encourage more births. For example, in 1952, the government published aregulation to restrict sterilization and abortions (Banerjee, Meng, Porzio, and Qian2013). The policy allowed a female to have an abortion only if the female was over35 or already had six or more children. Chairman Mao Zedong’s famous saying17“the more people, the stronger we are” is still a well-known phrase in China, evenfor the current generation.This large population growth was slowed by the second stage of family plan-ning policies implemented in 1972. During this stage, the government used theslogan “later, spaced, and few”: “later” for later marriage, “spaced” for spacedbirth, and “few” for fewer children. The policy emphasized birth spacing and didnot place a cap on the total number of children; however, the population controlpolicy at this stage was voluntary, and no punishment was meted out for violations.The decision to adopt birth control methods was left to the couples themselves.As a result of these population policies, China’s population almost doubled in just30 years, increasing from 540 million in 1949 to 960 million in 1978.The famous One-Child Policy stage represents the third stage of family plan-ning policies. This policy was introduced in 1978 and applied to the babies bornin 1979. In urban areas, each family was allowed only one child; however, in ruralareas, a second child was allowed if the first child was not male. Any additionalchildren resulted in large fines. Those families who violated the policy were re-quired to pay monetary penalties and could be denied bonuses at their workplaces.2.3 The Impact of the Number of Brothers onHouseholds’ Savings Rate2.3.1 IdentificationLet us first consider the following equation:18SavingRatei = αBroi +Xiγ + εi (2.1)The definition of the savings rate is given in Section 2.2.2. Broi is the numberof brothers of an individual. Xi is a set of individual characteristics and individ-ual’s household characteristics. α , the coefficient on Broi, is the parameter weare interested in. Broi could be correlated with unobserved family characteris-tics, such as parents’ economic conditions or their preferred number of children,which may be correlated with individual’s household savings. Thus, α cannot beconsistently estimated through equation 1.In order to identify the effect of brothers on the savings rate, I consider the fol-lowing case. If individuals’ parents cannot manipulate the gender of individuals’siblings, then given the number of siblings, the gender of siblings is only deter-mined by nature. The number of brothers is not correlated with any unobservedcharacteristics for a given number of siblings.If, given the number of siblings, having a brother instead of a sister is ran-domly assigned by nature, then the effect of having a brother instead of a sisteron the savings rate can be interpreted as a randomly assigned treatment. α can beconsistently estimated through equation 2.2. See Appendix A.2 for proof. Keepin mind that the interpretation of α is different in equation 2.2 from that in equa-tion 2.1, as α in equation 2.2 represents the effect on the savings rate of having abrother instead of a sister, for a given number of siblings.19SavingRatei = αBroi +δ (Sibi)+Xiγ + εi (2.2)The identification strategy compares the savings rate of individuals with dif-ferent numbers of brothers but with the same number of siblings. The upper panelof Figure 2.6 presents this variation. The figure suggests that for each siblinggroup, having more brothers is associated with a lower savings rate. As the sav-ings rate is defined as savings divided by income, one may be concerned that thenegative correlation between the number of brothers and the savings rate (condi-tional on the number of siblings) could be driven by the income correlation. Thelower panel of Figure 2.6 suggests that this is not a concern, as there is not a clearpattern of how income is correlated with the number of brothers given the numberof siblings.The assumption that, conditional on the number of siblings, the number ofbrothers is a random assignment requires that no predetermined family character-istics affect the assignment of the gender of the siblings (the only thing that candetermine the gender of the siblings is nature). Several papers in the “missingfemale” literature indicate that Chinese households have a son preference and thatthe sex ratio of newborns became distorted significantly following the introduc-tion of the One-Child Policy (1979) (Wei and Zhang 2011; Arnold and Liu 1986),because parents wanted to ensure that they had a son. The main reason for the sonpreference is that male children provide financial support to parents when parentsget old. Parents “chose” the gender of their children by practicing sex-selective20abortion or female infanticide, which was a practice sometimes found in ruralareas.I found that by restricting the sample to urban residents, and those born beforethe One-Child Policy (1979) and after World War II (1945), the evidence suggeststhat the gender of individuals’ siblings is exogenously assigned. In the rest of thepaper, I call this sample the restricted sample.There are several reasons that there is no gender distortion in the restrictedsample. First, the ultrasound technology required for sex-selective abortions wasonly introduced in the 1980s; households before the 1980s had no reliable methodfor performing sex-selective abortions. Second, female infanticide occurred mainlyin rural areas where households delivered babies at home. In urban areas, babieswere usually delivered in hospitals. In this case, it was unlikely that urban house-holds would risk criminal prosecution for son preference. Keep in mind that peo-ple born close to 1979 are unlikely to have siblings born after 1979 because ofthe One-Child Policy. Third, Chairman Mao largely enforced gender equality inChina before he passed away in 1976 (Li 2000). “Women hold half of the sky”is his famous slogan to enforce gender equality. In urban areas, females enjoyedas many job opportunities as males. The greater degree of gender equality ingeneral made parents in urban areas less likely to exhibit the same degree of sonpreference as before.Two sets of statistical tests examine whether the gender of children is exoge-nously assigned in the restricted sample. Table 2.1 reports the proportion of malesiblings given the number of siblings. The natural gender ratio is 106 males per21100 females (Jacobsen, Moller, and Mouritsen 1999). This implies that the natu-ral proportion of male siblings is 51.5%. If parents practice son preference, thisproportion would be significantly greater than 51.5%. The statistics computed inTable 2.1 show that the proportion of males is close to the natural level, regardlessof individuals’ number of siblings in the restricted sample.Table 2.2 provides a test of the random assignment of the number of brothersconditional on the number of siblings. In column 1, where the number of sib-lings is not controlled for, the number of brothers is significantly correlated withthe mother’s years of education. The Wald test suggests that all of the familycharacteristics are jointly significant. In contrast, once the number of siblings iscontrolled for in column 2, no parental characteristic is significantly correlatedwith the number of brothers, and the Wald test suggests that they are not jointlysignificant. I repeat the same test for the proportion of male siblings (column 3)and obtain similar results. The results in Table 2.2 provide strong evidence thatconditional on the number of siblings, the number of brothers is random amongurban residents born between 1945-1978.One may have concerns that parents might be practicing a son preference byadopting a stopping rule; that is to say, they keep having babies until they reach thedesired number of boys. This is also unlikely to happen in the restricted sample.An easy way to see whether parents adopted a stopping rule is to check the genderof their last child. If parents adopted a stopping rule, we are more likely to observethat the youngest child is a male. Recall that the natural proportion of males is51.5%. For urban residents born between 1945 and 1978, the proportion of males22as the youngest child of parents is 51.7% in the CGSS data and 50.4% in theCULS data (see Data appendix), and both are not significantly different from thenatural proportion of males.11One might also want to know the effect of sisters on a household’s savings rate.Ideally, we want to include the number of sisters in the regression to estimate theimpact of the number of sisters on the savings rate. However, such an estimateis not feasible because of the problem of collinearity (we cannot add both thenumber of brothers and sisters and siblings into one regression). As we control forthe number of siblings, α measures the difference between the effect of brothersand that of sisters. The coefficient on the number of siblings represents the effectof sisters with bias induced by endogeneity. See Appendix A.2 for the proof.Although the true effect of sisters could not be estimated, from Figure 2.2, itis more likely that sisters have no effect on a household’s savings rate. If this isthe case, the estimated relative effect of brothers compared with that of sisters, α ,also equals the absolute effect of brothers. If sisters behave like brothers, by alsoplaying a role with other siblings through risk sharing and supporting parents, theestimated brothers effect would be a lower bound of the absolute effect of brothers(see Appendix A.3).11The CGSS data do not provide the exact birth order of individuals’ siblings, because it onlylists the number of younger brothers and sisters, and older brothers and sisters. For this reason,I check the gender of an individual conditional on the individual being the youngest child in thefamily. The CULS 2001 data (see Data Appendix for details) provide the birth order of siblings. Irestrict the sample to urban residents born between 1945 and 1978. The sample size of the CULSdata is 5351.232.3.2 Results: the Impact of the Number of Brothers onHousehold Savings RateThe estimation results of equation 2.2 are presented in Table 2.3. Error terms areclustered at county level. Column 1 uses nonurban residents data. The rest of thecolumns use urban residents data because of the identification strategy discussedin Section 2.3.1. Both columns 1 and 2 control for the number of siblings, yearsof education, gender, age, age squared, household income, and marital status.In both urban and rural areas, we observe a negative effect of the number ofbrothers on the household savings rate. The coefficient on the number of brothersfor the sample of urban residents is −0.048 and statistically significant at the 1%level. This means that having one brother instead of one sister would, on aver-age, reduce the savings rate by 4.8 percentage points. Interestingly, the magnitudeof the brothers effect is larger in urban areas than in rural areas. The estimationresults in the first two columns may suggest that urban households rely more ontheir brothers than their rural counterparts. Rural areas usually have less devel-oped financial markets and experience higher risks associated with fluctuations inagricultural production. However, compared with urban households, rural house-holds can usually share risks with village members in addition to their familymembers and relatives. The larger brothers effect in urban areas may be becauseof the relative scarcity of sources of risk sharing besides family and relatives.Keep in mind that the coefficient on brothers in the rural sample may be biasedbecause of the potential female infanticide problem in rural areas.Column 3 adds a large set of demographic and characteristic controls that24could potentially affect a household’s savings rate: family size, parents-living-together dummy, Communist Party membership status, father’s and mother’s ed-ucation, and a send-down dummy.12 Chamon and Prasad (2010) indicate thatincreases in children’s education expenses and housing reform caused householdsto save more. For this reason, column 4 adds the number of children and chil-dren’s age group dummies in order to control for the potential education expenseeffect. Column 5 adds households’ housing characteristics: a dummy variable in-dicates whether each household owns the house, the mortgage value (if the houseis owned), and the value of the house that a household owns.13 Note that bycontrolling for these housing variables, I also control for the household asset ac-cumulation information because housing is the most important vehicle of house-hold asset accumulation ( Wei and Zhang 2011). Column 6 uses a set of siblingdummies instead of the number of siblings. This relaxes the specification of thefunctional form of δ (Sibi) in equation 2.2.In columns 2 to 6 of Table 2.3, the coefficient on the number of brothers is verystable at around−0.048. The fact that the coefficient on brothers is fairly constantalso provides evidence that the number of brothers is unlikely to be correlatedwith family characteristics once we have controlled for siblings. If the number of12Send-down was a program during the Chinese Cultural Revolution (1967–1977) in which thegovernment forced adolescents in urban areas to go to rural areas to do hard manual labor. Zhou(2013b) found that this event had a large impact on the send-down youths’ income and ability towithstand hard work.13The size of the mortgage is calculated as the percentage of the housing property that is stillunpaid multiplied by the housing value. Own housing is defined as a house owned by a familymember. Among urban individuals aged 28–60, 0.3% of individuals live with a working parentaged below 60; 5% of individuals live with married children.25brothers were correlated with any of the related individual and family characteris-tics used in the regressions, then the coefficient on the number of brothers shouldhave changed considerably in columns 2 to 6.One may worry about possible gender differences in the brothers effect. Malesmay be more likely than females to get help from their brothers. In the latter case,brother-in-laws of the female (i.e., the brothers of her husband) may be play-ing a more important role. I look for possible gender differences by introducingthe variable “Brothers of Female Respondents” into the regression (column 7 ofTable 2.3). This variable is generated by interacting Brothers with the Female re-spondent indicator dummy. The interaction variable Brothers of Female Respon-dents captures the brothers effect on females relative to males (the total brotherseffect for females is the sum of the coefficients on the main Brothers variable andBrothers of Female Respondents.) The coefficient of this variable is 0.02. How-ever, the standard error is relatively large, and the coefficient is not statisticallydifferent from zero. I conclude that the brothers effect on females is either equalto, or slightly smaller than, the brothers effect on males.I further restrict the analysis to both individuals and individuals’ parents withurban resident cards. This ensures that the individuals were born in urban ar-eas, where missing female problems are unlikely to occur. The sample becomesrelatively small; however, the coefficient of brothers is still around 0.05 and sta-tistically significant at the 1% level.The population policy switched from encouraging fertility to voluntary birthcontrol in 1972. The number of siblings of individuals declined gradually for26people born between 1972 and 1979 (Figure 2.4). In order to avoid the potentialeffect of this policy change, column 7 drops individuals born after 1971. Doingthis also allows us to estimate a relatively consistent sample of individuals with asimilar number of siblings. Column 8 drops individuals close to retirement age.The last column focuses solely on individuals who are between the ages of 35and 50. In these columns, brothers have a strong negative effect on the householdsavings rate.2.4 Why Brothers Reduce the Savings Rate: RiskSharing/Extending Borrowing Limits andSupporting ParentsIn this section, I propose that brothers reduce the savings rate through two chan-nels: (1) sharing their own risks and extending borrowing limits, and (2) sharingthe risks of their parents. A theoretical framework is provided in the online ap-pendix to support the arguments.A. Individual-Level Income Uncertainties and Health RisksI use the degree of uncertainty that individuals encounter to test the risk-sharing/extending-borrowing-limits effect. If brothers play roles in sharing risks/ex-tending borrowing limits, those individuals with larger uncertainties will have alarger brothers effect. Households with large uncertainties have a greater needto self-insure, so whether they have brothers (with whom they can share risks)will affect their savings rate considerably. By contrast, for those households with27fewer uncertainties, the presence of brothers might not matter so much; therefore,they are likely to have a small brothers effect. In equation 2.3, the size of α0 isexpected to be larger than the size of α1, where LargeUncertaintyi equals 1 ifindividual i encounters large uncertainties , and 0 otherwise; SmallUncertaintyiequals 1 if individual i encounters small uncertainties, and 0 otherwise.SavingRatei =α0Broi×LargeUncertaintyi+α1Broi×SmallUncertaintyi+δ (Sibi)+Xiγ+εi(2.3)I use individual income uncertainties and health risks as measures of the de-gree of uncertainty. The income uncertainty measures come from the questionsin the survey, “Is your basic monthly wage stable?” and “Is your monthly bonusstable?” A individual can choose among three possible responses: “very unsta-ble,” “a little unstable,” “stable.” The survey also asks, “How do you feel aboutthe condition of your health?” The answers are “very satisfied,” “satisfied,”“notsatisfied,” and “very unsatisfied.” Based on the answers, I evaluate the individ-ual’s health condition as “very good,” “good,” “bad,” or “very bad.” A bad healthcondition, unstable wage or bonus implies that individuals encounter greater un-certainty.The regression results are presented in columns 1 to 3 of Table 2.4. The resultsstrongly support the risk-sharing hypothesis: households with a large income un-certainty or health risks have a larger brothers effect, whereas households with asmall income uncertainty or health risks have a small brothers effect.28B. Regional Financial DevelopmentI test for the brothers effect of risk sharing/extending borrowing limits by ex-ploring the regional variations in financial development. If the incomplete state ofthe financial market makes household members rely on their brothers, we shouldobserve that households in financially developed regions have a smaller brotherseffect than households in regions where the financial market is underdeveloped.This is because formal credit market information is relatively widely availablein financially developed areas. In addition, households have more alternativesthrough which to borrow or lend funds in such areas. Therefore, households facea lower cost of accessing the financial market, and they can use the instrumentsavailable in financial markets to insure themselves. These households have lessneed to rely on brothers to borrow money or to share risks. In financially under-developed regions, the brothers effect should be large, because households haveno other alternative for acquiring insurance or borrowing money.I use the provincial-level insurance density and the number of foreign banksper capita in 2005 to measure regional financial development. See Appendix Table2 for the statistics of these two variables. Insurance density is provincial levelinsurance premiums per capita.14 Insurance density is used to capture overalldevelopment in the insurance market. The number of foreign banks per capita hasdirect and indirect effects on local financial development.1514The insurance premium is the sum of the private sector and public sector premia.15The number of consumer loans was almost zero in 1997 when the Chinese financial marketwas in its infancy. The direct effect of foreign banks on the financial market is reflected in the29SavingRatei = β0Broi +β1Broi×FianncialDevelopmenti +δ (Sibi)+Xiγ + εi(2.4)Equation 2.4 is estimated. Note that the city dummies are included in allregressions in this paper in order to control for the regional fixed effect. For thisreason, the provincial-level financial development indicators are not included inequation 2.4 because of collinearity with the city dummies.16 Regional financialdevelopment is usually correlated with regional GDP growth. In order to avoid thepotential concern that the brothers’ effect is driven by economic growth instead offinancial development, an interaction term of the number of brothers and regionalGDP growth is also included to control for the potential economic growth effect.The error term is clustered at province level to control for the random shockscorrelated within province.17The results in columns 4 and 5 of Table 2.4 show that the brothers effect isway that foreign banks offer more services and financial products to consumers in the market.The indirect effect is the spillover effect. Foreign banks bring to China experience and knowledgeaccumulated in well-developed markets abroad. Local Chinese banks can enjoy a spillover effectby observing the foreign banks’ ways of operating in the market and recruiting employees whohave accumulated expertise from foreign banks. We observe that the number of foreign banksin each province is determined primarily by government policies rather than by local consumers’demand for financial instruments. The Chinese government first allowed foreign banks to establishbranches in four cities in Guangdong and Fujian provinces. Only foreign currency businesses wereallowed to operate at that time. The next city to acquire permission was Shanghai in 1990. In 1992,the government granted permission to an additional seven cities located in Liaoning Shandong,Jiansu, ZheJiang, Fujian, and Guodong provinces, and Tianjin municipality. In 1996, foreignbanks were allowed to engage in business using Chinese currency in Shanghai. Later, this policywas extended to the provinces around Shanghai.16Cities dummies (in total 50 cities) absorb all the variation at the city and province (a lowerlevel of regional aggregation) level.17The significance level remains unchanged if I cluster the error term at county level.30indeed smaller in financially less developed regions. For example, in a provincewith the smallest insurance density (density=1), having an additional brother re-duces the savings rate by 9.1 percentage points (−0.093 + 0.002); in a provincewith the largest insurance density (density=32), having an additional brother re-duces the savings rate by only 2.9 percentage points (−0.093+32 × 0.002).C. Supporting ParentsIn Chinese culture, parents are supported primarily by their male children(Banerjee, Meng, and Qian 2010; Lee and Xiao 1998; Yu, Yu, and Mansfield1990; Ge, Yang, and Zhang 2012). By using China Health and Retirement Longi-tudinal Study (CHARLS) 2011 data, Table 2.5 shows that male children are morelikely to live with their parents and to make more regular and nonregular transfersto parents.Health care has become one of the major social issues in China in recent years.The rising private burden of health care is one of the main explanations of thehigh savings rate in China, in particular the high savings rate among the elderly(Chamon and Prasad 2010).18 A household with several brothers would needto save less for their parents’ risks—in particular, risks from medical expenditure,which is shared mainly among brothers. According to the CHARLS 2008 data, theconditional mean of transfers from male children to parents for medical expensesis almost twice the amount of that from female children: 2964 from male children18In 1978, out-of-pocket health spending was 20% of total health spending in China. In 2002,out-of-pocket health spending was 60% of total health spending (Yip and Hsiao 2008).31and only 1508 from female children.19If children do save for their parents, then once their parents have passed away,a household need no longer save for its parents. I utilize this idea of brothers toidentify the size of the brothers effect associated with supporting parents: I add(1) the number of a individual’s-parents-deceased term and (2) an interaction termbetween the number of (individual’s) brothers and the number of (individual’s)deceased parents. If parents are deceased, brothers will no longer be playing arole in sharing the risks of parents; therefore, the higher the number of parentswho have passed away, the smaller the size of the brothers effect. In Equation 2.5,we would expect δ2 to have the opposite sign to δ1.SavingRatei = δ1Broi +δ2Bro×ParentDeceasedi +δ3ParentDeceasedi+δ (Sibi)+Xiγ + εi(2.5)Table 2.6 reports the estimation results. First note that δ3 is significantly neg-ative. This suggests that households do save for their parents: once a parent haspassed away, a household saves less. Second, the brothers effect becomes smallerif the parents have passed away: δ2 has the opposite sign to δ1. When bothparents have passed away, having one brother reduces the savings rate by 0.028(0.026× 2− 0.8), and when no parents have passed away (the brother-parents-deceased interaction term also equals zero), the size of the brothers effect reachesits maximum value, |− .08|.19The sample is restricted to parents who were aged over 60 years in 2008.32Column 2 uses the One Parent Deceased and Two Parents Deceased dummiesinstead of the number of parents deceased variable. The estimation results revealthat the supporting parents effect is linear in the number of parents: the coeffi-cient on the two-parents-deceased interaction term (0.052) is almost twice thatof the coefficient on the one-parent-deceased interaction term (0.019). Similarly,linearity is observed between the parents-deceased dummies (the noninteractionterms).Note that two additional variables are also added to equation 2.5: the pres-ence of male children of an individual, and whether a parent (of an individual) isliving with that individual. The presence of male children reduces the householdsavings rate. This is consistent with the theory that male children carry out theduty of supporting parents. Bearing in mind that the financial support of the threegenerations is suggested here: individuals share the cost of supporting parentswith their male siblings. At the same time, individuals also expect their own malechildren to support them and therefore reduce their current savings rate. Second,the parents-living-together dummy has a negative sign. Households who live withtheir parents usually pay a large portion of their parents’ living expenses.20 Thus,if a individual lives with his/her parents, his/her household saves less. On theother hand, children who live with their parents are most likely to inherit the par-ents’ house after the parents have passed away. This leads to another important20According to the CLUS data, if a senior is living with his or her child, the senior only pays58% of his/her own living expenses; 38% of the living expenses are paid by the family memberswho live with him/her. However, seniors not living with a child pay 88% of their own expenses;the remainder is shared by those children not living with their parents and other family members.33interpretation of the negative coefficient of the parents-living-together dummy:children who live with parents will save less because they can expect a higherfuture income.Wei and Zhang (2011) suggests that parents tend to buy housing for their malechildren when they get married. One may worry that this may cause a potentialendogeneity problem because given that parents’ wealth is limited, individualswith fewer brothers (out of the total number of siblings) could anticipate a largerwealth inflow when they get married, and this may reduce the current savingsrate. However, among urban residents aged 28–60, 97% of the individuals havemarried (including 6% who are divorced or whose spouse has passed away). Onlyfewer than 3% of individuals were never married, and their average age is 38. Inaddition, a set of detailed housing information is included in all the regressions. Itis quite unlikely that Wei and Zhang (2011)’s suggestion could bias the results inthis paper.Other than purchasing housing for male children upon their marriage, it isquite rare for parents to provide transfers to their adult children in urban China.CHARLS 2011 data suggest that in urban China, only 0.2% of adult children agedabove 23 receive regular transfers from parents, and only 2.6% of children receivenonregular transfers from their parents.D. Brothers Effect in Different Income and Asset GroupsLow-income households usually have smaller emergency funds with which toprotect themselves from risks. In addition, it is common in China, and probably34in most other financially underdeveloped countries, for banks to lend money onlyto households with stable jobs and high income. This is consistent with the liter-ature that supports the idea that low-income households in developing countriesare usually borrowing constrained and have difficulty accessing the formal creditmarket (Morduch 1995). Households with low incomes or few assets may haveto rely mainly on their brothers; therefore, these households would have a largebrothers effect. 21I divide households into low- and high-income groups depending on whetherthe household income is below or above the median of the overall income distri-bution of the sample. The household income levels are used to approximate thedegree of demand for brothers because of extending borrowing limits or risk shar-ing. Column 1 of Table 2.7 reports the brothers effect for each income group. Thebrothers effect in the high-income group is calculated from the interaction term,high-income group dummy×brothers; the brother’s supporting parents effect inthe high-income group is calculated from a triple interaction term: high-incomegroup dummy×brothers×number of parents deceased. The results reveal that thebrothers effect is driven mainly by the low-income group. The coefficients ofboth the number of brothers and its interaction term with the number of parentsdeceased are much larger in the low-income groups compared with the previous21The 2002 CHIP data suggest that high-income households might have accumulated enoughemergency savings to insure themselves against a shock: 28% of the top-income tertile householdsstated that they had adequate savings in their bank to finance an emergency compared with only8% in the low-income tertile. These data relate to the CHIP 2002 question “If your householdsuddenly encountered difficulty and needed 10,000 RMB immediately, where or to whom wouldyou turn first?”35results (column 1 of Table 2.6). In contrast, both of these coefficients are not sta-tistically different from zero in the high-income group. I further restrict samples toindividuals with no living parents in column 2 to exclude the brothers’ supporting-parents effect. Although the standard errors of the coefficients are large becauseof the small sample size, the results are consistent with what we expected: thelow-income group has a much stronger effect of brothers compared with the high-income group.I further confirm the heterogeneity of the brothers’ effect by dividing house-holds by their assets instead of by their incomes (columns 3 and 4).22 Similar tocolumns 1 and 2, the brothers effect is larger in the low-asset group compared withthe high-asset group, which confirms the strong risk-sharing/extending-borrowing-limits effect of brothers.22Besides income, assets are also an indicator of the demand for brothers, for potentially tworeasons. First, a household with sufficient assets would be less likely to borrow money from broth-ers because it can finance consumption using its own emergency funds following shocks. Second,assets, especially housing assets, improve a household’s ability to access the formal financial mar-ket because assets could act as collateral when borrowing money from banks. Most bank loansin China require collateral, and the only acceptable collateral for most banks is buildings or land(Gregory & Tenev 2001; Ayyagari et.al., Cousin 2006). Only 4% of commercial loans are securedby movable assets in China. The value of housing assets is generated by subtracting mortgage bal-ances (unpaid amount) from the housing values owned by a household. Ideally, total assets valueis a better indicator than housing assets value. As CGSS does not provide total asset data, I usehousing value instead. This caveat is unlikely to cause problems because the rank of householdtotal assets and the rank of housing assets are highly correlated. Using the 2002 CHIP data, Igenerate the three-level (low, medium, high) housing value asset rank and total asset rank. Thesetwo ranks are highly correlated: the correlation coefficient is 0.77 and significant at the 1% level.362.5 Robustness Check2.5.1 Son PreferenceThe identification strategy in this paper relies on parents with a son preference notacting on it by selecting the gender of their children. In this section, I test to whatextent, if any, does the subjective preference of son bias our results by controllingan indicator of son preference.The indicator comes from the question in the Family Survey of CGSS 2006:“If you are only allowed to have one child, do you prefer a boy or a girl.” A re-spondent can choose “Boy,” “Girl,” or “Both boy and girl are the same for me.”(The Family Survey of CGSS 2006 is a subset of the China General Social Sur-vey.) The proportion of individual choices in each category is 20%, 12% and67%, respectively, in the restricted sample. Based on the answer to this question,I generated a son-preference indicator and a daughter-preference indicator, wherethe indicator equals one if a individual chooses a specific gender. The genderpreference question is only asked in the Family Survey of CGSS 2006, which is arelatively small sample. One limitation of this indicator is that the son preferenceis of individuals, not of individual’s parents. However, the literature has shownthat the gender preference is largely transmitted from parents to children within afamily (Escriche et al. 2004).Table 2.8 reports the estimation results for the Family Survey sample. Column1 does not control for the gender preferences, while column 2 controls for genderpreferences. The coefficient of brothers in column 1 is very close to the coef-37ficient in column 2. The coefficient of both gender preference indicators is notstatistically significant (the top panel of column 2). Interestingly, the estimatedcoefficient of son preference is the same as the coefficient of daughter preference.In the rest of the table, I repeat the same strategy in the estimation of the differentchannels of the brothers effect. The estimation results are almost identical with orwithout controlling for son preference and daughter preference. These estimationresults suggest that the brothers effect on the savings rate is unlikely to be affectedby the son preference.2.5.2 Brothers Effect of Individuals Born After the One-ChildPolicyIn the main sample, I use individuals who were born before the One-Child Policyto identify the brothers effect. In this section, I provide a robustness check for themain sample to show that for people born after the One-Child Policy, there is stilla strong brothers effect.Because of the missing female problem that started to prevail after the One-Child Policy, the control function approach is no longer valid for individuals bornafter the One-Child Policy. Instead, I use the One-Child Policy fines for unau-thorized births in urban areas as an instrument for the number of brothers that ahousehold has in urban areas. The One-Child Policy fines are set by local gov-ernment. Ebenstein (2010) shows that the One-Child Policy fine had a significantimpact on the gender ratio and fertility decisions. For this reason, we can expecta strong first-stage estimation.38Fines are set as a percentage of an individual’s annual income for a certainnumber of years. Following Ebenstein (2010), I calculated the present value oftotal fines to obtain a single value, which represents the percentage of a parent’sannual income needed to pay fully for an additional child.23 The typical finerequires each parent to pay 10% of his/her annual income for 14 years, which, ac-cording to my calculations, is equivalent to 123% of the combined annual incomeof each parent.The provincial level One-Child Policy fines have considerable regional andtemporal differences. Provincial fixed effects and time fixed effects are includedwhen using the instrumental variable. The provincial fixed effects control forprovincial level time invariant factors such as provincial initial conditions. Theyears fixed effects control for the factors that uniformly affected all provinces ineach year. This IV strategy identifies a local average treatment effect in the sensethat it identifies the effect of households’ parents who would have one more childif fines or bonuses were low, but not otherwise. For this group, the impact of thenumber of brothers might be large because they have strong preference for havingchildren and have strong family ties.In the IV estimation, the sample includes individuals born between 1979 and1984 (22 to 27 years old in the data) and who are urban residents. We should bearin mind that the IV estimation results might not be very precise because of thesmall sample size (355 observations).23A 2% annual discount rate is applied to calculate the present value of fines. The One-ChildPolicy fine data are collected in Scharping (2003).39Column 1 of Table 2.9 presents the results of the first stage. Fines significantlyreduce the number of brothers that a households has. Changing fines from zero to100% of annual income reduces, on average, the number of brothers by 0.779, andthis is statistically significant at the 1% level. The second-stage estimation resultsare reported in the remaining columns. The Anderson–Rubin weak IV robust 95%confidence intervals for key variables are provided in square brackets.The IV estimation results are consistent with the findings by using the sam-ple data from before the One-Child Policy: having an additional brother reducessavings rate (column 2). The results are also consistent with the previous find-ings that low-income or wage-unstable individuals have a larger brothers effectthan high-income or wage-stable individuals, which implies that brothers play arole in sharing risks and extending the borrowing limit.24 The estimation results,however, could not detect the supporting-parents effect (column 3), potentiallybecause there are too few households with deceased parents given their young age(22–28).2.6 How the Decline in the Number of Brothers inHouseholds Could Explain the Savings RatePuzzleData from the China Statistical Year Book indicate that the average savings ratein urban areas increased from 16% in 1990 to 24% in 2005, where the average24Because of the small sample size, I divided households into two income groups (low and high)instead of three. For the same reason, I also group “wage very unstable” and “wage unstable” intoone group.40savings rate is defined as “averagesaving/averagedisposable income.” In thissection, I calculate, holding everything else constant, to what extent the change inthe number of brothers can explain the change in the savings rate. I also assumethat sisters have no effect on the savings rate.From the estimation results of the previous sections, we know that the brotherseffect depends on the number of living parents and their average incomes. Thus,I divide households into six groups: two income groups times three age groups.The two income groups are low and high; they are equally divided over the incomedistribution. The three age groups are ages 22–39, 40–49, and 50–60. The changesin the savings rate in each group depend on the average income, the number ofparents deceased, the number of brothers and the estimated brothers effect in thatgroup. The total change in average savings is the sum of the change in savings ineach group weighted by each group’s density. Mathematically, it can be describedin the following way.4averagesaving =∑A∑IIncI,A( ̂broIncI,A +DPI,A× ̂broDPI,A)4broI,A f (I,A)(2.6)A denotes the age group, and I denotes the income group. Inc is the aver-age income. DP is the number of parents deceased. ∇bro denotes the changein the number of brothers between 1990 and 2005. f (·) is the density of eachgroup. ̂broInc is the estimated brothers effect. ̂broDP is the estimated brother-supporting-parents effect. The statistics of these variables based on the CGSS41data are presented in Appendix Table 3. Note that only the statistics of the low-income groups are presented, because the savings rate of the high-income groupis not affected by the number of brothers. The marriage rate is also used in thecalculation in order to take into account the change in the number of brothers ofboth the husband and wife of a household.The simple calculation suggests that declines in the number of brothers ofhouseholds explained 34.7% of the increase in the aggregate savings rate from1990 to 2005 in urban China. Be mindful that the estimated explained increasedwould be larger if sisters also behaved like brothers and affected the householdsavings rate.2.7 Conclusion and DiscussionIn this paper, I found that having one more brother of a individual reduces the indi-vidual’s household savings rate by at least five percentage points in urban China.Brothers reduce the savings rate because they share the risks/extend borrowinglimits, and share the cost of supporting their parents. The change in the number ofbrothers of households explained 34.7% of the increase in the household savingsrate.It is interesting to note that although China is the world’s second largest econ-omy, household financial markets are still underdeveloped even in urban areas.The Chinese government might consider developing household financial marketsas soon as possible. The baby boom generation can rely on their siblings to fi-nance themselves. They face few hurdles while household financial markets are42underdeveloped. However, the current and future younger generations have fewsiblings because of the One-Child Policy. They lack a family-based safety netand they carry the huge burden of supporting their parents. Developing householdfinancial markets is a necessary and urgent task.This paper is one of the first papers to estimate the number of siblings effecton the household savings rate. The results may not be limited to China only. Itwould be interesting to see whether other countries where households share riskswith their siblings and children to support their parents financially, such as Indiaand other East Asian countries, have a similar sibling effect on the savings rate.In addition, if these countries have cultures similar to that of China, that is to say,male siblings have stronger family ties compared with female siblings, we mayalso observe gender differences in the siblings effect on the savings rate.43Figure 2.1: Sources for Borrowing Money in Urban China: Self-Reports ofBorrowing Resource if One Encounters a Negative Shock (Percentageof Respondents)Note: The above results are calculated by the author based on a question in the Chinese House-hold Income Project 2002 Urban Sample:“If your household encountered an abrupt difficulty andneeded 10,000 RMB immediately, who (where) would you turn to first?” Sample size: 6779.44Figure 2.2: Age Profile Household Savings Rate by Number of Brothers andSistersData source: China General Social Survey 2006. Total sample size: 6886. Sample size of individ-uals with more than one brother in each age group (from young to old): 111 299 571 685 416 515469; zero or one brother: 613 628 640 573 406 481 482; more than one sister: 161 333 504 618411 477 418; zero or one sister: 563 594 707 640 408 519 53345Figure 2.3: Age Profile Household Savings Rate by Number of Brothers andSisters - Households with No Living ParentsData source: China General Social Survey 2006. Sample size: 1732. Sample size of individualswith more than one brother in each age group (from young to old): 152 145 266 295 ; zero or onebrother: 117 131 250 376, more than one sister: 147 143 244 286; zero or one sister: 122 133 27238546Figure 2.4: Number of Brothers and Sisters by Individuals’ Birth YearData source: China General Social Survey 2006. Sample of urban area residents are used. Samplesize: 323547Figure 2.5: Number of Brothers and Household Savings Rate in Urban Ar-easNote: The number of siblings are restricted to individuals aged 20-60. Death rates are used inorder to compute the number of siblings in early years. Saving rate is defined as 1-living ex-penditure/disposable income. Saving rate and death rate data source: China Statistical Yearbook.Siblings data source: China General Social Survey 2006.48Figure 2.6: Source of Variation: Average Household Savings Rate by Num-ber of Brothers for a Given Number of SiblingsNote: China General Social Survey 2006 is used. Sample is restricted to urban area residents bornbetween 1945 to 1978.49Table 2.1: Fraction of Male Siblings by Total Number of SiblingsNumber of Siblings Obs Fraction of Male 95% Conf. Interval1 572 0.52 [ 0.50, 0.55]2 846 0.52 [ 0.50, 0.53]3 756 0.49 [ 0.48, 0.51]4 or more 1085 0.48 [ 0.47, 0.49]Note: China General Social Survey 2006 is used. Sample is restricted to urban area residents bornbetween 1945 to 1978.50Table 2.2: Test of Random Assignment of the Number of Brothers Condi-tional on the Number of SiblingsDependent VariableBrothers Brothers FractionSiblings 0.485∗∗∗ -.009∗∗(0.012) (0.005)Mother Education -.042∗∗∗ -.008 -.003(0.009) (0.006) (0.003)Father Education -.010 0.008 0.004(0.009) (0.006) (0.003)Mother Communist Party 0.004 -.031 0.016(0.068) (0.052) (0.022)Father Communist Party 0.059 0.039 -.024(0.133) (0.098) (0.045)Mother Company Type -.060 -.044 -.029(0.06) (0.043) (0.02)Father Company Type 0.056 -.014 -.007(0.047) (0.034) (0.014)Mother Occupation Skill Level -.013 -.012 -.006(0.025) (0.02) (0.009)Father Occupation Skill Level -.013 0.002 -.0002(0.02) (0.014) (0.006)Mother Occupation Dummies Yes Yes YesFather Occupation Dummies Yes Yes YesObs. 2608 2608 2383Wald statistics 9.32∗∗∗ 1.39 1.45Note: China General Social Survey 2006 is used. Sample is restricted to urban area residentsborn between 1945 to 1978. The Wald test examines the joint significance of all the regressorsin column 1. In column 2 and 3, number of siblings is not included in the Wald test; all otherregressors are included. Standard errors are clustered at county level. *** p<0.01, ** p<0.05, *p<0.151Table 2.3: The Impact of Number of Brothers on Household Savings RateDependent Variable: Savings RateBorn 1946-1978 Born before 1972 Born after 1955 Born 1956-1971(Age 28-60) (Age 35-60) (Age 28-50) (Age 35-50)Rural (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)Brothers -.033∗ -.048∗∗∗ -.048∗∗∗ -.046∗∗∗ -.046∗∗∗ -.046∗∗∗ -.056∗∗ -.057∗∗∗ -.053∗∗∗ -.066∗∗∗ -.083∗∗∗(0.018) (0.017) (0.016) (0.016) (0.016) (0.016) (0.022) (0.02) (0.019) (0.019) (0.024)Siblings 0.017 0.011 0.014 0.016 0.017(0.012) (0.01) (0.01) (0.01) (0.01)Brothers of Female Respondents 0.021(0.024)Years of Education -.004 0.009∗ 0.01∗∗ 0.012∗∗ 0.012∗∗ 0.012∗∗ 0.013∗∗ 0.004 0.015∗∗ 0.012∗∗ 0.018∗∗(0.006) (0.005) (0.005) (0.005) (0.005) (0.005) (0.005) (0.005) (0.007) (0.005) (0.008)Household Income 1.935∗∗∗ 0.328∗∗∗ 0.335∗∗∗ 0.33∗∗∗ 0.333∗∗∗ 0.328∗∗∗ 0.328∗∗∗ 0.27∗∗∗ 0.315∗∗∗ 0.375∗∗∗ 0.354∗∗∗(0.371) (0.093) (0.09) (0.091) (0.092) (0.091) (0.092) (0.088) (0.093) (0.12) (0.123)Basic Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes YesDetailed Backgrounds Yes Yes Yes Yes Yes Yes Yes Yes YesChildren Yes Yes Yes Yes Yes Yes Yes YesHousing Yes Yes Yes Yes Yes Yes YesSibling Dummies Yes Yes Yes Yes Yes YesObs. 2364 2580 2539 2539 2502 2502 2502 1730 2067 1816 1381R2 0.179 0.175 0.18 0.227 0.225 0.229 0.23 0.243 0.209 0.27 0.243Note: China General Social Survey 2006 is used. Sample is restricted to individuals born between 1945 to 1978. Column 1 uses non-urban residents data. Column 2 to column 11 use urban area residents data. Standard errors are clustered at county level. *** p<0.01, **p<0.05, * p<0.1.Other variables included:1. Basic Controls: female, age, age squared, marital status, years of education, household income and city dummies.2. Detailed Backgrounds: mother education, father education, number of people in households, communist party membership andsend-down dummy.3. Children Information: number of children, children age group dummies: 0-6, 6-18 or 18 and above.524. Housing Information: housing dummy, value of mortgage and value of housing.53Table 2.4: Brother’s Sharing Risks / Extending Borrowing Limits EffectDependent Variable: Savings Rate(1) (2) (3) (4) (5)Stability of IncomeBrothers ×Wage Very Unstable -.137∗∗∗(0.046)Brothers ×Wage Unstable -.049(0.032)Brothers ×Wage Stable -.047(0.032)Brothers × Bonus Very Unstable -.114∗(0.065)Brothers × Bonus Unstable -.068∗(0.037)Brothers × Bonus Stable -.039(0.034)Personal HealthBrothers × Health Very Poor -.158∗∗(0.065)Brothers × Health Poor -.127∗∗∗(0.031)Brothers × Health Normal -.067∗∗∗(0.024)Brothers × Health Very Good -.093∗∗(0.037)Regional DevelopmentBrothers -.093∗∗∗ -.088∗∗∗(0.026) (0.026)Brothers × Insurance Density 0.002∗∗∗(0.0008)Brothers × # of Foreign Bank per Capita 0.032∗∗∗(0.008)Obs. 1407 1013 2499 2499 2499R2 0.337 0.314 0.254 0.248 0.248Note: Sample is restricted to urban area residents born between 1945 to 1978. Standard errors areclustered at county level. *** p<0.01, ** p<0.05, * p<0.1.Other variables included:1. Basic Controls: siblings, female, age, age squared, marital status, years of education,household income and city dummies.2. Detailed Backgrounds: mother education, father education, number of people in house-54holds, communist party membership and send-down dummy.3. Children Information: number of children, children age group dummies: 0-6, 6-18 or 18and above.4. Housing Information: housing dummy, value of mortgage and value of housing.5. Column 4 and 5 also include number of brothers × provincial level growth regional prod-uct.55Table 2.5: Gender Differences in Supporting ParentsMale FemaleLiving with ParentsProportion 23.9 4.5(0.5) (0.3)Regular Transfers to ParentsProportion 5.0 2.6(0.6) (0.4)Conditional Mean 7322 2833(1903) (798)Non-regular Transfers to ParentsProportion 32.2 31.8(0.6) (0.6)Conditional Mean 2248 1146(160) (78)Note: Authors’ tabulation based on the China Health and Retirement Longitudinal Study 2011.Sample is restricted to parents who are above 60 years old, and their children are above 23 yearsold. There are 2410 individual level observations. Proportion represents percentage of individuals.Standard errors in parentheses.56Table 2.6: The Impact of Number of Brothers on Household Savings Rates- the Effect of Supporting ParentsDependent Variable: Savings RateBrother -.080∗∗∗ -.078∗∗∗(0.026) (0.025)Brother × # of Parents Deceased 0.026∗(0.014)Brother × One Parent Deceased 0.019(0.029)Brother × Two Parent Deceased 0.052∗(0.029)# of Parents Deceased -.081∗∗(0.034)One Parent Deceased -.085∗(0.049)Two Parents Deceased -.160∗∗(0.07)Male Children Presence -.047∗∗ -.048∗∗(0.024) (0.024)Parents Live Together -.106∗∗∗ -.104∗∗∗(0.038) (0.039)Obs. 2500 2500R2 0.247 0.247Note: Sample is restricted to urban area residents born between 1945 to 1978. Standard errors areclustered at county level. *** p<0.01, ** p<0.05, * p<0.1.Other variables included:1. Basic Controls: siblings, female, age, age squared, marital status, years of education,household income and city dummies.2. Detailed Backgrounds: mother education, father education, number of people in house-holds, communist party membership and send-down dummy.3. Children Information: number of children, children age group dummies: 0-6, 6-18 or 18and above.4. Housing Information: housing dummy, value of mortgage and value of housing.57Table 2.7: The Brother Effect in Different Income Groups and Asset GroupsDependent Variable: Savings RateAll No Living Parents All No Living Parents(1) (2) (3) (4)Low IncomeBrothers -.122∗∗∗ -.054(0.027) (0.034)Brothers× # of Parents Deceased 0.039∗∗(0.017)High IncomeBrothers 0.014 0.007(0.016) (0.038)Brothers× # of Parents Deceased 0.016(0.014)Low AssetBrothers -.090∗∗∗ -.053(0.026) (0.043)Brothers× # of Parents Deceased 0.028(0.018)High AssetBrothers -.056∗∗∗ -.047(0.022) (0.033)Brothers× # of Parents Deceased 0.019(0.016)Obs. 2491 663 2312 615R2 0.313 0.239 0.238 0.21Note: Sample is restricted to urban area residents born between 1945 to 1978. Standard errors areclustered at county level. *** p<0.01, ** p<0.05, * p<0.1. The brothers effect in high incomegroup is calculated from the interaction term, high income group dummy×brothers. The brother’ssupporting parents effect in high income group is calculated from a triple interaction term: highincome group dummy×brothers×number of parents deceased.Other variables included:1. Basic Controls: siblings, female, age, age squared, marital status, years of education,household income and city dummies.2. Detailed Backgrounds: mother education, father education, number of people in house-holds, communist party membership and send-down dummy.3. Children Information: number of children, children age group dummies: 0-6, 6-18 or 18and above.4. Housing Information: housing dummy, value of mortgage and value of housing.585. Presence of male children.6. Column 1 and 3 also controls number of parents deceased and parents living togetherdummy.59Table 2.8: Robustness Check: Son PreferenceDependent Variable: Savings RateWithout Son, With Son,Daughter Preference Daughter Preference(1) (2)Basic ResultsBrothers -.089∗∗∗ -.087∗∗∗(0.027) (0.026)Son Preference 0.057(0.039)Girl Preference 0.057(0.054)Obs. 927 927Supporting ParentsBrothers -.109∗∗∗ -.108∗∗∗(0.04) (0.039)Brothers × # of Parents Deceased 0.027 0.028∗(0.017) (0.017)Obs. 927 927Individual Wage RisksBrothers ×Wage Unstable -.078∗ -.078∗(0.044) (0.044)Brothers ×Wage Stable -.010 -.010(0.038) (0.038)Obs. 511 511Regional Financial DevelopmentBrothers -.134∗∗∗ -.133∗∗∗(0.045) (0.044)Brothers × # of Foreign Bank per Capita 0.022∗ 0.021∗(0.013) (0.013)Obs. 927 927Income HeterogeneityBrothers × Low Income Dummy -.236∗∗∗ -.236∗∗∗(0.051) (0.051)Brothers × High Income Dummy -.029 -.029(0.037) (0.037)Obs. 927 927Note: The Family Survey of the China General Social Survey 2006 is used. Sample is restrictedto urban area residents born between 1945 to 1978. Wage Unstable equals one if a respondentcharacterized his/her wage is very unstable or unstable; 0 otherwise. Wage stable equals one ifa respondent characterized his/her wage is stable. Standard errors are clustered at county level.Standard errors are clustered at county level. *** p<0.01, ** p<0.05, * p<0.1.Other variables included:601. Basic Controls: siblings, female, age, age squared, marital status, years of education,household income and city dummies.2. Detailed Backgrounds: mother education, father education, number of people in house-holds, communist party membership and send-down dummy.3. Children Information: number of children, children age group dummies: 0-6, 6-18 or 18and above.4. Housing Information: housing dummy, value of mortgage and value of housing.5. Number of parents deceased, parents living together dummy and presence of male children.61Table 2.9: IV Estimation Results for Individuals Born after the One ChildPolicy1st Stage 2nd StageBrothers (1) Savings Rate (2-6)(1) (2) (3) (4) (5) (6)Fines -0.779∗∗∗(0.269)Brothers -0.373∗∗ -0.359∗∗(0.18) (0.177)[-1.090, -0.035] [-1.92, 0.045]Brothers × # of Parents Deceased -0.261(0.676)Brothers × Low Income Dummy -0.510∗∗∗(0.181)[-1.745, -0.110]Brothers × High Income Dummy -0.253(0.175)[-1.56, 0.23]Brothers ×Wage Unstable -0.484∗∗(0.219)[-1.785, 0.054]Brothers ×Wage Stable -0.377∗(0.446)[-1.92, 0.082]Brothers × Bonus Unstable -0.291(0.215)Brothers × Bonus Stable -0.106(0.166)Birth Year & Province Fixed Effect Yes Yes Yes Yes Yes YesObs. 355 355 355 355 300 2361st Stage F Statistics 10.217 0.016 5.579 4.322 6.473Note: Sample is restricted to urban area residents born between 1979-1984. Wage Unstable equalsone if a respondent characterized his/her wage is very unstable or unstable; 0 otherwise. Wagestable equals one if a respondent characterized his/her wage is stable. Standard errors are clusteredat county level. *** p<0.01, ** p<0.05, * p<0.1. Other variables included:1. Basic Controls: siblings, female, age, age squared, marital status, years of education,household income and city dummies.2. Detailed Backgrounds: mother education, father education, number of people in house-holds, communist party membership and send-down dummy.3. Children Information: number of children, children age group dummies: 0-6, 6-18 or 18and above.624. Housing Information: housing dummy, value of mortgage and value of housing.5. Number of parents deceased, parents living together dummy and presence of male children.63Chapter 3How Does a Hard Manual LaborExperience during Youth AffectLater Life? The Long-term Impactof the Send-down Program duringthe Chinese Cultural Revolution3.1 IntroductionThe adolescent and teenage years are important stages in the human lifespan. Dur-ing these years, lifelong habits and personality traits are easily shaped or changedby the outside environment. A good environment fosters positive thinking, mo-64tivating individuals to perform to the best of their abilities, and to generate highreturns for themselves (Borghans et al. 2008; Cunha and Heckman 2007). So far,most literature has focused on how a positive intervention could affect individualachievement later in life later(Rodrguez-Planas 2012; Schochet and McConnell2008; Heckman and Kautz 2014). However, what if people undergo hard manual-labor experiences during adolescence?During the 1960s and 1970s, under Mao Zedong’s leadership, China under-went the famous Cultural Revolution. The government forced more than 16 mil-lion adolescents from urban areas to move to rural areas to carry out agriculturalfield work. This event is known as the “send-down movement.1 The sent-downyouths were forced to engage in hard manual labor in the fields for as long as12 hours a day, 7 days a week (Zhou and Hou 1999). Although the sent-downswere allowed to return to urban areas after the Cultural Revolution, they werescarred by their difficult experience—an experience not shared by non-sent-downurban youths. This study investigates the long-term impact of such a challengingmanual-labor experience on these youths when they had reached the ages of 40 to55.One might expect that such experiences, which adolescents endured for ap-proximately five years, would have some effect on life outcomes, whether positiveor negative. Surprisingly, if we compare the income of those sent down with thosenot sent down, conditional on individuals having graduated from high school (in-1High school in this paper refers to a school comprising grades 7 through 12; it includes bothjunior and senior high schools.65cluding junior high) during the Cultural Revolution, we see virtually no incomedifference between the two. This paper finds that a very important step has beenneglected by previous literature, one that reveals a significant effect of the send-down experience. This key step is the upgrading of education after the CulturalRevolution.For several years during the Cultural Revolution, many senior high schoolsand universities stopped admitting new students due to the governments educa-tion policy (Meng and Gregory 2002; Giles, Park, and Wang 2008; Han, Suen,and Zhang 2011).2 This resulted in education interruption for teenagers gradu-ating from junior and senior high schools during the Cultural Revolution. Han,Suen, and Zhang (2011) find that, after the Cultural Revolution, many of theseindividuals went back to school to reinvest in their human capital in order to com-pensate for their interrupted schooling. In this paper, I refer to this “re-schooling”movement as education upgrading. I find that, among individuals who graduatedfrom high school during the Cultural Revolution, the sent-down males were morelikely to have upgraded their education compared to the non-sent-down males. Itappears that the hard manual labor experience has a strong positive effect on in-come. However, this effect is conditional on education upgrading. The sent-downmales who upgraded their education earn a 10% higher income than non-sent-down males who also upgraded their education. However, for those who did notupgrade their education, the send-down experience negatively affects income be-2In the first 2 years of the Cultural Revolution, not only senior high schools and universitiesbut also elementary and junior high schools were shut down.66cause of the loss of years of urban work experience.Furthermore, consistent with the finding in income, conditional on educationupgrading, the sent-down males are more likely to have computers at home thannon-sent-down males are. The social economic environment changed dramati-cally after the Cultural Revolution, with computers being one of the new hightechnologies favored by the rapid economic growth. Knowing how to operate acomputer could have positive affect on ones income; however, it is not easy for in-dividuals aged 40 or 50 to learn such a new technology. The education-upgradedsent-down males may have acquired skills during the send-down that helped themadjust to changes in their environment. The results are robust against the exclu-sion of individuals living with their children. Falsification test shows that bothsent-down males and non-sent-down males are equally likely to own other majorhome appliances, such as color TVs, air conditioners, or video cameras.It is natural to believe that agricultural field work in rural areas should haveno direct relation to academic education or urban work experience. One interpre-tation of the results is that the experience of years of forced hardship during theiryouth has helped the sent-down males develop an endurance or greater resistanceto future adversity. Many documents have reported that the difficult send-downexperience could have motivated these youths to study and work harder later in life(Yang 1992; Wang 2006; Liu 2012; Tang 2012). These documents reported thatsent-down youths learned that life is tough and, further, that hard manual-laborexperience made them stronger, helping them gain the ability to face adversity.Appendix B provides a conceptual framework to explain the empirical finding67that sent-down males are more likely to upgrade education, and-conditional uponeducation upgrading-why they earn higher incomes.Almost every urban family had at least one child sent down (Bernstein 1977).The accumulated number of send-downs during the 1960s and 1970s was equiva-lent to 10.5% of the total non-farming population in 1979 (Pan 2002). During theCultural Revolution, local governments had a quota of send-downs to fill everyyear (Pan 2002; Bernstein 1977; Singer 1971). The quota varied largely by year.Local government determined the send-down selection process based on the quotaand the number of age eligible youths (junior or senior high school graduates intheir graduation year). If the quota was high, all age eligible youths would be sentdown. If the quota was low, the local government would allow families who hadalready sent away a proportion of their children to keep their current age eligiblechild.3It has been well documented that parental social status or political capital didnot prevent youth from privileged classes from being sent down, as Mao wasenforcing social equality in China (Bernstein 1977; Singer 1971; Unger 1980;Zhou and Hou 1999; Xie et al. 2008). Some previous studies, however, havesuggested that the send-down program might have discriminated against a groupof children whose parents had college-level education.4 In order to avoid potentialbias induced by the selection of the policy, this study focuses only on childrenwhose parents had less than or equal to 12 years of education. (Note that the3China had a huge baby boom after the war; the average number of children per family duringthe 1960s and 1970s was four (Zhou 2013a; Banerjee et al. 2010).4Mao thought that high education was a main source of social inequality (Pan 2002).68results are robust when I restrict samples to those individuals whose parents hadonly equal to or less than 9 years of education.)In estimating the send-down effect conditional on education upgrading, par-ents education and job information are used to approximate individuals abilitywhich is unobserved to econometricians. In addition, I estimate a differential ef-fect by adopting a difference-in-differences type of specification, which is servedto control for the general difference between the send-downs and the non-send-downs (i.e., the difference between the two groups regardless of education-upgradingstatus), as well as the difference between the education-upgraded group and thenon-education-upgraded group (regardless of the send-down experience).One may still have concerns that the above two strategies may not fully solvethe endogeneity problem in education upgrading choice. In the robustness check, Ifurther use the relative number of full-time teachers during the Cultural Revolution(which was determined by the education policies during the Cultural Revolution)as IVs to instrument education upgrading choice. The IV estimation results areconsistent with the findings in the OLS estimations.The positive effects of the send-down experience on education and incomesfound in this study are robust and statistically significant even when I: (1) controlfor family connections; (2) drop all the send-down samples of those able to returnto urban areas before the end of the Cultural Revolution; (3) eliminate individualswhose parents had capitalist tendencies (worked in private firms or owned privatefirms); and (4) focus on individuals from disadvantaged family backgrounds. Therobust and significant results suggest that the findings in this paper are unlikely to69be altered by the youths’ family backgrounds.Li et al. (2010) suggest that parents were allowed to choose which child tosend away, and their empirical results suggest that parents chose to send away thechild with the lower ability. If this were the case, the selection within family wouldcause a downward bias in the estimated send-down effect. Given the findings in Liet al. (2010), the estimated positive effect of the send-down experience suggestedin this paper would be a lower bound. Li et al. (2010), however, discovered thisparent selection effect from a twin study.5 It is unlikely that the local governmentwould have allowed parents much freedom in planning and choosing which childto send if the children were not twins, given that the local government had a send-down quota to fill each year, which varied considerably from year to year.6This paper contributes to a large body of literature including research on thesend-down experience, military service, households in the conflict environment,education, and adolescent development. In the send-down literature, papers havefocused on the outcome of the send-down experience from different perspectives.By using the fact that parents were forced to choose one of their twins to senddown, Li et al. (2010) identify the roles of altruism, favoritism, and guilt in par-ents’ behavior towards their children. Among sociologists, Zhou and Hou (1999)along with Chen and Cheng (1999) report that the traumatic send-down experi-5Except for the first two years of the second stage of the send-down movement, the localgovernment usually required at most only one child to be sent down from each family each year.Therefore, parents with twins had to choose one of the twins to be sent down.6Bernstein (1977); Singer (1971); Unger (1980); Zhou and Hou (1999); Xie et al. (2008) sug-gest that the number of send-downs varied largely from year to year due to the changes in send-down policies. Figure 3.1 illustrates this variation.70ence had a positive effect on the future income of those sent down. However, Xieet al. (2008) suggest that the send-down experience does not affect their income.This paper focuses on education attainment after the send-down movement andsuggests that education upgrading was a key factor that led to a large positiveoutcome following the send-down experience.Because of the hardships induced by the send-down experience, the effects ofthe send-down experience might be comparable with the effect of military serviceon an individual. Studies suggest that military experience combined with finan-cial support has a positive effect on education attainment for returning veterans(Bound and Turner 2002; Lemieux and Card 2001). On the other hand, there ismixed evidence across countries regarding the effect of military experience on in-come (Card and Cardoso 2011; Earnings and Records 1990; Joshua et al. 2011;Albrecht et al. 1999; Imbens and an der Klaauw 1995). This paper provides evi-dence that a forced hard experience might have a positive effect on education at-tainment without the financial support offered by associated programs, such as the“G.I. bill” in the US. Furthermore, the hard experience could have a positive ef-fect on earnings depending on whether individuals upgraded their education afterthe hardship. Future research might seek to investigate the education-upgrading-dependent heterogeneous effect of military service in other countries.The remainder of this paper is constructed as follows. After providing back-ground information and documentation on the send-down policy in the followingsection, I introduce the process of sample restriction in Section 3. Section 4 de-scribes the education interruption during the Cultural Revolution, the education-71upgrading movement after the Cultural Revolution, and the impact of the send-down experience on the choice to upgrade education. Sections 5 and 6 present theestimation results of the send-down effect on income and computer ownership,respectively. Section 7 provides the robustness check, and Section 8 concludesthe paper.3.2 Background3.2.1 The Send-down PolicyThe send-down movement is also known as the “rustication movement.” In Chi-nese, it was called “Shang Shan Xia Xiang,” meaning “going up to the mountainsand down to the villages” (Bernstein 1977). The send-down program began in1960 and ended around 1978.Before 1967, the targets of the send-down program were workers, employees,and jobless city dwellers, as well as elementary and junior high school graduates.At this point, people were mostly persuaded—not forced—to go to rural areas.Voluntary send-down numbers dropped when urban people learnt more about therealities of rural life; they were troubled by the hardship of manual labor and theinability to support themselves (Pan 2002). Approximately one million individu-als were sent down during this stage.The second stage of the send-down movement was initiated by Mao’s speechin 1968: “It is necessary for educated young people to go to the countryside tobe reeducated by the poor and lower middle class peasants. Cadres and other city72people should be persuaded to send their sons and daughters who have finishedjunior or senior high school, college, or university to the countryside”(Pan 2002).The second stage of the send-down movement came to be regarded as a politicalcommand. It was primarily forced rather than voluntary. More than 16 million in-dividuals were sent down between 1968 and 1978. From economic administrators,cadres to students and their parents, if one refused to take part in the send-downprogram, they could be accused of opposing the great strategy of Chairman Mao(Zhang 2000; Pan 2002).7The massive send-down movement resulted in 10.5% of China’s total non-farming population in 1979 being sent down(Pan 2002), with almost every urbanfamily having at least one child sent down (Bernstein 1977). Every year, localgovernments had a quota of send-downs to fill (Pan 2002; Bernstein 1977; Singer1971). The quota varied largely by year. Local government determined the send-down selection process based on the quota and the number of age eligible youths(junior or senior high school graduates in their graduation year). If the quota washigh, all age eligible youths would be sent down. If the quota was low, the localgovernment would allow families who had already sent away a proportion of theirchildren to keep their current age eligible child.8The blue bar of Figure 1 indicates the number of individuals sent down eachyear in the China General Social Survey (CGSS) 2003 data (see data appendix for7Pan (2002) documented that Mao was essentially anti-urban, anti-intellectual, and pro-rural.This was at the root of his support for the send-down movement.8China had a huge baby boom after the war; the average number of children per family duringthe 1960s and 1970s was four (Zhou 2013a; Banerjee et al. 2010).73details). There was a substantial increase in the number of people sent down in1968, the year Mao made his famous speech about the send-down policy. Statis-tics also show that the number of send-downs varied considerably from year toyear—a variation caused by the differences in yearly send-down quotas (Pan 2002;Bernstein 1977; Singer 1971). The send-down policy was intensively executedthroughout the country at the beginning of the second stage of the send-down.The number of sent-down youths decreased between 1970 and 1972 and reachedanother peak in 1974 and 1975.The red line in Figure 1 indicates the total number of individuals sent down ineach year, as reported in Pan (2002). The two data sources show a very similartendency in the number of send-downs for each year. The send-down movementended in 1978 when the new leadership of the Communist Party took control ofthe government and most of the sent-down youths were allowed to return to urbanareas (Zhou and Hou 1999).93.2.2 Send-down Experience and DocumentationsBeing sent down was an extremely difficult experience for affected adolescents.Several studies (Zhou and Hou 1999; Bernstein 1977; Li et al. 2010) have reportedthat most of the sent-down youths were forced to carry out hard manual labor in9In 1985, the government introduced a policy to compensate the sent-down youths, countingtheir work experience in rural areas as work experience in their current job. The money would beadded to their salaries for the rest of their careers. However, salary increase due to work experiencewas minimal. In 2003, for example, government occupations paid only 1 RMB (0.15 USD) peryear of work experience. Thus, five years of the send-down experience only counted for 5 RMB,which is less than 1% of the average income. In calculating salaries, many companies do notaccount for experience beyond 10 years. Thus, the send-down compensation policy is unlikely toconsiderably affect people’s income and employment.74the fields for as long as 12 hours per day and 7 days per week. On average, theywere forced to stay in rural areas for about five or six years. Disdain for the send-down policy was widely documented following the Cultural Revolution.However, the forced years of hard manual labor could have helped urban youthdevelop a kind of endurance for, or resistance against, future hardships. In ruralareas, without parental support, youth were forced to acclimate to an entirely dif-ferent environment. The process of overcoming difficulty and surviving in a harshenvironment at a young age proved to be an important life experience. A sub-stantial number of documents report that the sent-down youths developed a toughworking spirit through the hard manual labor experience (Yang 1992; Wang 2006;Liu 2012; Tang 2012). Wang 2006, for instance, reports as follows: “Through thesend-down experience in the rural area, we learned the spirit of hard work frompeasants. We learned that life is tough. The hard experience made us strongerand trained us to have the ability to encounter difficulties ....” Similarly, Liu 2012documented a story of a sent-down individual who succeeded in later life. Thissent-down male suggested that the hard training experience helped him to build astrong spirit for bearing hard work. China’s current president Xi Jinping was alsosent down and received education upgrading after his return to the urban area. XiJinping describes the send-down experience as having motivated him to have thecourage to face difficulties later in his life (Xi 2003).753.3 Sample RestrictionsIn order to estimate the send-down effect, it is necessary to first investigate thecharacteristics and family backgrounds of those who were sent down. The idealcomparison group for those who were sent down should be a group of individualswho were not sent down but had similar characteristics and family backgroundsto those who were sent down during the Cultural Revolution.3.3.1 Treatment Group and Comparison GroupI only focused on the second stage of the send-down (1968–1978) because thiswas a forced movement and was announced without anticipation. The targets ofthe second stage of the send-down program were urban junior and senior highschool students upon their graduation. Therefore, the comparison group consistsof urban residents who had graduated from junior and senior high school duringthe Cultural Revolution.10The youngest send-downs were born in 1963 (graduated from junior highschool in 1978) and the oldest send-downs were born in 1948 (graduated fromsenior high school in 1966).11 This calculation of the birth years of the send-downs is supported by Figure 3.2. The figure presents the proportion of urbanhigh school graduates sent to rural areas by year of birth. As illustrated in Fig-ure 3.2, in the peak year, almost 50% of high school graduates were sent down.In order to avoid potential cohort and age differences between the treatment10Urban residents are defined as individuals with an urban resident card.11Because of the education interruption between 1966 and 1968, individuals sent down in 1968included students who had graduated between 1966 and 1968 Meng and Gregory 2002.76and comparison groups, I restricted the comparison group to individuals born be-tween 1948 and 1963. Note that these individuals were between 40 and 55 yearsold in the CGSS 2003 data.3.3.2 Family BackgroundThe send-down movement during the second stage was forced and unavoidable.It has been well documented that parental social status or political capital did notprevent the youths in certain privileged classes from being sent down (Bernstein1977; Singer 1971; Unger 1980; Zhou and Hou 1999; Xie et al. 2008). The chil-dren of many communist party leaders and government officials were also sentdown. The daughter of Deng Xiaoping (China’s Chairman in the 1980s) and thenephew of Zhou Enlai (China’s first Prime Minister who served between 1949 and1976) were among the privileged children not given preferential treatment.Almost every family in the affected generation had at least one child sent down(Bernstein 1977). The send-down selection was not based on children’s personaltraits; however, previous literature suggests that the send-down program discrimi-nated against a group of children whose parents had college-level education, cap-italist tendencies, were working for a private organization, or owned a privatebusiness (Bernstein 1977; Pan 2002; Zhou and Hou 1999).12The CGSS 2003 provides a detailed set of information about both parentswhen the respondents were 18 years old, which is very close to the time the re-spondents would have been selected to be sent down. The information includes12 During the Cultural Revolution, university education was seriously criticized, as Mao be-lieved high-level education to be a source of inequality.77mother’s and father’s (1) years of education, (2) Communist Party membershipstatus, (3) leadership status—whether they were chief officers of a branch of gov-ernment or leaders in the Communist Party, and (4) capitalist tendencies—whetherthey worked in a private sector or owned a business.Table 3.1 presents descriptive statistics on the family backgrounds of thosesent down and those not sent down. The regression results are reported in Ta-ble 3.2). I divide parent education into three groups: (1) equal to or less thanjunior high school, (2) senior high school, and (3) college-level or above. Paternaleducation is the only statistically significant family background element on send-down probability, such that children whose fathers had college-level educationor higher were more likely to be sent down (column 2 of Table 3.2). Note that,among parents with senior high school education, the proportion of send-downsand non-send-downs is equally distributed.In order to avoid a potential correlation between parents’ education and per-sonal unobserved characteristics, in all further regressions, I focuse only on indi-viduals whose parents (both father and mother) had 12 or fewer years of education.Note that the results in this paper are robust to the exclusion of people whose par-ents had more than 9 years of education or capitalist tendencies (see Section 3.7).In summary, I restrict samples to individuals who (1) were born between 1948and 1963, (2) were junior high school or senior high school graduates between1966 and 1978, (3) were sent down after 1967 if they were sent down, and (4) didnot have parents with more than 12 years of education. Note that the restricted78sample is used to conduct analyses henceforth.13 Further sample restrictions areemployed in the robustness checks.3.4 Education3.4.1 Education Interruption during the Cultural RevolutionThe Chinese Cultural Revolution (1966–1977) caused a large-scale education in-terruption (Meng and Gregory 2002; Giles et al. 2008; Han et al. 2011). Duringthe first two years of the Cultural Revolution, schools at all levels were closed andadmission of new students was stopped. Although high schools were gradually re-opened as of 1968, admission of students to universities resumed only after 1969and on a small scale. Academics-based entrance examinations were not availablefor any level of school during the Cultural Revolution.Table 3.3 shows the number of students by education level for each year.The education policy during the Cultural Revolution significantly affected thenumber of students enrolled in universities and senior high schools. The studentratio of university to senior high school to junior high school was 1:2:9 in 1960(i.e., for every 9 junior high school students, there were 2 senior high schoolstudents and 1 university student). This number jumped to 1:73:479 in 1970 andwent back to 1:18:58 in 1978 when the Cultural Revolution ended. A substantialnumber of individuals lost the opportunity to go to university, and some could not13Nine individuals in the sample were able to return to school after entering the labor forceduring the Cultural Revolution. As returning to school during the Cultural Revolution was anuncommon event, I drop this sample for potential endogeneity, although the estimation results donot change when it is included.79even enter senior high school.143.4.2 Education Upgrading after the Cultural RevolutionAfter the Cultural Revolution, the education system resumed normal operation.Schools that were closed during the Cultural Revolution were reopened. Therewas high demand for reinvesting in education among individuals who had experi-enced education interruption because of the Cultural Revolution (Han et al. 2011).Based on this demand, China gradually increased the number of institutions offer-ing degree programs to people in the labor force. Some programs, such as adulteducation, offered courses at night or on weekends to accommodate students’schedules. The degree programs included senior high school degrees and 3- and4-year university bachelor degrees.15 Many individuals utilized these options togo back to school to compensate for their lost opportunities.In this paper, I refer to the reinvestment in education as “education upgrading.”Specifically, education upgrading applies to individuals who left school duringthe Cultural Revolution but acquired a higher degree of education—senior high14The number of students in university, senior high school and junior high school was 962000,1675000, 8585000 in 1960, 48000, 3497000, 22922000 in 1970 and 856000, 15531000, 49952000in 1978, data source: Comprehensive Statistical Data and Materials on 50 years of New China.Note that due to the population expansion policy during the 1950s and 1970s, the number ofindividuals aged between 10 to 20 has increased from 140 million in 1960 to 235 million in 1978.The absolute number of students in elementary schools and high schools has also increased due tothe expansion of population.15Adult education initially started in China in the 1950s on a very small scale owing to low de-mand. During the Cultural Revolution, adult education, both general and technical, was regardedas heresy and nearly stopped entirely. After the Cultural Revolution, especially after 1980, it wasrestored and quickly came to be offered by large-scale institutions (Duke 1987). The length ofthe degree program offered in the adult education system was approximately equal to that of thenormal degree program.80school or university—after the Cultural Revolution. According to the CGSS 2003data, almost one-fifth of the affected generation upgraded their education after theCultural Revolution.3.4.3 The Send-down Effect on Education UpgradingIn the restricted sample of the CGSS data (See Section 3.3), 24.1% of sent-downmales upgraded their education, compared to 19.6% of the non-sent-down males(Table 3.4). For females, the difference between the two groups was smaller—15.9% of the sent-down group and 14.4% of the non-sent-down group upgradedtheir education. Conditional on education upgrading, on average, the sent-downsbegan upgrading their education in 1985, one year earlier than the non-sent-downs.I use a probit model to test whether the send-down experience statisticallyraised the probability of upgrading one’s education. The results are presented inTable 3.5.EduU pgradei = β0Senddowni +β1Fi +Xiγ +ui (3.1)EduU pgradei is a dummy variable that equals one if an individual’s educationwas upgraded after the Cultural Revolution and zero otherwise. Senddowni is adummy variable that equals one if an individual has been sent down and zero oth-erwise. Xi is a set of observed individual characteristics. It includes the number ofyears of education an individual received before 1978, age, years of CommunistParty membership, and province dummies. The CGSS 2003 data reports individ-uals’ full education history, including the start and end years of each education81program. The education-upgrading and years of education before 1978 dummiesare constructed from these education history data. Age represents the difficultyof returning to school because of biological reasons. The education system wentback to normal in 1978 and gradually expanded thereafter. The older the individ-ual, the more difficult it was to return to school. ui is an error term clustered at theprovince level.Fi is a measure of family background that controls individual i’s unobservedability. It is a linear function of both father’s and mother’s years of education,Communist Party membership, leadership status, and capitalist tendencies. Thesquared term of parents’ years of education is also included. Note that, as long assend-down status is not correlated with family background or ability, excluding Fifrom the regression should not affect the coefficient of senddowni.Columns 1 through 6 only use male samples. In the first column, none of thefamily background variables are controlled. The estimated result suggests that thesend-down experience increased the probability of individuals receiving educationupgrading by 10%. From columns 2 through 6, more and more family backgroundvariables are controlled. The send-down coefficient is highly significant. It isalso fairly constant and close or equal to 10%. This suggests that the send-downselection is unlikely to be correlated with family background or ability in therestricted sample; otherwise, we would observe large changes in the magnitudeof the send-down coefficient.Several years of hard manual labor could have cultivated a strong motivationto avoid manual labor later in life among those sent down, thereby encouraging82their pursuit of higher levels of education upon their return to urban areas. Theyknew that higher education could substantially increase their chances of avoidinghard manual labor.In female samples (column 7), the coefficient is much smaller with a largestandard error. The non-significance of the send-down coefficient among femalescan be explained as follows: when female send-downs returned to urban areas,they had already reached 23 years of age, a typical age for Chinese women to getmarried. Most of the females, therefore, spent more time looking to get marriedand raise children than to further their education. However, after they were mar-ried and had children, it became more difficult for them to go back to school thanthe males did.3.5 The Send-down Effect on IncomeThe long-term send-down effect on income could be ambiguous. It could be pos-itive because the hard manual labor experience could have motivated those sentdown to work harder later in their lives. On the other hand, an average five-yearloss of urban work experience and network connections could have a negativeimpact.Table 3.6 reports the average incomes of the send-down group and the non-send-down group by gender. If we only examine the numbers in the first panelof Table 3.6, it appears that the send-down experience had no impact on incomefor either gender. However, when I further divide income by whether individualsupgraded their education, there is a large difference between the income of those83sent down and those not sent down. For males who upgraded their education, theaverage income of the send-down group is 1587 RMB, higher than the incomeof the non-send-down group by 343 RMB. In contrast, for those who did notupgrade their education, the sent-down males earn an income 165 RMB lowerthan the non-sent-down males.The pattern of income difference in the female samples is similar to that in themale samples. However, the magnitude of the difference is not as large.3.5.1 IdentificationGiven the income differences described in Table 3.6, which suggest a large posi-tive effect of the send-down experience, conditional on education upgrading, I usethe following regression model to estimate the send-down effect on income formales.Incomei =α0Senddown×EduU pgradei+α1Senddowni+α2EduU pgradei+α3Fi+Xiγ+εi(3.2)Incomei is the log monthly income of the individual i. EduU pgradei is aneducation-upgrading indicator dummy that is equal to one if one received educa-tion upgrading and zero otherwise. Xi is a set of individual characteristics thatincludes total work experience, total years of education, employment status, age,number of years of Communist Party membership, and province dummies. Sameas in Equation 3.1, Fi is a function of family background that controls for unob-84served ability.Equation 3.2 is a difference-in-difference type regression model. α1 estimatesthe general difference between send-downs and non-send-downs (i.e., the differ-ence between the two groups regardless of education-upgrading status). For ex-ample, losing several years of urban work experience during send-down representsa common experience between education-upgraded and non-education-upgradedsend-downs. Note that, if the send-down policy generated any other differencesbetween the send-downs and the non-send-downs, these differences are also cap-tured by α1. The difference between the education-upgraded group and the non-education-upgraded group (regardless of the send-down experience) is included inα2. Note that the total years of education includes the years of education upgrad-ing. Therefore, the dummy variable EduU pgradi captures the additional premiumof an individual having upgraded their education.α0 is the variable of interest. The income differences illustrated in the thirdrow of Table 3.6 is captured by α0. α0 estimates the differential effect, which isthe additional difference between send-downs and non-send-downs among onlythose who had upgraded their education.3.5.2 Estimation ResultsThe OLS estimation results of Equation 3.2 are reported in Table 3.7. Column1 includes Send-down, Education Upgrading, and their interaction term, withoutany additional controls. Column 2 through column 8 include controls for individ-ual characteristics and family backgrounds to assess robustness.85Through columns 1–9, α0, the coefficient of the interaction between senddown and education upgrading is significant and stays around 0.2. The magnitudeis also twice the magnitude of the negative send-down coefficient (columns 3–8).This suggests that, conditional on education upgrading, the send-down experiencehas a strong positive effect on income for males. It is worth noting that the condi-tional difference between the send-downs and the non-send-downs is robust evenafter controlling for occupation dummies (columns 7–8).The send-down coefficient is negative and significant at 10% from columns3–8. In column 9, where I excluded the years of send-down experience fromtotal work experience, the negative coefficient for send-down becomes smallerand non-significant. This suggests that the negative effect of send-down could bedriven by the loss of urban area work experience. In rural areas, the send-downswere usually assigned to do agricultural work. The agricultural-work experiencewould hardly contribute to an urban job.The coefficient of the education-upgrading dummy is positive but non-significant.The results suggest that there is no additional premium for upgrading educationamong non-send-downs. This is not surprising, as the number of years of educa-tion upgrading is included in the total years of education. In addition, Fi controlsfor unobserved ability. The education-upgrading dummy might only capture thedifference in education quality before and after 1978. The education quality af-ter the Cultural Revolution was, in general, higher than that during the CulturalRevolution. However, if we account for the fact that when those individuals up-graded their education, they had already reached the age of 30 and likely had a86daytime job, we might not observe a substantial increase in return to educationamong non-send-downs.Column 8 adds a government-related work place indicator and its interactionterm with send down. After the Cultural Revolution, if the government providedany informal compensation for people who were sent down, sent-down individualswho work in a government-related workplace would be more likely to have ahigher income. The non-significance of the send-down by government interactionterm suggests that it is unlikely that the government compensated the sent-downpeople in any informal manner. Similar to the estimation results in Table 3.5, thesend-down experience does not have a significant effect on females (column 10).3.6 The Send-down Effect on Computer OwnershipAfter the Cultural Revolution, China had a series of economic reforms. The socialeconomic environment changed dramatically. As an example, computers are oneof the new technologies favored by rapid economic and technological growth.Knowing how to use a computer could have potentially benefited individualsduring the period of socioeconomic and technological growth. However, learningto use a computer might have been a challenge for both the send-down and thenon-send-down groups. Computers made their presence in China in the early1990s and came to prevail only after 2000. It takes time and effort to learn to usea computer even for the young, let alone for individuals who are 40 or 50 yearsold. Owning a computer could serve as an indicator of an individual’s ability toquickly adapt to technological change.87The CGSS data ask respondents various home appliances they own: comput-ers, color TVs, air conditioners, and video cameras. Unlike computers, homeappliances such as color TV, air conditioners, and video cameras require little orno learning skills and bring almost no benefit to an individual’s earnings or em-ployment opportunities. As the send-down experience should have no effect onownership of these non-skill-related appliances, I estimate this effect using falsi-fication tests in my investigation of the send-down effect on computer ownership.Panel A of Table 3.8 presents the statistics of the dummy variable computerownership by gender, send-down experience, and education-upgrading status. Inthe education-upgraded male samples, the sent-down group has 20% more indi-viduals have computers than the non-sent-down group. There is a similar tendencyin the female sample; however, the difference is much smaller. In panel B, the es-timation results suggest that, conditional on education upgrading, the sent-downmales own more computers than non-sent-down males do. However, there are nosignificant differences in ownership of other major household appliances.One may have a concern that computers are used by the children of respon-dents rather than the respondents themselves. In order to limit this bias, I restrictedthe samples to individuals who are not living with their children or do not havechildren. The regression results are presented in the last column of Table 3.8.The estimation results are consistent with the finding in column 1, although thestandard errors increased because of the small sample size.883.7 Robustness Check3.7.1 IVA potential concern in the identification strategy is that parents education andjob information can not fully control for unobserved ability, and in addition, thedifference-in-difference type of specification can not fully control for the generaldifference between education upgraded group and non-education upgraded group.If this is the case, it would result in endogeneity in education-upgrading choice.For this reason, I use exogenous variation from the school closure policy duringthe Cultural Revolution to instrument the education-upgrading choice. Introduc-ing IVs helps solve the endogeneity problem; the trade-off is that it only identifiesa local effect.Individuals upgraded their education because their education was interruptedduring the Cultural Revolution. During the Cultural Revolution, at least two ex-ogenous factors determined whether a senior high school student could move onto university after having graduated: the number of full-time teachers employedat the university and the number of senior high school students. The former mea-sures the number of universities or schools that had not been closed; the lattermeasures the number of individuals that could potentially compete for admission.The number of full-time teachers was exogenous because it was determined byeducation policies during the Cultural Revolution, such as school closures. Thenumber of students can be considered exogenous because it was affected by theeducation policy as well as by the population expansion policies during the 1950s89and 1960s.16I divide the number of full-time university teachers by the number of seniorhigh school students to measure the possibility of education interruption that a se-nior high school student could have experienced education interruption during theCultural Revolution. If there were relatively fewer full-time university teachers forthe number of senior high school students in the region in which the senior studentgraduated, it would be more likely that this student’s education was interrupted.The student would, therefore, be more likely to have sought education upgradingafter the Cultural Revolution. It might seem plausible to divide number of uni-versity teachers by the population to calculate the per capita number of teachers,rather than dividing the number of teachers by the number of senior high schoolstudents. Note, however, that only a subset of the population had possibility ofattending university; only senior high school students could potentially have thisopportunity. Therefore, dividing the number of university teachers by the numberof senior high school students would better capture competitiveness.By the same logic, I use the ratio of senior high school teachers to junior highschool students to measure the probability of a junior high school student experi-encing education interruption. The teacher-student ratios varied across provinceand years. I match the teacher-student ratio with the individuals’ end-of-schoolingyear (during the Cultural Revolution), the level of schools these individuals couldpotentially attend (either university or high school), and the province in which16The Chinese government introduced population expansion policies during the 1950s and1960s, which resulted in substantial population growth.90they lived. For example, the measure for XiaoMing, who graduated from a seniorhigh school in Shanghai in 1972, is the university teacher to senior high schoolstudent ratio in Shanghai in 1972, while the measure for HaiLiang, who graduatedfrom a junior high school in Beijing in 1969, is the senior high school teacher tojunior high school student ratio in Beijing in 1969. In the rest of the paper I referto this instrument as “Teacher Ratio.”The variation is based on the differences in Teacher Ratio across the yearswithin each province. Note that province dummies are included in all regressionsin this paper. They control for all provincial-level time-invariant factors.Because the Teacher Ratio measures the probability of students having goneto upper degree schools during the Cultural Revolution, the smaller the TeacherRatio, the more likely an individual’s education was interrupted during the Cul-tural Revolution, and therefore, the more likely an individual would have chosento upgrade their education after the Cultural Revolution. That is, we would expectTeacher Ratio to have a negative effect on education upgrading.I also interact Teacher Ratio with age and use it as the second instrumentvariable. As shown in Section 3.4.3, age is also an important factor affectingeducation upgrading. When the education system resumed normality, the olderthe individual, the higher the cost of education upgrading. This could be due toboth biological reasons and family reasons, such as raising children. As they growolder, individuals would be less likely to upgrade their education. Therefore, the“lost opportunity” effect might diminish with age. In the first stage, therefore, wewould expect the coefficient of the interaction between teacher ratio and age to91have an opposite sign to the teacher ratio coefficient.Table 3.9 reports the IV estimation results. Birth year dummies are includedto control for cohort effects.17 In column 1, the coefficient for TeacherRatio isnegative and its interaction term with age is positive. This is consistent with what Iexpected: individuals who graduated in a low provincial teacher ratio year duringthe Cultural Revolution were more likely to have experienced education interrup-tion; therefore, they would have been more likely to upgrade their education afterthe Cultural Revolution. This lost opportunity effect diminishes if the individualwas older.I further divide samples by non-send-down and send-down (columns 2 and 3)instead of reporting the regression results of the interaction between send-downand education upgrading. Thereby, we can gain a better understanding of how theteacher ratio, the (“lost opportunity”), affects each group. Section 3.4.3 suggestedthat the harsh manual-labor experience induced the sent-downs to upgrade theireducation. Similarly, the estimation results in columns 2 and 3 suggest that thesent-downs were more affected by the “lost opportunity” than were the non-sent-down individuals. The size of the coefficient for teacher ratio within the send-down group is much larger than that of the non-send-down group.The second stage estimation results are consistent with the OLS results. Con-ditional on education upgrading, sent-down males earn significantly more incomeand are more likely to own a computer than those who also received education17CGSS 2003 is a one year individual level data, therefore birth year dummies are equivalent toage dummies.92upgrading but had not been sent down. The size of the IV estimates is largerthan that of the OLS estimates. The estimation results in column 3 indicate thatan education-upgraded sent-down male will earn a 35% higher income comparedto one who also received education upgrading but had no send-down experience(subtract 0.18 from 0.53 in column 4).The IV estimates are more than two times greater than the OLS estimates.One reason is that the instruments identify a local average treatment effect. Peo-ple might upgrade their education for many reasons, such as new schools openingnear their home. However, the compilers in the IV strategy are those who up-graded their education only because of the education interruption. The compilerslikely would have been qualified to go to upper-level school had there been no Cul-tural Revolution. Compared to those who would have been disqualified for highereducation regardless of the education policies, the compilers potentially have ahigher return to education. They had been denied the opportunity to achieve theirdesired level of education. In addition, by going to rural areas to carry out hardmanual labor, they were set back significantly. The joint experience of educationinterruption and hard manual labor could have motivated some send-downs to up-grade their education and work harder once they regained the opportunity to doso.3.7.2 Other Robustness ChecksSocial networks play an important role in affecting individuals’ wage and employ-ment opportunities in China (Wang 2013). Several years away from urban areas93could have potentially weakened the network of connections among sent-downs,thereby causing an income discrepancy compared with non-sent-downs. The firstrobustness check exercise adds a family connection indicator in the regression tocontrol for any potential correlation between the sent-downs and social networks.The indicator comes from the survey question “How many of your relatives orfriends helped get you your job?” The estimation results are presented in panelA of Table 4.7. The coefficients of both Senddown alone and its interaction termwith EduUpgrade remain nearly unchanged from corresponding estimations inprevious tables. This suggests that family connections are unlikely to affect ourestimation results.Early in the 1970s, the government began allowing some sent-down youths toreturn to urban areas if they could find a job or if they were accepted at a schoolin an urban area. Li et al. (2010) and Zhou and Hou (1999) suggest that well-connected families were able to get their children back to urban areas earlier thanothers. Thus, it is possible that controlling the family connection indicator may notfully solve the problem here. In order to avoid the potential endogeneity problemresulting from early return events, I dropped all sent-down individuals who wereable to return to urban areas before the end of the Cultural Revolution.18 Theseresults are presented in panel B of Table 4.7. The results suggest that family con-nections and early returns are unlikely to have affected the estimated send-downeffects. The coefficients for send-down and its interaction term remain statisticallysignificant and the sizes approximated those previously estimated.18 This accounts for 29% of the total male send-down population in the data.94I further tested the send-down effect among individuals with different familybackgrounds. Specifically, I focused on the following family backgrounds: (1)parents who did not work in a private firm and did not own a private business (i.e.,did not have capitalist tendencies); (2) parents who were not Communist Partymembers; (3) parents with only junior high school education or lower; (4) fatherwho worked in nongovernment sectors; and (5) father who was in an unskilledwhite collar or blue collar occupation. In (4) and (5), I do not restrict by mother’swork place or occupation because relatively few individuals had a working motherwhen they were 18 years old.Samples (2) through (5) include individuals with “disadvantaged” family back-grounds. Children from these family backgrounds likely had less political power,less government-related connections, and/or less motivation for higher education.From the estimation results in the previous sections, we generated several signif-icant positive effects for the send-down experience: sent-down males are morelikely to have upgraded their education and, conditional on education upgrading,they earn higher incomes and are more likely to have computers at home. There-fore, I focus on individuals with “disadvantaged” family backgrounds, investigat-ing whether the positive effects of the send-down experience could be driven bydifferences in family backgrounds.The results are reported in the remaining panels in Table 4.7. All the coef-ficients in Table 4.7 have the correct sign, and all of them are not statisticallydifferent from the regression results in the previous sections. Overall, the resultsreported in Table 4.7 suggest that the send-down effects are robust against various95types of family backgrounds.3.8 ConclusionThe forced send-down movement affected more than 16 million urban youths inChina. Several years of manual labor experience in rural areas were undeniablyhard on those urban youths who were, on average, only 17 years old when theywere sent down. The loss of years of urban work experience caused a negativeeffect on income. However, the estimation results suggest that the hard manuallabor experience induced those urban youths to upgrade their education after theCultural Revolution, and conditional on upgrading education, the send-downs earnhigher incomes than the non-send-downs.In the current political environment, no policy makers would consider initiat-ing a similar send-down movement again. However, the send-down event mightelucidate some important factors in the education of teenage children. Hard men-tal and physical training might not be as detrimental as once thought. Childrenexperiencing difficulties and overcoming these difficulties independently mightbecome stronger and work harder in their later life, just as numerous send-downshave described how the hard send-down experience had cultivated in them a strongspirit (Yang 1992; Wang 2006; Liu 2012; Tang 2012). More evidence is needed tounderstand how adversities could affect youth and shape their path in the future.Future study could focus more on the effect of adversity during adolescence orchildhood. It would also be interesting to compare the short term and long termeffect of adversities.96Figure 3.1: Number of Youth Sent to Rural Areas by YearNote: The blue bars shows the number of individuals were sent to rural areas each year reported inthe individual level survey data, China General Social Survey 2003. The red line shows the totalnumber of individuals were sent to rural areas each year. It is calculated by author based on thedata reported in Pan (2002).97Figure 3.2: Send-down Proportion by Year of BirthNote: The proportion is among junior high school and senior high school graduates in urban areas.Data sources: China General Social Survey 2003.98Table 3.1: Individual Characteristics and Family Background during theCultural RevolutionSend-down Non-send-downStandard StandardMean Deviation Mean DeviationVariable (1) (2) (3) (4)Family Backgrounds at Age 18Father:Years of Education 6.22 4.66 5.09 4.45Proportion of Junior High School or below 0.84 0.37 0.88 0.326Proportion of Senior High School 0.08 0.27 0.08 0.26Proportion of College or above 0.08 0.28 0.05 0.21Proportion of Leader 0.05 0.21 0.03 0.17Proportion with Communist Party Membership 0.31 0.46 0.28 0.45Proportion with Capitalism Traits 0.02 0.13 0.02 0.15Mother:Years of Education 3.37 4.29 2.84 4.05Proportion of Junior High School or below 0.92 0.27 0.94 0.23Proportion of Senior High School 0.05 0.23 0.04 0.18Proportion of College or above 0.02 0.15 0.02 0.14Proportion of Leader 0.01 0.08 0.00 0.06Proportion with Communist party membership 0.07 0.26 0.06 0.23Proportion with Capitalism Traits 0.01 0.11 0.01 0.08Send-down Duration 5.33 3.41Age upon send-down 17.95 1.39Female 0.55 0.50 0.46 0.50Proportion Junior High School Graduates 0.66 0.47 0.65 0.48Obs. 333 970Note: I restrict samples to individuals who (1) were born between 1948 and 1963, (2) were juniorhigh school or senior high school graduates between 1966 and 1978, (3) were sent down after1967 if they were sent down.99Table 3.2: Probit Estimation of Send-downDependent Variable: Send-down(1) (2)Family Backgrounds at Age 18Father:Years of Education 0.01∗∗∗(0.003)Senior High School 0.01(0.05)College or above 0.16∗(0.10)Leader 0.04 0.07(0.06) (0.07)Communist Party Membership 0.02 0.02(0.03) (0.04)Capitalism Traits -.08 -.06(0.10) (0.11)Mother:Years of Education 0.002(0.003)Senior High School 0.09(0.07)College or above -.09(0.07)Leader -.03 -.03(0.14) (0.14)Communist Party Membership 0.01 0.03(0.07) (0.07)Capitalism Traits 0.10 0.08(0.20) (0.19)Obs. 1203 1203Pseudo R2 0.11 0.11Note: Marginal effects are reported. Dependent variable Send-down is a dummy variable equal toone if an individual were sent down, 0 otherwise. All regressions control for age, gender, educationdegree during the Cultural Revolution and province dummies. I restrict samples to individuals who(1) were born between 1948 and 1963, (2) were junior high school or senior high school graduatesbetween 1966 and 1978, (3) were sent down after 1967 if they were sent down. Province dummies100are included. Standard errors in parentheses are clustered at the province level. *** p<0.01, **p<0.05, * p<0.1.101Table 3.3: Number of Students (10,000 person)University Senior High Junior High1960 96.2 167.5 858.51961 94.7 153.3 698.51962 83.0 133.9 618.91963 75.0 123.5 638.11964 68.5 124.7 729.41965 67.4 130.8 803.01966 53.4 137.3 1112.51967 40.9 136.5 1097.21968 25.9 140.8 1251.51969 10.9 189.1 1832.41970 4.8 349.7 2292.21971 8.3 558.7 2568.91972 19.4 858.0 2724.41973 31.4 923.3 2523.21974 43.0 1002.7 2647.61975 50.1 1163.7 3302.41976 56.5 1483.6 4352.91977 62.5 1800.0 4979.91978 85.6 1553.1 4995.2Data source: Comprehensive Statistical Data and Materials on 50 years of New China.102Table 3.4: Education UpgradingMale FemaleSend-down Non-Send-down Send-down Non-Send-downProportion Edu Upgrade 24.1% 19.7% 15.9% 14.4%among junior high 20.2% 16.4% 14.8% 12.3%among senior high 31.9% 25.3% 17.9% 18.7%Age Upgraded 31 30 31 29Year Upgraded 1985 1986 1985 1986Note: I restrict samples to individuals who (1) were born between 1948 and 1963, (2) were juniorhigh school or senior high school graduates between 1966 and 1978, (3) were sent down after1967 if they were sent down. (4) did not have parents with more than 12 years of education.103Table 3.5: Probit Estimation: the Impact of Send-down Experience on Edu-cation UpgradingDependent Variable: Education UpgradeMale (1)-(6) Female(1) (2) (3) (4) (5) (6) (7)Send-down 0.10∗∗∗ 0.09∗∗∗ 0.09∗∗∗ 0.09∗∗∗ 0.09∗∗∗ 0.10∗∗∗ 0.02(0.04) (0.03) (0.03) (0.03) (0.03) (0.04) (0.05)Years of Education Before 1978 -.01∗ -.02∗∗ -.02∗∗ -.02∗∗ -.02∗∗ -.02∗∗ -.01(0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01)Age -.01∗∗∗ -.01∗∗∗ -.01∗∗∗ -.01∗∗∗ -.01∗∗∗ -.01∗∗∗ -.06(0.004) (0.01) (0.01) (0.01) (0.01) (0.004) (0.06)Parents Education Y Y Y Y Y YParents Education Squared Y Y Y Y YParents Communist Party Y Y Y YParents Leaders Y Y YParents Capitalism Y YObs. 618 618 618 618 618 618 562Pseudo R2 0.20 0.21 0.22 0.22 0.22 0.23 0.20Note: Marginal effects are reported. The dependent variable is a dummy indicator equal to oneif an individual upgraded education after the Cultural Revolution, zero otherwise. All regressionscontrol for years of communist party member and province dummies. I restrict samples to individ-uals who (1) were born between 1948 and 1963, (2) were junior high school or senior high schoolgraduates between 1966 and 1978, (3) were sent down after 1967 if they were sent down, (4) didnot have parents with more than 12 years of education. Province dummies are included. Standarderrors in parentheses are clustered at the province level. *** p<0.01, ** p<0.05, * p<0.1.104Table 3.6: Descriptive Statistics of Monthly Income by Gender and Educa-tion UpgradingSend-down Non-send-downStandard StandardMean Deviation Mean Deviation(1) (2) (3) (4)Male 1011 762 1023 1043Female 757 593 792 624MaleEducation Upgraded 1587 1034 1244 711Not Education Upgraded 795 485 960 1113FemaleEducation Upgraded 1181 830 1067 531Not Education Upgraded 659 476 733 627Note: I restrict samples to individuals who (1) were born between 1948 and 1963, (2) were juniorhigh school or senior high school graduates between 1966 and 1978, (3) were sent down after1967 if they were sent down, (4) did not have parents with more than 12 years of education. Unit:RMB.105Table 3.7: The Impact of Send-down Experience on IncomeDependent Variable: IncomeMale(1)-(9) Female(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)Send-down × EduUpgrade 0.2∗ 0.2∗ 0.2∗∗ 0.19∗∗ 0.19∗∗ 0.19∗∗ 0.2∗∗ 0.2∗∗ 0.2∗∗ -.09(0.1) (0.1) (0.1) (0.09) (0.09) (0.09) (0.09) (0.1) (0.09) (0.11)Send-down -.03 -.07 -.10∗ -.10∗ -.10∗ -.09∗ -.09∗ -.09 -.04 -.02(0.08) (0.06) (0.06) (0.06) (0.06) (0.05) (0.05) (0.05) (0.05) (0.09)Edu Upgrade 0.42∗∗∗ 0.07 0.07 0.07 0.07 0.06 0.05 0.05 0.05 0.13(0.07) (0.08) (0.08) (0.08) (0.08) (0.08) (0.09) (0.09) (0.09) (0.1)Send-down × Government 0.00(0.2)Total Years of Education 0.06∗∗∗ 0.05∗∗∗ 0.05∗∗∗ 0.05∗∗∗ 0.05∗∗∗ 0.05∗∗∗ 0.05∗∗∗ 0.05∗∗∗ 0.08∗∗∗(0.01) (0.01) (0.01) (0.01) (0.01) (0.02) (0.02) (0.02) (0.02)Experience 0.02∗∗∗ 0.02∗∗∗ 0.02∗∗∗ 0.01∗∗∗ 0.01∗∗∗ 0.02∗∗∗ 0.02∗∗∗ 0.02∗∗∗(0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01)Parents Education Y Y Y Y Y Y Y YParents Communist Party Y Y Y Y Y Y YParents Leaders Y Y Y Y Y YParents Capitalism Y Y Y Y YOccupation Dummies Y Y Y YGovernment YExperience w/o SD Years YObs. 583 583 583 583 583 583 571 571 571 505R2 0.07 0.41 0.42 0.43 0.43 0.43 0.45 0.45 0.45 0.39Note: All regressions control for years of communist party member, employment status and province dummies. Column 9 uses experience,which excludes send-down years. Government is a dummy variable which equals to one if an individual works in government relatedwork place or state-owned firms. I restrict samples to individuals who (1) were born between 1948 and 1963, (2) were junior high schoolor senior high school graduates between 1966 and 1978, (3) were sent down after 1967 if they were sent down, (4) did not have parentswith more than 12 years of education. Standard errors in parentheses are clustered at the province level. *** p<0.01, ** p<0.05, * p<0.1.106Table 3.8: The Impact of Send-down on Having ComputersPanel A: Descriptive Statistics of Computers at HomeMale FemaleSend-down Non-send-down Send-down Non-send-downEducation Upgraded 0.59 0.39 0.50 0.45(0.09) (0.05) (0.10) (0.06)Not Education Upgraded 0.21 0.24 0.31 0.29(0.04) (0.02) (0.04) (0.02)Panel B: Regression Results of Home AppliancesComputers Color TVs Air Conditioners Video Cameras ComputersSend-down × EduUpgrade 0.24∗∗ 0.028 -.050 0.033 0.396∗(0.115) (0.027) (0.083) (0.037) (0.209)Send-down -.031 0.007 0.067 0.005 -.005(0.029) (0.016) (0.056) (0.018) (0.068)Edu Upgrade 0.043 -.015 0.047 -.026 -.097(0.046) (0.023) (0.061) (0.017) (0.187)Obs. 619 619 619 619 136R2 0.31 0.21 0.33 0.20 0.54Note: Standard deviations are presented in parentheses of panel A. The regression in the last column of Panel B uses only individualswho are not living with children or they don’t have children. All regressions control for individual income, household income, number ofchildren, age of the youngest child, presence of female children, one digit occupation dummies, years of education, experience, years ofcommunist party member, employment status, family backgrounds. I restrict samples to individuals who (1) were born between 1948 and1963, (2) were junior high school or senior high school graduates between 1966 and 1978, (3) were sent down after 1967 if they were sentdown, (4) did not have parents with more than 12 years of education. Province dummies are included. Standard errors in parentheses areclustered at the province level in panel B. *** p<0.01, ** p<0.05, * p<0.1.107Table 3.9: The Impact of Send-down Experience on Males’ Income and Computer Ownership (IV)1st Stage 2nd StageEduUpgrade Income Income Income Income Computer Color TVs Air Conditioners Video CamerasAll Non-Send-down Send-down (4) (5) (6) (7) (8) (9) (10) (11)Teacher Ratio -.02∗∗ -.02∗∗∗ -1.57∗∗(0.009) (0.009) (0.62)Teacher Ratio × Age 0.0004∗∗ 0.0004∗∗ 0.03∗∗(0.0002) (0.0002) (0.01)Send-down × Teacher Ratio 0.01(0.03)Send-down × EduUpgrade 0.53∗∗ 0.36∗∗ 0.52∗∗ 0.45∗∗ 0.734∗∗∗ 0.115 0.508 -.101(0.22) (0.17) (0.25) (0.19) (0.262) (0.85) (0.511) (0.085)[0.06,1.06] [-0.02,1.09] [-0.01,1.35] [0.03,0.97] [0.28,1.12]Send-down 0.03 -.18∗ -.04 -.13 -.11 -.118 0.148 -.041 0.041(0.11) (0.10) (0.09) (0.08) (0.1) (0.076) (0.184) (0.125) (0.03)Send-down × Government -.21(0.22)EduUpgrade -.48 -.38 -.66 -.46 -.255 -1.400 -.505 -.016(0.41) (0.38) (0.45) (0.37) (0.869) (1.678) (1.204) (0.09)Experience w/o SD Years YOccupation Dummies Y Y Y Y Y YGovernment YAll Family Backgrounds Y Y Y Y Y Y Y Y Y Y YF-Statistics 10.42 3.71 3.67Obs. 542 423 119 542 542 532 542 581 581 581 581Note: All regressions control for total years of education, experience, age, years of communist party member, employment status andprovince dummies. Only male sample are used. Column 5 uses experience, which excludes send-down years. Government is a dummyvariable which equals to one if an individual works in government related work place or state-owned firms. I restrict samples to individualswho (1) were born between 1948 and 1963, (2) were junior high school or senior high school graduates between 1966 and 1978, (3) weresent down after 1967 if they were sent down, (4) did not have parents with more than 12 years of education. Standard errors in parentheses108are clustered at the province level. Anderson-robin weak IV robust 90% confidence intervals were presented in square parentheses. ***p<0.01, ** p<0.05, * p<0.1.109Table 3.10: Other Robustness ChecksDependent VariablesEducation Upgrade Income (IV) Computers (IV)(1) (2) (3)Panel A. Family Connection ControlledSend-down × EduUpgrade 0.51∗ 0.80∗∗∗(0.27) (0.30)Send-down 0.10∗∗∗ -.18∗ -.13∗(0.04) (0.09) (0.08)Obs. 617 541 580Panel B. Early Return DroppedSend-down × EduUpgrade 0.48∗∗ 1.08∗(0.21) (0.65)Send-down 0.11∗∗ -.18∗∗ -.18(0.05) (0.07) (0.13)Obs. 577 507 540Panel C. Parents Working in Private Firms DroppedSend-down × EduUpgrade 0.49∗ 0.68∗∗∗(0.26) (0.25)Send-down 0.10∗∗ -.16∗ -.09(0.04) (0.09) (0.08)Obs. 600 527 565Panel D. Parents Non-communist Party MemberSend-down × EduUpgrade 0.96∗ 1.27(0.52) (1.12)Send-down 0.09∗ -.28∗∗∗ -.24(0.05) (0.09) (0.18)Obs. 417 401 432Panel E. Parents with Junior High Education or BelowSend-down × EduUpgrade 0.69∗∗∗ 0.94∗∗∗(0.16) (0.25)Send-down 0.08∗∗ -.16∗∗ -.14∗(0.04) (0.08) (0.07)Obs. 550 498 535Panel F. Father in Non-government Sector OnlySend-down × EduUpgrade 0.41∗∗ 0.83∗∗∗(0.21) (0.31)Send-down 0.09∗∗ -.13∗ -.13∗(0.04) (0.08) (0.07)Obs. 590 508 561Panel G. Father Non-skilled White or Blue Color OccupationSend-down × EduUpgrade 0.83∗ 1.26∗(0.50) (0.67)Send-down 0.09∗ -.26∗∗ -.17(0.05) (0.11) (0.11)Obs. 487 436 419110Note: Only male samples are used. All regressions control for experience, age, years ofcommunist party member, employment status, family backgrounds and province dummies. Inaddition, column 1 controls for years of education during the Cultural Revolution; column 2 and3 controls for years of education; Column 3 further controls for personal income, householdincome, number of children, age of the youngest child, presence of female children, one digitoccupation dummies, experience. Column 1 reports the marginal effect of the probit model. Irestrict samples to individuals who (1) were born between 1948 and 1963, (2) were junior highschool or senior high school graduates between 1966 and 1978, (3) were sent down after 1967 ifthey were sent down, (4) did not have parents with more than 12 years of education. Standarderrors in parentheses are clustered at the province level. *** p<0.01, ** p<0.05, * p<0.1.111Chapter 4The Long-term Impact of theSend-down Experience: Happinessin Life, Political Attitudes, andInvestment in Children4.1 IntroductionThe send-down movement (1968-1978) during the Chinese Cultural Revolutioncaused more than 16 million adolescents in urban areas to move to rural areasto carry out hard manual labor. The traumatic experience significantly affectedthe education attainment of those people who were sent down and their income(Chapter 3). This paper investigates the send-down effect on outcomes other than112education and income. Specifically, it investigates the feeling of happiness in thelives of those who were sent down, their political attitudes, and their investmentin the education of their children.The hard manual experience during adolescence certainly did not make thesent-down youth happy during the send-down period. However, it is unclear howsuch a traumatic experience would impact the youths’ feelings about life whenthey reached age 40s or 50s. This paper suggests that the traumatic experienceduring adolescence has had a surprisingly long-lasting effect. Even decades afterthe traumatic experience, those who were sent down still feel significantly lesshappy about their lives than do those who were not sent down. The hard experi-ence during adolescence could have permanently fostered bitterness in the mindsof those sent down.This paper also investigates the effect of the send-down experience on the po-litical attitudes of those who were sent down. The send-down movement was ini-tiated by Chairman Mao Zedong during the Cultural Revolution. The sent-downyouth experienced great hardship because of a decision made by a few individ-uals in government. Would such a traumatic experience enforced by a stronglydictatorial government affect individuals’ beliefs in democracy and their attitudestowards the government? This paper suggests that the sent-down group are lesslikely to agree with the statement that “democracy means the government shouldmake decisions on behalf of the people.” Furthermore, this paper finds that in-dividuals who experienced send-down are less likely to become members of theCommunist Party. The communist political party has ruled the People’s Republic113of China since the republic was established in 1949.This paper further investigates the impact of the send-down on investment inthe education of the next generation. This paper uses the education fees paid“apart from the tuition uniformly regulated by the state and the local government”as a measure of the investment made towards children’s education. This papersuggests that the sent-down males invested more in their children’s education re-gardless of their own education-upgrading choices. Their hard manual-labor expe-rience as adolescents could have taught them the bitterness of hard manual labor.They hope for a better education and therefore a better life for their children.The data used in this paper comes from the China General Social Survey 2003.The send-down was forced in the second stage of the send-down movement(1968-1978) and its target were junior and senior high school graduates. Furthermore,Chapter 3 provides suggestive evidence that send-down selection was random ifrestrict to individuals who’s parents have less than 12 years of education. Fol-lowing the estimation strategy used in Chapter 3, this paper makes the followingsample restriction in order to identify the send-down treatment effect. I restrictsamples to individuals who (1) were born between 1948 and 1963, (2) were juniorhigh school or senior high school graduates between 1968 and 1978, (3) were sentdown after 1967 if they had been sent down, (4) did not have parents with morethan 12 years of education, and (5) were not able to return to school before 1978after they had entered the labor force. Further sample restriction would be con-ducted in the robustness checks. The estimation results are robust to various typesof family backgrounds. The effect is much stronger upon males than it is upon114females.In the rest of the paper, sections 2, 3, and 4 investigate the send-down expe-rience on happiness in life, political attitudes, and investment in children’s edu-cation, respectively. Section 5 provides the robustness checks. The final sectionconcludes the paper.4.2 HappinessHappiness in life, as an important measure of well-being, has attracted more andmore attention from economists and policy makers in recent decades (Gleibs et al.2013; Helliwell and Wang 2012). Researchers are questioning what factors couldexplain the differences in happiness after controlling for health, wealth, and mar-ital status as those others. Helliwell and Huang (2008), using cross-country-leveldata, suggest that the quality of a political institution could be an important factorin an individual’s well-being or lack thereof.The sent-down youth were forced by the government to move to rural areasand to carry out hard manual labor in the fields. They worked 12 hours a day, 7days a week (Zhou and Hou 1999). This overbearing experience of the send-downduring their adolescence certainly did not make of the youths happy individuals,and it was unclear just how much effect such a traumatic experience would stillhave on their feelings about life when they turned 40 or 50 years old.The CGSS 2003 asks, “Generally speaking, how do you personally feel aboutyour life?” Respondents can choose one of the following five options: 1) Very un-happy 2) Unhappy 3) So-so 4) Happy 5) Very happy. The basic descriptive statis-115tics by the send-down experience are presented in Table 4.1. Table 4.2 presentsthe estimation results by using ordered logit models. I divide the sample by gen-der. Columns 1 and 4 include basic controls; columns 2 and 5 add a set of detailedpersonal characteristics such as employment status, family income, occupation,and information regarding the children; columns 3 and 6 further include a set ofdetailed information of both parents of respondents: education, communist partymembership, a dummy variable equals one if mother or father was a senior man-ager a leader in the work place; a dummy variable equals one if either mother orfather worked in a private firm or owned a private business.Table 4.2 suggests that those who were sent down are significantly less likelyto be happy than are those who were not sent down. Moreover, sent-down malesexperiences a greater negative effect than did the sent-down females. Educationand income have positive effects on happiness in life. This is consistent with mostof the findings in recent research (Frijters et al. 2013; Clark et al. 2008). It isworth noticing that Communist party members are also significantly happier thannon-Communist Party members.The results in this section suggest that a traumatic experience during adoles-cence could have surprisingly long-lasting effects. Even decades after the trau-matic experience, the sent-down group still feels significantly less happy aboutlife than does the non-sent-down group.4.3 Political AttitudesPiketty (1995) provides a theoretical framework and suggests that people formu-116late and modify their attitudes towards government policy based on their personalexperience of it. Recent empirical evidence has shown that belief in or lack ofconfidence in government could be modified through aggregate-level economicshocks or business cycles (Stevenson 2011; Grosjean et al. 2013). The send-downmovement provides an opportunity to test the level to which a bad experiencecaused by the government affects individual’s attitude to this government. In par-ticular, when the bad experience has resulted from a decision taken by a dictatorialgovernment, it would be interesting to know how this decision has affected the in-dividuals’ belief in the government’s ability to have a democracy.The CGSS asks respondents whether they agree or disagree with the statementthat “Democracy means the government should be for the people.” The originalChinese means to ask respondents whether they believe in government that gov-ernment could achieve democracy. The regression results presented in Table 4.3show that individuals who were sent down are significantly less likely to agreewith the statement. The effect is stronger for males than for females.The sent-down experience also resulted in those sent down being less likelyto become members of the Communist Party as compared to those who were notsent down (Table 4.4). The Communist party has been the ruling political partyin China since the birth of the People’s Republic of China. Although the newleaders of the Communist party had promised to bring China great prosperityafter the Cultural Revolution, the past mistakes of the government held a strongpersistent effect on people’s choice not to become party members. In keeping withthe previous findings, the effect is stronger upon males than it is upon females.117The results suggest that experiencing a government-caused bad outcome couldhave significant impacts on individuals’ political altitudes. Note that the estima-tions compare the sent-down youths to the non-sent-down youths in the samecohort, those latter having known about the send-down movement and perhapshaving had friends or siblings who were sent-down. Individuals who experiencedthe bad outcome have significantly different political attitudes compared to thegroup that did not experience the bad outcome but only knew of its existence. Thisfinding may suggest a significant “experiencing” effect relative to the “knowledgeeffect.”The survey also asks respondents whether they agree or disagree with the fol-lowing statements regarding democracy: “It will be democracy only when ordi-nary people have direct voices and decision-making powers on important stateand local matters.” “It is also democracy if ordinary people have the right to votefor their own representatives and to discuss important state and local matters.”The regression results in Table 4.5 show that the send-down experience doesnot affect individuals’ opinion in regard to letting ordinary people have decision-making powers or voting. The send-down experience only informed a limitationin the government’s ability in making democracy. Interestingly, however, such aexperience resulted from a dictatorship does not make people think that lettingordinary people exercise power through decision making or voting are better so-lutions. A potential reason for these results is that there is no difference betweenthe send-downs and the non-send-downs in experiencing seeing ordinary peopleexercise power through decision making or voting.118The CGSS also asks “Generally speaking, how much do you trust strangers?”Respondents are given five options: “Highly untrusted, Untrusted, So-so, Trusted,Highly trusted.”The regression results presented in Table 4.5 suggest that the send-down ex-perience does not have an impact on trust. It is well known that the slave tradecaused Africa to have a low level of social trust (Nunn and Wantchekon 2011).Being sent down and being enslaved are both traumatic experiences (though be-ing sent down may not be as traumatic as being enslaved). There are fundamentaldifferences in the causes of such experiences. Individuals become slaves becausethey have been cheated and sold by strangers, neighbors, relatives, and even thelocal government. Such an experience has caused the low level of general trust inAfrica. However, a traumatic experience caused by a dictatorial government hasnothing to do with mistrust of strangers, neighbors, or relatives. This finding fur-ther confirms the suggestion that attitudes and beliefs are formulated by personalexperience.4.4 Intergenerational Effect: Investment in theNext Generation’s EducationCould the hard manual experience during youth affect individuals’ education in-vestment towards their children? Ideally, we would like to have the informationof total education expenditure on children to estimate the send-down effect. In theabsence of the total education expenditure in the data set, I use the following mea-sure as the best alternative measure I can find. The CGSS 2003 asks, “Apart from119tuition uniformly regulated by the state and the local government, did your familyever make sponsorship contributions, pay self-financing fees, or pay charges forchoosing a school to attend?”After the Cultural Revolution, the education system went back to normal.1 Tu-ition is regulated by the state and the local government. However, apart from theregulated tuition, both public schools and private schools are allowed to chargestudents additional fees (Tsang 2001). These additional fees are referred to as“sponsorship fees.” Some call it “school-choosing fees” or “self-financing fees.”Sponsorship fees vary according to the level of the schools and the students’ aca-demic achievement when entering the school. Often, in public schools, studentsare charged sponsorship fees if they want to go to a particular school but do notmeet the required academic standing. Private schools usually charge fees otherthan the government-regulated tuition fees in order to offset their operating costs.2These fees are sometimes called the “self-financing fees.” Most fees are set byschools when students are accepted at that school.The sponsorship fees caused a large social problem.3 Starting around 2005,the government introduced policies to regulate sponsorship fees; however, parents1 Students are assigned to the local public elementary and junior high schools in their ownvicinities; however, for entering top schools in the local level (town, city, or province), merit-based entrance exams are usually required. Entrance exams are required for entering senior highschools and universities.2 Private schools counted for a very small proportion in the total educational institutions inChina in the early 2000s. Private schools were all abandoned during the Cultural Revolution dueto Mao’s legislation policies. The development of private schools in China began in the later1990s.3Charmon and Prasad suggest that rising education expenditure caused the rising householdsaving rate.120deem the fees to still be far beyond the acceptable level.4 If a student wants to goto a better school, the probability of this student having to pay sponsorship fees ishigher.5In the restricted samples, the proportion of the individuals answered that theyhad paid sponsorship fees to schools for their own children at least once in thepast is 23%, which is a very large proportion.6 The conditional mean of the totalsponsorship fees ever paid is 6294 RMB, which is more than a quarter of the yearlyhousehold income in 2002. The survey also asks the reasons why sponsorshipfees were paid. Conditional on having ever paid sponsorship fees, 64% of therespondents said the fees were for attending a better school or for having failed topass the entrance examinations. Less than 1% of the individuals gave relocationdue to job transfer as the reason for paying the fees. About 38% of the respondentsgave other unknown reasons.The upper panel of Table 4.6 reports the descriptive statistics of the sponsorship-fee payments by send-down and gender. It first presents the statistics of fees paidfor any reason (attending a better school, relocation, or other). Then I calculatethe fees without taking into account the fees paid for relocation or other unknown4According to an investigation conducted by the National Bureau of Statistics of China, 90% ofthe parents think that sponsorship fees are too high (Wang 2004). In 2011, the Beijing governmentabandoned the sponsorship-fee system in all public-founded kindergartens.5We can know this tendency even from policy regulations. Government usually allows top levelschools to collect more sponsorship fees. For example, in 2005 the HuBei provincial governmentintroduced a policy for high schools which dictated that the maximum percentage of enrolledstudents that a school can collect sponsorship fees from in the top provincial-level schools, topcity-level schools, and other schools was limited to 30%, 20%, and 10%, respectively (Ma 2007).6Of those sponsorship fees, 7% were paid to private schools, the remaining were paid to publicschools.121reason, and I call the remaining subgroup “Attending Better Schools.” For bothcategories, the fees paid by the sent-down males are much higher than those paidby the non-sent-down males, regardless of their own education-upgrading choices.The difference between the sent-down and the non-sent-down in the female sam-ple is much smaller than it is in the male sample.In the lower panel of Table 4.6, I present the regression results of the sponsorship-fee payment. In addition to the basic controls included in the income regressions,I also control for personal income, household income, the ratio of personal in-come to household income, number of children, children’s gender, and children’sage.7 The ratio of personal income to household income is included for the con-sideration of potential household bargaining power between husband and wife.The regression results further confirm the suggestion in the upper panel that thesend-down experience significantly increased the sponsorship-fee payments in themale sample.The sent-down group knows the bitterness of doing hard manual labor throughtheir own painful experience. As a potential explanation for the estimation results,the send-down experience made them invest more in their children’s education inorder to help their children avoid hard manual labor in their lives and to have abetter future; they might hope that their children could have a better educationwhich would in turn lead to a better life.7children’s gender is defined as presence of female children, children’s age is defined as theage of the youngest child if there is more than one child.1224.5 Robustness CheckFollowing Zhou (2013b), this paper provides the following robustness check.Panel A adds a family connection indicator; the indicator comes from the sur-vey question “how many of your relatives or friends helped you get your job?”Panel B drops all the individuals who were able to return to urban areas before theend of the Cultural Revolution. I further test the send-down effect among indi-viduals with different family backgrounds. Specifically, I focus on the followingfamily backgrounds: (1) parents who were not working in a private firm and didnot own a private business (did not have capitalist tendency); (2) parents who werenot communist party members; (3) parents with only junior high school educationor lower; (4) father who was working in nongovernment sectors; and (5) fatherwho was in an unskilled white-collar occupation or blue-collar occupation. In (4)and (5) I do not restrict by mother’s work place or occupation because relativelyfew individuals had a working mother when they were 18 years old.The estimation results of robustness checks are presented in Table 4.7. Thecoefficients are almost all statistically significant at least at the 10% level. Overall,the results reported in Table 4.7 suggest that the send-down effects are robust to avariety of controls for family background.4.6 ConclusionThe send-down experience forced by the government had a surprisingly long-lasting effect on the well-being of those sent down and their attitudes towards thegovernment. Those who were sent down are significantly less likely to be happy123in their lives than their counterparts who were not sent down; they are less likelyto believe in the government’s ability to have a democracy. They are also lesslikely to become Communist Party members. However, the send-down experiencedoes not affect individuals’ political attitude towards voting or in seeing ordinarypeople exercise power through decision making; it also has no significant impacton how individuals trust others. The potential reason for this could be that thesend-down experience was caused by a decision made by almost only one personin the government; it was not caused by being cheated by strangers; neither thosewho were sent down nor those who were not sent down have any experience invoting or seeing ordinary people make decisions. All this evidence supports theidea of an experience-based process for formulating attitudes and beliefs.The hard manual labor experience, however, induced those sent down to in-vest significantly more in their children’s education in the hope of helping theirchildren have a better life. The send-down experience could have made those sentdown learn the hardship of doing manual labor and hence made them come torealize that only education could help their children avoid the experience of hardmanual labor.124Table 4.1: Descriptive Statistics, by Send-down ExperienceSend-down Non-send-downMean Std. Dev. Mean Std. Dev.Happiness 3.14 0.77 3.25 0.78(1=Very Unhapppy, 5=Very Happy)Democracy Means Government Should Make Decisions 0.72 0.45 0.77 0.42(1=Agree, 0=Disagree)Proportion of Communist Party Member 0.22 0.41 0.26 0.44Democracy Means Voting 0.87 0.33 0.89 0.31(1=Agree, 0=Disagree)Democracy Means Ordinary People Have Decision Power 0.67 0.47 0.68 0.47(1=Agree, 0=Disagree)Trust 2.13 0.65 2.17 0.63(1=Highly Untrusted, 5=Highly Trusted)Sponsorship Fee 2.47 7.45 1.77 4.92Sponsorship Fee for Better School 2.00 6.69 1.29 4.29Obs. 304 918Note: I restrict samples to individuals who (1) were born between 1948 and 1963, (2) were juniorhigh school or senior high school graduates between 1968 and 1978, (3) were sent down after1967 if they had been sent down, (4) did not have parents with more than 12 years of education,and (5) were not able to return to school before 1978 after they had entered the labor force.125Table 4.2: Ordered Logit Regression Results: the Impact of the Send-downExperience on Life HappinessMale Female(1) (2) (3) (4) (5) (6)Send-down -.005 -.005∗∗ -.005∗∗ -.004∗ -.004∗ -.005∗( 0.003) (0.002) (0.002) (0.003) (0.003) (0.003)Education 0.001 0.002 0.001 0.002 0.003 0.003(0.002) (0.003) (0.003) (0.003) (0.004) (0.003)Income 0.005∗∗ 0.004∗∗ 0.005∗∗ 0.007∗∗∗ 0.006∗∗∗ 0.006∗∗∗(0.002) (0.001) (0.002) (0.002) (0.001) (0.001)Communist Party 0.004∗ 0.003∗ 0.003 0.004∗ 0.003∗ 0.003∗∗(0.002) (0.001) (0.002) (0.002) (0.002) (0.002)Age 0.001∗ 0.001∗∗ 0.001∗∗ 0.0003 0.0004 0.0004(0.0004) (0.0003) (0.0003) (0.002) (0.008) (0.009)Detailed Characteristics Y Y Y YFamily Backgrounds Y YObs. 641 615 615 581 553 553Note: The table shows ordered logit regressions on the dependent variable, life happiness, scaledfrom 1 to 5. Marginal effects are reported. Only the restricted samples are used. Detailed Charac-teristics includes household income, employment status, one digit occupation dummies, numberof children, age of the youngest child, presence of female children. Family Backgrounds includesthe following information of both parents when individuals were 18 years old: years of education,Communist Party membership status, an indicator of leadership in company or Communist Party,an indicator of owning private firms or working in private firms. Province dummies are included.Standard errors in parentheses are clustered at the province level. *** statistically significant at1% , ** statistically significant at 5%, * statistically significant at 10%.126Table 4.3: Probit Regression Results: Democracy Means GovernmentShould Make Decisions on Behalf of PeopleMale Female(1) (2) (3) (4) (5) (6)Send-down -.101∗∗ -.117∗∗ -.118∗∗ -.047 -.049 -.058(0.050) (0.055) (0.055) (0.045) (0.046) ( 0.045)Education -.053∗∗∗ -.049∗∗∗ -.047∗∗∗ -.033∗∗∗ -.028∗∗ -.029∗∗(0.011) (0.012) (0.043) ( 0.011) (0.045) (0.047)Income 0.018 0.028 0.025 0.048 0.063 0.052(0.025) (0.029) (0.028) (0.040) (0.040) ( 0.039)Communist Party -.083∗ -.046 -.041 -.104∗∗ -.049 -.054(0.047) ( 0.067) (0.057) (0.041) (0.044) (0.045)Age -.002 -.0001 -.0004 -.003 -.002 -.001(.005) (0.006) (0.006) (0.004) (0.006) (0.007)Detailed Characteristics Y Y Y YFamily Backgrounds Y YObs. 583 551 551 583 551 551Note: The dependent variable equals one if a respondent agrees with the statement “Democracymeans government should make decisions on behalf of people”, zero otherwise. Marginal effectsare reported. Only the restricted samples are used. Detailed Characteristics includes household in-come, employment status, one digit occupation dummies, number of children, age of the youngestchild, presence of female children. Family Backgrounds includes the following information ofboth parents when individuals were 18 years old: years of education, Communist Party member-ship status, an indicator of leadership in company or Communist Party, an indicator of owningprivate firms or working in private firms. Province dummies are included. Standard errors inparentheses are clustered at the province level. *** statistically significant at 1% , ** statisticallysignificant at 5%, * statistically significant at 10%.127Table 4.4: Probit Regression Results: the Impact of the Send-down Experi-ence on Communist Party Membership StatusMale Female(1) (2) (3) (4) (5) (6)Send-down -.108∗∗ -.108∗∗ -.093∗∗ -.007 -.008 -.006(0.045) (0.044) (0.047) (0.036) (0.036) (0.034)Education 0.101∗∗∗ 0.102∗∗∗ 0.091∗∗∗ 0.051∗∗∗ 0.051∗∗∗ 0.042∗∗∗(0.015) (0.016) (0.016) (0.007) (0.008) (0.008)Income 0.048 0.052 0.052∗ 0.049∗∗∗ 0.046∗∗∗ 0.0375∗∗∗(0.031) (0.032) (0.032) (0.016) (0.015) (0.014)Age 0.028∗∗∗ 0.028∗∗∗ 0.021∗∗∗ 0.014∗∗∗ 0.014∗∗∗ 0.012∗∗(0.005) (0.005) (0.005) (0.004) (0.004) (0.004)Detailed Characteristics Y Y Y YFamily Backgrounds Y YObs. 640 640 612 573 572 527Note: The dependent variable equals one if a respondent is a Communist Party member, zerootherwise. Marginal effects are reported. Only the restricted samples are used. Detailed Charac-teristics includes household income, employment status, one digit occupation dummies, numberof children, age of the youngest child, presence of female children. Family Backgrounds includesthe following information of both parents when individuals were 18 years old: years of education,Communist Party membership status, an indicator of leadership in company or Communist Party,an indicator of owning private firms or working in private firms. Province dummies are included.Standard errors in parentheses are clustered at the province level. *** statistically significant at1% , ** statistically significant at 5%, * statistically significant at 10%.128Table 4.5: The Impact of the Send-down Experience on Other AttitudesDependent VariableOrdinary People’s Decision Power Voting TrustMaleSend-down -.064 -.128 0.015(0.067) (0.198) (0.019)Obs. 491 564 615FemaleSend-down 0.006 -.042 0.002(0.057) (0.047) (0.028)Obs. 470 473 553Note: Only Males sample are used. The dependent variable in column 1 equals one if a respondentagree with the statement “It will be democracy only when ordinary people have direct voices anddecision power on important state and local matters”, zero otherwise. The dependent variable incolumn 2 equals one if a respondent agree with the statement “It is also democracy if ordinary peo-ple have rights to vote for their own representatives to discuss important state and local matters”,zero otherwise. The dependent variable in column 3 is a measure of “ Trust against strangers”,scaled 15. Marginal effects are reported. Only the restricted samples are used. Education, Income,Communist Party member status, age, province dummies, as well as the Detailed Characteristicsand the Family Backgrounds are included. Standard errors in parentheses are clustered at theprovince level. *** statistically significant at 1% , ** statistically significant at 5%, * statisticallysignificant at 10%.129Table 4.6: Intergenerational Impact of Send-down: Investment on Children’s EducationDescriptive Statistics: Sponsorship Fee PaidMale FemaleMean Standard Deviation Mean Standard DeviationAll ReasonsSend-down 2.99 8.60 2.01 6.26Non-send-down 1.73 4.50 1.66 4.69For Attending a Better SchoolSend-down 2.35 7.49 1.69 5.91Non-send-down 1.25 3.97 1.19 3.77Dependent Variable: Sponsorship Fee PaidMale (1)-(3) Female (4)-(6)All Better School Better School All Better School Better School(1) (2) (3) (4) (5) (6)Send-down 2.40∗∗ 2.18∗∗ 2.02∗ 0.05 0.15 -.08(1.03) (1.02) (1.19) (0.73) (0.66) (0.74)Send-down × EduUpgrade 0.96 1.44(3.52) (2.04)Edu Upgrade 1.75 0.82 0.65 0.98 -.12 -.57(1.08) (0.91) (0.65) (1.50) (0.95) (0.7)All Family Backgrounds Y Y Y Y Y YObs. 472 472 472 409 409 409Note: Unit of Sponsorship Fee is 1000 RMB. All regressions control for individual income, household income, number of children, ageof the youngest child, presence of female children, bargaining power, years of education, experience, years of communist party member,employment status, family backgrounds and province dummies. Only the restricted samples are used. Standard errors in parentheses areclustered at the province level. *** statistically significant at 1% , ** statistically significant at 5%, * statistically significant at 10%.130Table 4.7: Robustness CheckDependent VariablesDemocracy by Community Party Sponsorship FeeHappiness Government Membership Status (Better School)(1) (2) (3) (4)Panel A. Family Connection ControlledSend-down -.005∗ -.115∗∗ -.091∗ 2.04∗∗(0.003) (0.050) (0.048) (0.990)Obs. 615 536 612 464Panel B. Early Return DroppedSend-down -.008∗∗ -.100∗ -.110∗∗ 2.32∗(0.004) (0.059) (0.044) (1.27)Obs. 575 519 572 439Panel C. Parents Working in Private Firms DroppedSend-down -.006∗ -.115∗∗ -.085 2.12∗∗(0.003) (0.051) (0.052) (1.03)Obs. 599 536 597 451Panel D. Parents Non-communist Party MemberSend-down -.006∗∗ -.104∗∗ -.162∗∗∗ 1.44∗(0.002) (0.050) (0.038) (0.74)Obs. 449 399 442 333Panel E. Parents with Junior High Education or BelowSend-down -.005∗ -.110∗∗ -.091∗ 1.88∗(0.003) (0.048) (0.049) (1.05)Obs. 563 495 560 428Panel F. Father in Non-government Sector OnlySend-down -.005∗ -.098∗∗ -.109∗∗ 1.88∗∗(0.003) (0.048) (0.044) (0.94)Obs. 592 527 588 444Panel G. Father Non-skilled White or Blue Color OccupationSend-down -.005∗ -.082∗ -.123∗∗ 2.11∗∗(0.003) (0.043) (0.051) (1.06)Obs. 489 428 487 367131Note: Marginal effects are reported in column 1-3. Only the restricted samples are used. Edu-cation, Income, age, province dummies, as well as the Detailed Characteristics and the FamilyBackgrounds are included. Column 1, 2, 4 controls for Communist Party member status. Standarderrors in parentheses are clustered at the province level. *** statistically significant at 1% , **statistically significant at 5%, * statistically significant at 10%.132Chapter 5ConclusionsThis thesis provides several interesting results about how policies introduced sev-eral decades ago could affect current individuals’ economic outcomes, well-beingsand political attitudes. It provides us with an important path to understand thecurrent economic outcomes, and maybe the origins of the social conflicts as well.Knowing the existence of the events and studying those events may not be so dif-ficult. However, it may require fundamental changes in order to utilize what wehave learned.133BibliographyAlbrecht, J. W., P.-A. Edin, M. Sundstrom, and S. B. Vroman (1999). CareerInterruptions and Subsequent Earnings: A Reexamination Using Swedish Data.Journal of Human Resources 34(2), 294–311. → pages 71Allen, F., J. Qian, and M. Qian (2005). Law, Finance, and Economic Growth inChina. Journal of Financial Economics 77(1), 57–116. → pages 5, 12Anderson, S. and D. R. Ray (2010). Missing Women: Age and Disease. Reviewof Economic Studies 77(4), 1262–1300. → pages 8Arnold, F. and Z. Liu (1986). Fertility, Sex Preference, and Family Planning inChina. Population and Development Review 12(2), 221–246. → pages 20Ayyagari, M., A. Demirguc-Kunt, and V. Maksimovic (2010). Formal VersusInformal Finance: Evidence from China. Review of Financial Studies 23(8),3048–3097. → pages 5, 12Banerjee, A., X. Meng, T. Porzio, and N. Qian (2013). Fertility Change andHousehold Savings in China: Evidence from a General Equilibrium Model andMicro Data. Working Paper. → pages 10, 17Banerjee, A., X. Meng, and N. Qian (2010). The Life Cycle Model and HouseholdSavings : Micro Evidence from Urban China. Working Paper. → pages 31, 68,73Bernstein, T. (1977). Up to the Mountains and Down to the Villages: The Transferof Youth from Urban to Rural China. Yale University Press. → pages 68, 70,72, 73, 74, 77Borghans, L., A. L. Duckworth, J. J. Heckman, and B. T. Weel (2008). TheEconomics and Psychology of Personality Traits. Journal of Human Re-sources 43(4), 972–1059. → pages 65134Bound, J. and S. Turner (2002). Going to War and Going to College: Did WorldWar II and the G. I. Bill Increase Educational Attainment for Returning Veter-ans? Journal of Labor Economics 20(4). → pages 71Cai, H., H. Li, A. Park, and L. Zhou (2013). Family Ties and OrganizationalDesign: Evidence from Chinese Private Firms. The Review of Economic Statis-tics 95, 850–867. → pages 5Card, D. and A. R. Cardoso (2011). Can Compulsory Military Service RaiseCivilian Wages? Evidence from the Peacetime Draft in Portugal. IZA Discus-sion Paper (5915). → pages 71Chamon, M., K. Liu, and E. Prasad (2010). Income Uncertainty and HouseholdSavings in China. Working Paper. → pages 6, 13Chamon, M. D. and E. S. Prasad (2010). Why Are Saving Rates of Urban House-holds in China Rising? American Economic Journal: Macroeconomics 2(1),93–130. → pages 6, 12, 13, 25, 31Chen, K. and X. Cheng (1999). Comment On Zhou & Hou: A Negative Life EventWith Positive Consequences? American Sociological Review 64(1), 37–40. →pages 70Chen, Y., Y. Ma, and K. Tang (2011). The Chinese Financial System at the Dawnof The 21st Century: An Overview. The IEB International Journal of Finance 2,2–41. → pages 5, 12Choukhmane, T., N. Coeurdacier, and K. Jin (2013). The One-Child Policy andHousehold Saings. LSE Working Papers. → pages 10Clark, A. E., P. Frijters, and M. A. Shields (2008). Relative Income, Happiness,Utility: An Explanation for the Easterlin Paradox and Other Puzzles. Journalof Economic Literature 46(1), 95–144. → pages 116Coeurdacier, N., S. Guibaud, and K. Jin (2013). Credit Constraints and Growth ina Global Economy. LSE Working Papers. → pages 12Cunha, F. and J. Heckman (2007, May). The Technology of Skill Formation.American Economic Review 97(2), 31–47. → pages 65135Earnings, L. and A. Records (1990). Lifetime Earnings and the Vietnam Era DraftLottery: Evidence from Social Security Administrative Records. AmericanEconomics Review 80(3), 313–336. → pages 71Ebenstein, A. (2010). The “Missing Girls” of China and the Unintended Con-sequences of the One Child Policy. The Journal OF Human Resource 45(1),87–115. → pages 8, 38, 39Escriche, L., G. Olcina, and R. Sanchez (2004). Gender Discrimination and In-tergenerational Transmission of Preferences. Oxford Economic Papers 56(3),485–511. → pages 37Estrin, S. and M. Prevezer (2011). The role of informal institutions in corporategovernance: Brazil, Russia, India, and China compared. Asia Pacific Journalof Management 28, 41–67. → pages 5Fafchamps, M. (2008). Risk Sharing Between Households. Handbook of SocialEconomics, 1255–1279. → pages 6Fafchamps, M. and A. R. Quisumbing (2008). Risk Sharing Between Households.Handbook of Development Economics, 3187–3247. → pages 6Frijters, P., J. P. Haisken-denew, and M. A. Shields (2013). Money Does Matter!Evidence from Increasing Real Income and Life Satisfaction in East GermanyFollowing Reunification. American Economic Review 94(3), 730–740.→ pages116Gan, L. (2012). Findings from the China Household Finance Survey. WorkingPaper. → pages 12Ge, S., D. T. Yang, and J. Zhang (2012). Population Policies, Demographic Struc-tural Changes, and the Chinese Household Saving Puzzle. IZA DiscussionPapers 7026. → pages 10, 31Giles, J., A. Park, and M. Wang (2008). The Great Proletarian Cultural Rev-olution, Disruptions to Education, and Returns to Schooling in Urban China.World Bank Policy Research Working Paper 4729. → pages 66, 79Gleibs, I. H., T. a. Morton, A. Rabinovich, S. A. Haslam, and J. F. Helliwell(2013, March). Unpacking the Hedonic Paradox: A Dynamic Analysis of theRelationships between Financial Capital, Social Capital and Life Satisfaction.The British Journal of Social Psychology 52(1), 25–43. → pages 115136Gordon, R. H. and W. Li (2003). Government as a Discriminating Monopolist inthe Financial Market: The Case of China. Journal of Public Economics 87(2),283–312. → pages 12Grosjean, P., F. Ricka, and C. Senik (2013, May). Learning, Political Attitudesand Crises: Lessons from Transition Countries. Journal of Comparative Eco-nomics 41(2), 490–505. → pages 117Guariglia, A., X. Liu, and L. Song (2011). Internal Finance and Growth: Mi-croeconometric Evidence on Chinese Firms. Journal of Development Eco-nomics 96(1), 79–94. → pages 5, 12Han, J., W. Suen, and J. Zhang (2011). Picking up the Losses: The Impact of theCultural Revolution on Human Capital Re-investment in Urban China. WorldBank Policy Research Working Paper 4729. → pages 66, 79, 80Heckman, J. J. and T. Kautz (2014). Fostering And Measuring Skills: Interven-tions That Improve Character And Condition. NBER Working Paper 19656.→pages 65Helliwell, J. F. and H. Huang (2008, July). How’s Your Government? Interna-tional Evidence Linking Good Government and Well-Being. British Journal ofPolitical Science 38(04), 595–619. → pages 115Helliwell, J. F. and S. Wang (2012). The State of World Happiness. World Hap-piness Report, pp. 10–57. → pages 115Horioka, C. Y. and J. Wan (2007). The Determinants of Household Saving inChina: A Dynamic Panel Analysis of Provincial Data. Journal of Money, Creditand Banking 39(8), 2077–2096. → pages 11Imbens, G. and W. an der Klaauw (1995). Evaluating the Cost of Conscriptionin the Netherlands. Journal of Business & Economic Statistics 13(2), 207–215.→ pages 71Jacobsen, R., H. Moller, and A. Mouritsen (1999). Natural Variation in the humanSex Ratio. Human Reproduction 14(12), 3120–5. → pages 22Joshua, D., S. H. Chen, J. S. Long, B. J. D. Angrist, and J. Song (2011). Long-term Consequences of Vietnam-Era Conscription: New Estimates Using SocialSecurity Data. American Economic Review 101(3), 334–38. → pages 71137Lardy, N. (2012). Interest Rate Liberalization and the International Role of theRMB. Working Paper. → pages 12Lee, Y. and Z. Xiao (1998). Children’s Support for Elderly Parents in Urbanand Rural China: Results from a National Survey. Journal of Cross-CulturalGerontology 13, 39–62. → pages 31Lemieux, T. and D. Card (2001, May). Education, Earnings, and the Canadian‘G.I. Bill’. Canadian Journal of Economics 34(2), 313–344. → pages 71Li, H., M. Rosenzweig, and J. Zhang (2010). Altruism, Favoritism, and Guiltin the Allocation of Family Resources: Sophie’s Choice in Mao’s Mass Send-Down Movement. Journal of Political Economy 118(1), 1–38. → pages 70, 74,94Li, X. (2000). Where Do We Go during 50 Years?– A Review on ChineseWomen’s Liberation and Development Process. Zhejiang Academic Journal 2.→ pages 21Liu, X. (2012). Shi Nian Zhi Qing Yi Sheng Cai Fu . West China Develop-ment (04). → pages 67, 75, 96, 149Ma, Y. (2007). From Order to Fairness: Guideline of Allevaiting the Burdenof Charge for Higher School Selection. China Education Economics annualmeeting, papers and proceedings, 544–549. → pages 121Meng, X. and R. G. Gregory (2002, July). The Impact of Interrupted Educationon Subsequent Educational Attainment: A Cost of the Chinese Cultural Revo-lution. Economic Development and Cultural Change 50(4), 935–959. → pages66, 76, 79Modigliani, F. and S. L. Cao (2004). The Chinese Saving Puzzle And Life-CycleHypothesis. Journal of Economic Literature 42(1), 145–170. → pages 11Morduch, J. (1995). Income Smoothing and Consumption Smoothing. Journal ofEconomic Perspectives 9(3), 103–114. → pages 35Nunn, B. N. and L. Wantchekon (2011). The Slave Trade and the Origins ofMistrust in Africa. American Economic Review 101, 3221–3252. → pages 119138Pan, Y. (2002). An Examination of the Goals of the Rustication Program in thePeople’s Republic of China. Journal of Contemporary China 11(31), 361–379.→ pages 68, 72, 73, 74, 77, 97Piketty, T. (1995). Social Mobility and Redistributive Politics. Quarterly Journalof Economics CX(3), 551–584. → pages 116Qian, N. (2008). Missing Women and the Price of Tea in China: The Effectof Sex-Specific Earnings on Sex Imbalance. The Quarterly Journal of Eco-nomics 123(3), 1251–1285. → pages 8Rodrguez-Planas, N. (2012). Longer-term impacts of mentoring, educational ser-vices, and learning incentives: Evidence from a randomized trial in the UnitedStates. American Economic Journal: Applied Economics 4, 121–139. → pages65Scharping, T. (2003). Birth Control in China 19492000. New York, RoutledgeCurzon. → pages 39Schochet, Peter Z., J. B. and S. McConnell (2008). Does Job Corps work? ImpactFindings from the National Job Corps Study. American Economic Review 98,1864–1886. → pages 65Singer, M. (1971). Educated Youth and the Cultural Revolution in China. Michi-gan Papers in Chinese Studies 10. → pages 68, 70, 73, 74, 77Song, Z., K. Storesletten, and F. Zilibotti (2011). Growing Like China. AmericanEconomics Review 101, 196–233. → pages 5, 12Stevenson, B. (2011). Trust in Public Instituions over the Business Cycle. NBERWorking Paper 16891. → pages 117Tang, Z. (2012). “Zhi Qing” Shi Bi Cai Fu . Business Weekly (12). → pages 67,75, 96, 149Tsang, M. C. (2001). School Choice In The People’s Republic Of China.Teacher’s College Columbia University Working Paper (June). → pages 120Unger, J. (1980). Bending the School Ladder : The Failure of Chinese EducationalReform in the 1960s. Comparative Education Review 24(2), 221–237.→ pages68, 70, 77139Wang, H. (2006). Zhi Qing Jing Li Rang Wo Jian Reng Bu Ba . China StateFarms (09). → pages 67, 75, 96, 149Wang, S.-Y. (2013). Marriage Networks,Nepotism, and Labor Market Outcomesin China. American Economic Journal: Applied Economics 5(3), 91–112. →pages 93Wang, T. (2004). Ge Di Qu Qiou Ji Gao Zhong Ze Xiao Fei Shou Qu Qing Kuangde Diao Cha. China Statistics, 60. → pages 121Wei, S.-J. and X. Zhang (2011). The Competitive Saving Motive: Evidence fromRising Sex Ratios and Savings Rates in China. Journal of Political Econ-omy 119(3), 511–564. → pages 10, 20, 25, 34Xi, J. (2003). Fu Jian Bo Shi Fen Cai, Volumn 1. Hai Chao She Ying Yi Shu ChuBan She 1. → pages 75Xie, Y., Y. Jiang, and E. Greenman (2008). Did Send-down Experience Bene-fit Youth? A Reevaluation of the Social Consequences of Forced Urban-ruralMigration during China’s Cultural Revolution. Social Science Research 37(2),686–700. → pages 68, 70, 71, 77Yang, F. (1992). Hui Cheng Zhi Qing Sheng Cun Xian Zhuang Pou Xi . ChinaState Farms (01). → pages 67, 75, 96, 149Yao, R., F. Wang, R. O. Weagley, and L. Liao (2011). Household Saving Motives:Comparing American and Chinese Consumers. Family and Consumer SciencesResearch Journal 40(1), 28–44. → pages 12Yip, W. and W. C. Hsiao (2008). The Chinese Health System at a Crossroads.Health affairs (Project Hope) 27(2), 460–8. → pages 31Yu, L. C., Y. Yu, and P. K. Mansfield (1990). Gender and Changes in Supportof Parents in China: Implications for the One-Child Policy. Gender and Soci-ety 4(1), 83–89. → pages 31Zhang, X. (2000). Children of the Cultural Revolution: Family Life and PoliticalBehavior in Mao’s China. Boulder, Colo. : Westview Press, 200. → pages 73Zhou, W. (2013a). Brothers, Household Financial Markets and China’s SavingRate. Working Paper. → pages 68, 73140Zhou, W. (2013b). How Does A Traumatic Experience During Youth Affect LaterLife?The Long Term Impact of the Send-down Program during the Chinese CulturalRevolution. Working Paper. → pages 25, 123Zhou, X. and L. Hou (1999). Children of the Cultural Revolution: The Stateand the Life Course in the People’s Republic of China. American SociologicalReview 64(1), 12–36. → pages 65, 68, 70, 74, 77, 94, 115141Appendix AAppendix for Chapter 2A.1 DataThe primary data source for this study is the China General Social Survey (CGSS)2006 urban areas sample. It is an individual-level cross-sectional dataset. Thedata were collected jointly by the Sociology Department of People’s University ofChina and the Hong Kong University of Science and Technology Survey ResearchCenter. It covered 24 provinces and 4 municipalities. Only three autonomousprovinces were not included in the survey: Tibet, Qinghai, and Ninxia. The surveywas conducted based on a probabilistic sample. The stratification design wasbased on the 2000 population census.According to the CGSS documentation, the survey only asked one randomlyselected household member, between 19 and 70 years old, to answer all the ques-tions. I dropped all students from the CGSS sample. Among urban area residentswho born between 1945–1978 (aged 28–60 in 2006), 91% of respondents aremarried and 7.3% of respondents were living alone. In the following cases, re-spondents may not be counted as household head: a respondent is living with asibling or with working parents under 60 years old. Fortunately, only 1.4% ofrespondents are living with siblings. This suggests that brothers are most likelyto be the members of extended families of respondents. Furthermore, only 0.3%of the respondents live with a working parent under age 60. The estimation re-sults remain essentially unchanged when these 1.7% of respondents are excludedfrom the sample. Situation in which a respondent lived with her/his uncle or auntmight also be of concern. Unfortunately, aunts and uncles were not part of thelist of possible relationships with the respondent that are queried in the CGSS142questionnaire. This might be due to the fact that, in urban areas, it is quite rarefor an individual to live with an uncle or aunt. Furthermore, since only 0.3% ofrespondents lived with working parents under 60, it is unlikely that any significantnumber of respondents lived with an uncle or aunt who is working and under 60years old.The basic summary statistics for all variables used in the regression are pre-sented in Appendix Table 2.Three other supplementary datasets are used in this chapter China FamilyPanel Study (CFPS), Chinese Household Income Project (CHIP) urban area sam-ple, and Chinese Health and Retirement Longitudinal Study (CHARLS). CFPSwas conducted by the Peking University Institute of Social Science survey in Bei-jing, Shanghai, and Guangdong province. This study was also based on a proba-bilistic sample and stratified design. It is currently available for the 2008 and 2009series. CHIP was conducted under the auspices of the Chinese Academy of So-cial Science. The sampling frame is a subsample of the official household surveyconducted by the National Bureau of Statistics (NBS). The 2002 CHIP survey isused in this study. CHARLS was conducted by the National School of Develop-ment (China Center for Economic Research) at Peking University. Currently, onlythe 2008 survey is available. The provincial-level data were primarily collectedfrom the China Statistical Year Book published by the NBS. The provincial-levelfinancial development data were collected from the Almanac of China’s Financeand Banking. The China Urban Labor Survey (CULS) was administered fromNovember 2001 to January 2002 in five large Chinese cities: Shanghai, Shenyang,Wuhan, Xian, and Fuzhou. The survey was administered by the Institute for Pop-ulation Studies at the Chinese Academy of Social Sciences (CASS-IPS), in col-laboration with local offices of the NSB in each of the five cities.A.2 Proof of the Identification StrategyIn this appendix, I show that under the assumption that εi is conditional indepen-dence of number of brothers given number of siblings; that is,εi⊥broi|sibi.α can be consistently estimated in the following equation.(For simplicity I ignoreother controls.)Yi = αbroi +δ (sibi)+ εi (A.1)143where δ (sibi) is a function of sibi.Proof:Use the definition of conditional independence, we havef (εi|sibi,broi) =f (εibroi|sibi)f (broi|sibi)=f (εi|sibi) f (broi|sibi)f (broi|sibi)= f (εi|sibi)where f (·) is the density function. Thus,E(εi|sibi,broi) =∫εiε f (εi|sibi,broi)dεi=∫εiε f (εi|sibi)dεi= E(εi|sibi)Since E(εi|sibi) is a function of sibi, letδ˜ (sibi) = E(εi|sibi)where δ˜ (sibi) is an unknown function of sibi.AssumeYi = αbroi +β sibi + εiSince E(εi|sibi,broi) is not depend on broi, we haveE(Yi|broi,sibi) = αbroi +β sibi +E(εi|sibi,broi)= αbroi +β sibi + δ˜ (sibi)Thus, α can be consistently estimated under equation A.1, where δ (sibi) =144β sibi + δ˜ (sibi). δ (sibi) is a control function, in order to consistently estimate α .A.3 Relative Effect of Number of BrothersIn this section, I show that if sisters have an effect on savings rate, I can still havethe difference of the effect between brothers and sisters. Suppose we are interestedin estimating equation A.2. (For simplicity I ignore other controls).Yi = αbbroi +αssisi + εi (A.2)where sisi is the number of sisters.αb and αs cannot be consistently estimated because broi and sisi are correlatedwith the error term εi. For this reason, we use the control function approachexplained in appendix C by adding a function of sibi into equation A.2. We canhaveE(Yi|broi,sisi,sibi) = αbbroi +αssisi +δ (sibi)Due to collinearity, αb and αs cannot be estimated together. However,E(Yi|broi,sibi,sisi) = αbbroi +αs(sibi−broi)+δ (sibi)= (αb−αs)broi +δ ′(sibi)where δ ′(sibi) = δ (sibi)+αssibi Thus we can still identify the effect of brothersrelative to sisters which is αb−αs.145Table A.1: Household Expenditure and Total IncomeAge Group Living Cost Education Expenditure Medical Expenditure Disposable Income25-30 12219 541 660 3385930-35 11950 1324 824 3097635-40 11790 2203 1119 3004040-45 10745 3830 885 2495345-50 12144 4218 1109 2626150-55 11429 2267 1539 2489155-60 12043 882 1755 27360Note: Chinese RMB is presented in the table. Exchange rate in 2006: 1 US Dollar = 7.97 RMB.China General Social Survey 2006 is used. Sample is restricted to urban area residents bornbetween 1946-1978.146Table A.2: Summary StatisticsVariable Obs Mean Std. Dev. Min Max UnitSavings Rate 2634 0.260 0.545 -5.000 0.947Number of Brothers 2634 1.432 1.188 0 8Number of Siblings 2634 2.823 1.745 0 9Year of Education 2581 10.244 3.057 1 22Age 2634 43.992 8.925 29 60Household Yearly Income 2634 0.285 0.332 0.009 6 100,000 RMBMarital Status 2634 0.908 0.289 0 1Female 2634 0.527 0.499 0 1Mother’s Years of Education 2609 4.746 3.620 1 17Father’s Years of Education 2597 6.447 3.810 1 17Family Size 2634 2.867 1.039 1 9Communist Party 2634 0.126 0.332 0 1Send-down 2634 0.113 0.317 0 1Number of Parents Deceased 2608 0.809 0.837 0 2Presence of Male Children 2634 0.528 0.499 0 1Parents Live Together 2634 0.161 0.368 0 1Children Age <6 2634 0.097 0.296 0 1Children Age 6−12 2634 0.169 0.374 0 1Children Age 12−18 2634 0.179 0.384 0 1Children Age >18 2634 0.116 0.320 0 1Number of Children 2634 1.070 0.665 0 6No House 2634 0.306 0.461 0 1Value of Mortgage 2616 0.013 0.130 0 4.9 100,000 RMBValue of Other Houses 2612 0.124 0.772 0 20 100,000 RMBFather’s Huko 2634 0.696 0.460 0 1Mother’s Huko 2634 0.647 0.478 0 1Mother’s Company Owner Ship 2634 0.552 0.839 0 2Father’s Company Owner Ship 2634 1.099 0.941 0 2Mother’s Skill Level 2631 0.291 0.884 0 4Father’s Skill Level 2632 0.857 1.457 0 4Father’s Communist Party 2634 0.139 0.346 0 1Mother’s Communist Party 2634 0.031 0.174 0 1Provincial Level Data:Number of the Branchesof Foreign Banks 28 0.360 0.761 0 3.093Insurance Density 28 7.171 9.157 1.030 32.930 100 RMB / Person147Table A.3: Statistics Used for Calculating Increased Savings Rate due toDecreased Number of BrothersAge Group22-39 40-49 50-60̂broIncl,A -0.069 -0.116 -0.155̂broDPl,A 0.037 0.025 0.057Average Income (100,000RMB) 0.12 0.11 0.12Distribution of Each Group 0.18 0.17 0.16Marriage Rate 0.78 0.87 0.85Decreased Number of Brothers 0.71 0.32 0.45Note: The statistics of low income households are presented.148Appendix BAppendix for Chapter 3B.1 DataThe main data used in this paper are the China General Social Survey (CGSS)2003. The CGSS 2003 data are also part of the East Asian General Social Sur-vey. The data were collected jointly by the Hong Kong University of Scienceand Technology Survey Research Center and the Sociology Department of Peo-ple’s University of China. CGSS 2003 was an individual level survey and wasconducted in city areas. It covered 24 provinces and four municipalities. Onlythree autonomous provinces were not included in the survey: Tibet, Qinghai, andNinxia.1 The survey was conducted based on a probabilistic sample and stratifieddesign.B.2 Conceptual FrameworkUpgrading education is a time- and energy-consuming endeavor for individualswho are already 30 years old. Such individuals often have a job during the dayand a family to take care of at home. However, as suggested by numerous doc-uments, such as Yang (1992); Wang (2006); Liu (2012); Tang (2012), the send-down experience improved their capability to bear such hardships. Therefore, forthe send-downs, exerting effort toward upgrading education is not as costly asit might be among non-send-downs. Because of the low cost of exerting effort,1Qinghai is a province next to Tibet. Ninxia is another minority province located in inlandChina. The 2003 survey was conducted in October and November.149send-downs are more likely to upgrade education and exert more effort towardfurther studies. If the return to education depends on the effort put into study, wewould also find that send-downs who upgraded education would on average earnhigher incomes compared to non-send-downs who also upgraded their education.The following simple model illustrates the above idea.ei is the effort an individual i put into study when upgrading their education.ei is non-negative; it equals 0 if individual i choose not to upgrade their education.An individual chooses the level of effort to maximize his/her utility.maxeiw(ei,ai)−C(ei) (B.1)The wage function w(·) depends on an individual’s effort in study and his abilityai ∈A, where A is the space of ability. Both send-downs and non-send-downs drawa from same distribution F(.). The wage function satisfies properties wa > 0,we >0,wea > 0. The last condition indicates that the return to effort is increasing inability. There is a trade-off in exerting effort: exerting effort towards studying canincrease wages; however, exerting such effort is costly. Denote the cost function asC(ei) of effort. This satisfies the condition Ce(·) > 0 and Cee(·) > 0. For the sent-down group, providing additional effort is less costly. A simple cost function forsend-downs could be C(ei)− θei with θ > 0. For simplicity, the wage functiondoes not depend on experience. (We can think of this is as a case in which wecompare individuals with identical years of experience.)The first-order condition of Equation B.1 iswe(ei,ai) = Ce(ei) ∀ai ∈ A (B.2)Let a∗NS where a∗S denotes the ability of the marginal individual who is indif-ferent to upgrade education for the non-send-down group, NS and the send-downgroup, S respectively. This satisfieswe(0,a∗NS) = Ce(0)we(0,a∗S) = Ce(0)−θWe have a∗S < a∗NS, since we(0,a∗S) < we(0,a∗NS) and wei,ai > 0. Thus, forany increasing CDF of a, F(.)1−F(a∗S) > 1−F(a∗NS)150That is, more people in the send-down group upgraded their education.Denote the solution of the first-order condition as eSi (ai) for send-down groupand eNSi (ai) for non-send-down group. Combine first-order conditions with theassumption that send-downs have lower marginal cost of effort, we havewei(eNSi (ai),ai) > wei(eSi (ai),ai) ∀ai > a∗SThus, for a given ability, send-downs earn higher income than non-send-downsw(eNSi (ai),ai) < w(eSi (ai),ai) ∀ai > a∗SFor individuals who do not upgrade education, their incomes are same (given sameyears of experience).151

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0167500/manifest

Comment

Related Items