One-ChildPolicy,FertilityandGrowth

按一下以編輯母片標題樣式,按一下以編輯母片,第二層,第三層,第四層,第五層,*,*,按一下以編輯母片標題樣式,按一下以編輯母片,第二層,第三層,第四層,第五層,*,*,One-Child Policy, Fertility and Growth,Hongbin Li,Junsen Zhang,The Chinese University of Hong Kong,2024/9/10,1,Paper 1:How Effective is Chinas One-Child Policy?,Hongbin Li* Junsen Zhang*,Yi Zhu*,* The Chinese University of Hong Kong,* Michigan State University,2024/9/10,2,Chinas one-child policy,Unique one-child policy (starting from1979),one of the largest social experiments in human history, involving over one billion people,a counter-natal policy rarely seen in other countries,Decline of fertility since 1970s,total fertility drops from 6 before 1970 to 2.75 by 1980 and less than 2 since 1992 (Peng, 1996),How large is the effect of the one-child policy on fertility?,2024/9/10,3,Literature,Most are indirect tests: relate variation in fertility to variation in the implementation of policy across localities,fines for above-quota births (McElroy and Yang, 2000, etc.),one-child subsidy (Short and Zhai, 1998, etc.),provision of contraceptives (Johnson, 1994, etc.),Problems,local variations may be endogenous: higher fertility leads to more strict implementation,local data may not be representative,None directly measure the policy effect,2024/9/10,4,Objectives,Directly measure the effect of one-child policy on fertility (probability of having a second child),Test the variation of effect across sub-groups,rural vs. urban,less-educated vs. better-educated mother,Others,Empirical approach: Differences-in-differences based on a unique aspect of the one-child policy,2024/9/10,5,Institutional Background (1),Han,Chinese (more than 90% of the population),national policy: each woman allowed one child,local policy: method for implementing national policy,e.g., fines for above-quota births,local officials may be demoted for allowing too many above-quota births,Minorities,two children allowed in most regions through 1980s,even less restrictive for some small ethnic groups,same policy as,Han,for groups with population of more than 10 million (only,Zhuang,qualified in the end of 1980s),2024/9/10,6,Institutional Background (2),Local policy: urban vs. rural,urban: more strictly enforced,severe punishment for above-quota births (e.g., wage cut, ineligible for promotion in state-owned enterprises),rural: less strictly implemented,fine is the primary penalty, but not very effective for the poor (Li and Zhang, 2005),large variation in policy (fine) across localities,in some areas and in certain years, a second child is allowed if the first one is a girl,2024/9/10,7,Identification,Affirmative national policy,One-child only applies to,Han,Chinese,one child allowed for each,Han,woman,Minorities not subject to the one-child quota,allowed to have two children (some groups even more),Differences-in-Differences (DD),treatment:,Han,Chinese,control: ethnic minorities,no need to rely on local policies,2024/9/10,8,DD Strategy,DD Estimation,Y: having a second child,H:,Han,dummy,T: treatment dummy,3,: DD estimate,2024/9/10,9,Data,Chinese Population Census (1% sample),2 rounds: 1982 and 1990,representative national data: cover all provinces,Sample,women aged 20-64,household head or spouse,Han,Chinese: 94% (1982) and 93% (1990),2024/9/10,10,Pre-treatment group,How to determine the timing of treatment?,no simple distinction between pre- and post-treatment,treatment is a matter of degree that decreases with a womans age (women of younger birth cohorts are more likely to be affected by the policy),Pre-treatment group,women who already had the second child if they wanted and were able to do so by 1979 (thus not affected by one-child policy),determine the cutoff age (see following figures),2024/9/10,11,Pre-treatment group,Stable for 1935-45 cohorts at 96-97% (close to the biological limit),Stable for 1950 and earlier cohorts between 1982 and 1990,2024/9/10,12,Pre-treatment group,Women who were 37 or older (1945 and earlier cohorts in 1982) should have had the second child if they wanted and were able to,It is safe to set the cutoff age at 37 in 1979, which means 1942 and earlier cohorts are our pre-treatment group,2024/9/10,13,Results: Full Sample,Estimate for women of each birth cohort,Significant DD estimates for most cohorts,Robust to household and geographic controls,2024/9/10,14,2024/9/10,15,2024/9/10,16,2024/9/10,17,Results: Rural vs. Urban,Larger DD estimate in urban areas,Average effect: -7.5% (rural) vs. -16.8% (urban),2024/9/10,18,Results: Education Level,Smaller DD estimate for better-educated women,Average effect: -0.8% (illiterate) vs. -15.5% (high education),2024/9/10,19,Results: Education and Area,Variation of effect across education levels is smaller in urban areas,2024/9/10,20,Sensitivity Tests,Can DD estimate pick up other socioeconomic factors in the same period?,Test: use having a first child and being married as dependent variables, which may reflect parental preferences but are not affected by the one-child policy,Results: very small estimates, only among young women,Conclusion: DD method is not picking up the change of the,Han,-minority gap in fertility preference or other determinants of fertility; it is mainly the policy effect,2024/9/10,21,Results: Full Sample,Estimate for women of each birth cohort,Significant DD estimates for most cohorts,Robust to household and geographic controls,2024/9/10,22,A summary,Main findings,a large effect of one-child policy on fertility,larger effects for urban women and better-educated women DD effect is not very likely to be driven by other policy or socioeconomic changes,Implications,a new identification approach based on the affirmative policy,can be used to break endogeneity in fertility-related research,Fertility and labor supply (Li and Zhang, 2006),Fertility and marriage stability (Li, Zhang and Zhu, 2006),2024/9/10,23,Paper 2:Do High Birth Rates Hamper Economic Growth?,Forthcoming Review of Economics and Statistics,Hongbin LiJunsen Zhang,The Chinese University of Hong Kong,2024/9/10,24,The Malthus debate,Theoretical debate,Malthusian school: population hampers growth due to limited resources,Boserupian school: population is good, or at least neutral,Scale effect, endogenous technological progress,Empirical tests,Go either way (negative, or non-negative)no definite conclusion (Kelley, 1988; Temple, 1999),2024/9/10,25,Hard to establish causality,Population growth/birth rate in a growth regression is endogenous,Simultaneity,Income has a negative effect on fertility: quantity-quality tradeoff (Becker and Lewis, 1973; Barro and Becker, 1989; Wang et al., 1994),Income raises the real wage of women and leads to lower fertility (Galor and Weil, 1992),Omitted variable bias,Hard to control all variables or come up with a valid IV using cross-country data (Mankiw et al., 1992),2024/9/10,26,Our contributions,Establish causality by using Chinas unique population control policy as an instrument,Understand and evaluate how much Chinas one-child policy has contributed to growth since 1978,2024/9/10,27,Chinas fertility policies,National policies,The one-child policy,The affirmative policy: minorities are allowed to have two or more children,Community policies,Need birth permits,For above-quota births,Fines, no public school and other benefits,Government officials be demoted,Providing birth control facilities,2024/9/10,28,Empirical tests,Growth = a + b*BR + x*c + e,BR: birth rate,x: other determinants of growth,BR is endogenous,Reverse causality,Omitted variables,2024/9/10,29,Identification strategy,Drawing on provincial-level data from China,Using % of minority population in a province as IV for BR,It is highly correlated with birth rate,It should not be correlated with growth if other determinants of growth are controlled for,2024/9/10,30,Data,Provincial-level data for the period 1978-1998,Divide into 4 5-year intervals: so, it is a panel of 4 periods,Data sources: various statistical yearbooks,2024/9/10,31,One more econometric issue,When we do the fixed effects estimations, we will introduce serial correlation because the lagged dependent variable will be on the RHS of the growth equation,2024/9/10,32,Econometric method: GMM,The GMM method developed by Arellano and Bond (1991), first used in growth regressions by Caselli et al. (1996),Can deal with the issue of lagged dependent variable, allowing additional IVs for BR,Steps,Take first difference to eliminate the fixed effects,Apply the IV estimations to the first differences,Two ways: DIF or SYS,2024/9/10,33,Results: “first stage”,First, we examine whether our main IV is good,In the first differenced BR equation, the proportion of minority population is highly correlated with BR,Population-Growth_Tables.pdf,2024/9/10,34,GMM results,High birth rates hamper economic growth,The result is robust when we control for other determinants of growth,The contribution of the one-child-policy in Chinas growth,Raises growth by 1 percentage point a year,Raise the steady stage real per capita GDP level by 14.3 percentage points,Population-Growth_Tables.pdf,2024/9/10,35,Conclusions,We use a unique aspect of Chinas population policy to identify,the effect of the one-child policy on fertility,The causal effect of population on growth,We find that the one-child policy indeed,has a large effect on fertility,has contributed to Chinas growth,2024/9/10,36,