The Impacts of the Gender Imbalance on Marriage and Birth: Evidence from World War II in Japan
TThe Impacts of the Gender Imbalance on Marriage and Birth:Evidence from World War II in Japan ∗ Kota Ogasawara and Erika IgarashiFebruary 2, 2021
Abstract
This study uses the unprecedented changes in the sex ratio due to the lossesof men during World War II to identify the impacts of the gender imbalance onmarriage market and birth outcomes in Japan. Using newly digitized census-basedhistorical statistics, we find evidence that men had a stronger bargaining positionin the marriage market and intra-household fertility decisions than women. Underrelative male scarcity, while people, especially younger people, were more likely tomarry and divorce, widowed women were less likely to remarry than widowed men.We also find that women’s bargaining position in the marriage market might nothave improved throughout the 1950s. Given the institutional changes in the abortionlaw after the war, marital fertility and stillbirth rates increased in the areas thatsuffered relative male scarcity. Our result on out-of-wedlock births indicates thatthe theoretical prediction of intra-household bargaining is considered to be robustin an economy in which marital fertility is dominant.
Keywords:
Fertility; Gender imbalance; Marriage market; Second World War; Sexratio;
JEL Codes:
J11; J12; J13; J16; N30; N35; ∗ Kota Ogasawara: Department of Industrial Engineering, School of Engineering, Tokyo In-stitute of Technology, 2-12-1, Ookayama, Meguro-ku, Tokyo 152-8552, Japan (E-mail: [email protected]). Erika Igarashi: Department of Economics, The University of Tokyo, 7-3-1,Hongo, Bunkyo-ku, Tokyo 113-0033, Japan (E-mail: [email protected]). The workwas supported by the JSPS KAKENHI Grant Number 19K13754. There are no conflicts of interest todeclare. All errors are our own. a r X i v : . [ ec on . GN ] F e b Introduction
War causes enormous losses of people. These losses are generally concentrated on men,who are drafted into battle, rather than women, leading to a substantial reduction in thesex ratio in an economy (i.e., the ratio of men to women in the population). Theoretically,relative male scarcity improves the bargaining position of men in the marriage market andthus their intra-household allocations (Becker 1973, 1974, 1991; Chiappori et al. 2002). Agrowing body of empirical research has validated this prediction, providing solid evidencethat the wartime losses of men affect not only marriage market outcomes but also birthoutcomes both inside and outside of marriage (Abramitzky et al. 2011; Bethmann andKvasnicka 2012; Brainerd 2017). This study examines the impacts of relative male scarcity caused by World War IIon marriage market and birth outcomes in postwar Japan. Japan’s wartime losses ofmen led to an unprecedented decline in the sex ratio similar to in post-World War IFrance (Abramitzky et al. 2011), post-World War II Bavaria, Germany (Bethmann andKvasnicka 2012), and Russia (Brainerd 2017). However, in contrast to those countries,Japan experienced rapid democratization induced by the General Headquarters of theAllied Powers. The fertility rate had declined dramatically immediately after the war,whereas the share of out-of-wedlock births had remained only a few percent. Stillbirthrates, however, substantially increased in the mid-1950s due to the enactment of theMaternal Health Act. Considering these, in this study, we contribute to the literature byinvestigating the consequences of the wartime losses of men on marriage market and birthoutcomes in this unique context of postwar Japan.To do so, we newly digitize census-based historical statistics and apply the difference-in-differences estimation strategy using the exogenous variations in the sex ratio fromthe wartime losses of men. We find that while people who faced relative male scarcitywere more likely to marry and divorce, the gender differences in the estimates show men’sstronger bargaining position in the marriage market. Widowed women were less likely toremarry than widowed men, who faced better outside options. While the result of our See also Acemoglu and Autor (2004) for the impacts of war-induced changes in the female laborsupply on earnings inequalities. Another strand of the literature exploits different semi-experimentalapproaches. For example, Angrist (2002), Lafortune (2013), and Porter (2015) use exogenous changes inthe sex ratio due to the inflow of immigrants and famine to analyze the impacts on pre-marital investment,marriage and labor markets, consumption behavior, and the health status of offspring. Charles and Luoh(2010) analyze the impacts of male incarceration rates on women in the marriage market. Wei and Zhang(2011) and Edlund et al. (2013) employ the culture-induced gender imbalance to investigate the impactson saving behavior and crime rates in China, respectively. Regarding postwar Japan, Ogasawara and Komura (2021) analyzethe impacts on the fertility rate. To the best of our knowledge, Brainerd (2017) is thefirst study to analyze the influence of an unbalanced sex ratio on marriage market andbirth outcomes. Our empirical analysis builds on this approach as well as adds variablesnot considered by Brainerd (2017), including the proportion of single men and women,proportion of widowhoods, and average age at first marriage. In addition, to investigatethe potential gender heterogeneity in the effects of the wartime losses of men, this studyis the first to consider all the outcome variables by sex. Thus, our census-based datasetenables us to paint a broader picture of the marriage market and fertility decision after(and before) marriage in the aftermath of the Second World War. Specifically, the im-pacts of relative male scarcity on marriages among widowed women have thus far beenunderstudied. The military pension was partly abolished in 1946 by the General Head-quarters of the Allied Powers, which might have encouraged widowed women to remarryin postwar Japan. Our result on widowed women is generally consistent with that ofSalisbury (2017), who finds evidence that widowed women were less likely to marry underthe Civil War Pension Act of 1862 than beforehand.Third, this study is the first to provide empirical evidence in an economy that ex-perienced a unique trend of out-of-wedlock births. Previous studies show that relativemale scarcity due to war raised the share of out-of-wedlock births in the cases of France,Bavaria, and Russia (Abramitzky et al. 2011; Bethmann and Kvasnicka 2012; Brainerd2017). While these countries experienced an increase in the share of out-of-wedlock births,however, the share was considerably lower in postwar Japan. Despite this difference, wefind a result consistent with those of previous studies, implying that the theoretical pre-diction of intra-household bargaining is robust in an economy in which marital fertility isdominant (Becker 1991; Willis 1999).The remainder of this paper is organized as follows. Section 2 describes the genderimbalance, marriage market, and demographic trends after the war in Japan. Section 3 While both studies examine the economies after World Wars I and II, respectively, Bitler and Schmidt(2012) investigate the impacts of the draft during a more recent war, namely, the Vietnam Conflict, onbirth rates. Specifically, while Ogasawara and Komura (2021) focus on the birth rates obtained from the predictednumber of people, we consider both marital and out-of-wedlock birth rates calculated using census-basedstatistics. We also use a different identification strategy than theirs.
As in many countries, Japan lost a number of men during World War II. Statistics indicatethat 1,864,710 military personnel died or were missing in action during the war Nakamuraand Miyazaki (1995, p. 289). A survey conducted in May 1928 shows that 323,495homefront people died or were missing mainly due to bombing Nakamura and Miyazaki(1995, p. 277). To understand the magnitude and persistence of the wartime losses of men,we first digitize the number of men and women using population censuses conducted afterthe war. Figure 1 illustrates the national average sex ratios by age in 1947, 1950, 1955,and 1960. The distributions of these sex ratios are considerably different than those inthe prewar period (Figure B.1 in Online Appendix B.2 illustrates the sex ratios in 1930and 1935). This means that the clear reductions in the sex ratios in Figure 1 were causedby the wartime losses of men. We can highlight a few important features in those figures.First, Figure 1a indicates a clear and dramatic decline in the sex ratio soon after thewar. The figures decline from age 21 and bottom out around age 26 with approximately0 . − .
76) losses at their maximum. These relative declines in the sex ratio areobserved by age 40. Second, these declines persist until the 1950s. Figure 1b shows thatthe sex ratio bottoms out around age 29 with slightly more than 0 . − .
78) lossesat their maximum. Figures 1c and 1d confirm the roughly 0 . . This figure includes victims who died or were missing between 1942 and 1948. Deaths due to executionand diseases contracted on the frontline as well as those during the early stage of the Sino-Japanese warare not included Nakamura and Miyazaki (1995, p. 289). Thus, the overall death toll was greater thanthe figure reported herein. This finding is consistent with the fact that most of the repatriation had finished by 1947. In the 17months from May 1946 to September 1947, approximately 3,149,000 people (mostly men) were repatriated .700.720.740.760.780.800.820.840.860.880.900.920.940.960.981.001.021.041.061.081.10 M a l e s / F e m a l e s Age measured in 1947 (a) 1947 Population Census M a l e s / F e m a l e s Age measured in 1950 (b) 1950 Population Census M a l e s / F e m a l e s Age measured in 1955 (c) 1955 Population Census M a l e s / F e m a l e s Age measured in 1960 (d) 1960 Population Census
Figure 1: Sex ratios measured in the 1947, 1950, 1955, and 1960 PopulationCensuses
Notes: The sex ratio is defined as the number of men divided by the number of women. All the ratios are the nationalaverages based on the 1947, 1950, 1955, and 1960 Population Censuses. The vertical dotted lines show the minimum valuesof the sex ratios in each census year. Source: Created by the authors using Statistics Bureau of the Prime Minister’s Office(1948), Bureau of Statistics, Office of the Prime Minister (1951a), and Bureau of Statistics, Office of the Prime Minister(1956a). .0500.0550.0600.0650.0700.0750.0800.0850.0900.0950.1000.1050.1100.1150.1200.125 N u m be r o f d i v o r c e s pe r peop l e N u m be r o f m a rr i age s pe r peop l e Census Years
Marriage rates (left axis)Divorce rates (right axis) (a) Marriage and divorce rates A v e r age age a t fi r s t m a rr i age Census Years
FemaleMale (b) Age at first marriage N u m be r o f s t ill b i r t h s pe r , b i r t h s N u m be r o f li v e b i r t h s pe r , m a rr i ed w o m en Census Years
Marital fertility rates (left axis)Stillbirth birth rates (right axis) (c) Marital fertility and stillbirth rates N u m be r o f abo r t i on s o r m i sc a rr i age s (d) Miscarriages and abortions Figure 2: Marriage market and birth outcomes by census year
Notes: Figure 2a presents the marriage rate (number of marriages per 100 people) and divorce rate (numberof divorces per 100 people). Figure 2b presents average age at first marriage by gender. Figure 2c presents themarital fertility rate (number of live births per 1,000 married women) and stillbirth rate (number of stillbirthsper 1,000 births). Figure 2d presents the number of miscarriages and number of abortions. Source: Createdby the authors using Statistics Bureau of the Cabinet (1935), Statistics Bureau of the Cabinet (1939b),Bureau of Statistics, Office of the Prime Minister (1951b) Bureau of Statistics, Office of the Prime Minister(1956b), Statistics Bureau of the Cabinet (1931), Statistics Bureau of the Cabinet (1936), Division of Healthand Welfare Statistics, Welfare Minister’s Secretraiat (1953), and Division of Health and Welfare Statistics,Welfare Minister’s Secretraiat (1957).
Figure 2a illustrates the marriage and divorce rates in the census years, indicating aclear hump in 1950. A large number of people who could not marry during the warstarted to marry thereafter, creating a clear marriage boom in the early 1950s (Yuzawa1977). Correspondingly, average age at first marriage also decreases in 1950, as shown in to Japan (Statistics Bureau of the Prime Minister’s Office 1948, p. 1). This does not simply represent the decline in the population but rather the rises in the numberof marriages and divorces. The numbers of marriages and divorces in 1935 are 556,730 and 48,528,respectively, whereas those are 715,081 and 83,680 in 1950 and 714,861 and 75,267 in 1955 (Division ofHealth and Welfare Statistics, Welfare Minister’s Secretraiat 1953, 1957; Statistics Bureau of the Cabinet1936). N u m be r o f s i ng l e s pe r peop l e Age measured in 1950
FemaleMale (a) Singles in 1950 N u m be r o f s i ng l e s pe r peop l e Age measured in 1955
FemaleMale (b) Singles in 1955 N u m be r o f m a rr i ed peop l e pe r peop l e Age measured in 1950
FemaleMale (c) Married people in 1950 N u m be r o f m a rr i ed peop l e pe r peop l e Age measured in 1955
FemaleMale (d) Married people in 1955 N u m be r o f d i v o r c ed peop l e pe r peop l e Age measured in 1950
FemaleMale (e) Divorced people in 1950 N u m be r o f d i v o r c ed peop l e pe r peop l e Age measured in 1955
FemaleMale (f) Divorced people in 1955 N u m be r o f w i do w hood s pe r peop l e Age measured in 1950
FemaleMale (g) Widowhoods in 1950 N u m be r o f w i do w hood s pe r peop l e Age measured in 1955
FemaleMale (h) Widowhoods in 1955
Figure 3: Marriage status measured in the 1950 and 1955 Population Censuses
Notes: Figures 3a and 3b present the proportion of singles per 1,000 people (women and men). Figures 3cand 3d present the proportion of married people per 1,000 people (women and men). Figures 3e and 3fpresent the proportion of divorced people per 1,000 people (women and men). Figures 3g and 3h presentthe proportion of widowhoods per 1,000 people (women and men). All those rates are the national averagesbased on the 1950 and 1955 Population Censuses. Source: Created by the authors using Bureau of Statistics,Office of the Prime Minister (1951a) and Bureau of Statistics, Office of the Prime Minister (1956a). An explanation of theinstitutional context might also be useful. The Civil Code of 1947 abolished the patriarchyfamily system ( ie seido ) and, correspondingly, inheritance by new heads of households wasreplaced by the equal distribution of inheritance. The Code also allowed divorces becauseof the infidelity of husbands and division of properties at divorce. To see the postwar changes in the marriages and divorces, Figure 3 illustrates theproportion of singles, married people, divorced people, and widowhoods by census yearand age. Figures 3a and 3b indicate that the proportion of single women starts to declinearound age 18 and that most women (men) marry by age 40. Accordingly, Figures 3cand 3d show that a large proportion of women marry by age 30. Similar but slightly latertrends can be found for men in the same figures.An important trend in bargaining position in the marriage market can be observed indivorces and widowhoods. Figure 3e indicates that while there is no clear trend in theproportion of divorced men, the proportion of divorced women rises considerably betweenages 25 and 37 and peaks at age 29. This trend is consistent with the wartime losses ofmen relative to women shown in Figure 1b. In addition, Figure 3f presents the rightwardshift of the distribution, which moves the peak of the proportion of divorced women toaround 34 years old. This also corresponds to the losses of men in Figure 1c. Sucha trend suggests that a large proportion of divorced women remained in the marriagemarket, whereas men did not.A similar trend can be seen in the proportion of widowhoods (Figures 3g and 3h). To secure human resources, the government was proactive in encouraging people to marry. Hence, anumber of couples married hastily before the men were sent to the frontline, leading the poor matching(called kakekomi kon (hasty marriages)). Indeed, in prewar Japan, arranged marriages ( omiai-kon ) werealso used, which reduces the risk of mismatch because they were preceded by documentary examinationand organized meetings with the partner (Yuzawa 2005, pp. 190–192). We summarize the prewar institution and postwar reforms under the General Headquarters of theAllied Powers in Online Appendix A.1. See also Hayashi and Prescott (2008) for the economic impactsof the Old Civil Code. Under Family Laws within the Meiji Civil Code dating from 1898, although bigamy was forbiddenfor both husband and wife, adultery committed by a wife was recognized grounds for divorce, whileadultery on the husband’s part could only be a reason for divorce if he were found guilty of the crime ofillicit intercourse. The new Civil Code abolished this institution and established the equality of men andwomen upon divorce (Online Appendix A.1).
Ie-no-Hikari (light of a house) at that time claimed: “You (a widowed female) are not mostlikely to get remarried because there are a large number of women and a small numberof men in the marriage market (Kawaguch 2003, pp. 124–125). In our empirical analysis,we consider information on marriage status (single, married, divorced, and widowed) andmatching (age at first marriage) to investigate the impacts of the wartime losses of menon the marriage market.
Figure 2c illustrates the trends of marital fertility and stillbirth rates. While the maritalfertility rate shows a clear decreasing trend in the postwar period, the stillbirth rate showsthe opposite trend. Note, here, that stillbirths include losses due to not only miscarriagesbut also artificial abortions: Figure 2d indicates that roughly half of stillbirths are fromabortions. While abortion was basically forbidden in the prewar period, the establishmentof the Eugenic Protection Law of 1948 started to allow artificial abortions for economicreasons. An important fact here is that out-of-wedlock births were rare in postwar Japan.The share of out-of-wedlock births to total births was only a few percent in the 1950s(Section 5.2). This means that most abortions occurred within marriage. Indeed, a surveyconducted in 1964 by the Ministry of Welfare indicates that more than 40% of marriedwomen had experienced at least one abortion and roughly 80% of those women had twoor more children (Nakagawa 2000, p. 282). Given the poor knowledge of contraception,abortion was a frequently used means of reducing the number of births among families inthe 1950s (Norgren 2008).As discussed in Section 5, this historical context on births is different than thosein countries investigated by previous studies (Abramitzky et al. 2011; Bethmann andKvasnicka 2012; Brainerd 2017). Given this uniqueness, we consider marital fertility,stillbirth rates, and out-of-wedlock fertility to investigate the association between relative This official survey investigated 2 ,
547 randomly sampled married women aged 20–39 in Japan. Forthe details, see the Ministry of Welfare’s webpage on the Survey on Family Planning (https://survey.gov-online.go.jp/s39/S39-12-39-09.html [in Japanese], accessed January 25, 2021).
While statistics on the socioeconomic outcomes after the war are scarce as in other coun-tries, the 1950 and 1955 Population Censuses document prefecture-age-level informationon marriage market outcomes as well as the number of people in Japan (Bureau of Statis-tics, Office of the Prime Minister 1951a, 1956a). We digitize those to prepare for thedata on the sex ratio and several measures on marriage market outcomes in 1950 and1955. We digitize the data using 92 (46 editions for each census year) reports of the censuses in total. Toconserve space, we display those as one citation (say, Bureau of Statistics, Office of the Prime Minister(1951a) and Bureau of Statistics, Office of the Prime Minister (1956a)) for each census year throughoutthis paper. We cannot include Okinawa prefecture in our analytical sample because the island of Okinawaremained under the exclusive control of the American military until May 1972. a b l e : Su mm a r y S t a t i s t i c s : P o pu l a t i o n C e n s u s S t a t i s t i c s P a n e l A : P r e f ec t u r e - Y e a r - A g e - L e v e l S t a t i s t i c s U n i t O b s e r v a t i o n s M e a nS t d . D e v . M i n i m u m M a x i m u m P r o p o r t i o n o f s i n g l e s ( p e r , p e o p l e ) W o m e n P r e f ec t u r e - Y e a r - A g e . . . . M e n P r e f ec t u r e - Y e a r - A g e . . . P r o p o r t i o n o f m a rr i e dp e o p l e ( p e r , p e o p l e ) W o m e n P r e f ec t u r e - Y e a r - A g e . . . . M e n P r e f ec t u r e - Y e a r - A g e . . . . P r o p o r t i o n o f d i v o r ce dp e o p l e ( p e r , p e o p l e ) W o m e n P r e f ec t u r e - Y e a r - A g e . . . . M e n P r e f ec t u r e - Y e a r - A g e . . . . P r o p o r t i o n o f w i d o w h oo d s ( p e r , p e o p l e ) W o m e n P r e f ec t u r e - Y e a r - A g e . . . . M e n P r e f ec t u r e - Y e a r - A g e . . . . S e x r a t i o ( A d j u s t e d : E q u a t i o n ) P r e f ec t u r e - Y e a r - A g e . . . . P a n e l B : P r e f ec t u r e - Y e a r - L e v e l S t a t i s t i c s U n i t O b s e r v a t i o n s M e a nS t d . D e v . M i n i m u m M a x i m u m A v e r ag e ag e a t fi r s t m a rr i ag e W o m e n P r e f ec t u r e - Y e a r . . . . M e n P r e f ec t u r e - Y e a r . . . . M e n - W o m e n P r e f ec t u r e - Y e a r . . . . F e r t ili t y a nd s t ill b i r t h r a t e s M a r i t a l f e r t ili t y r a t e ( li v e b i r t h s p e r , m a rr i e d w o m e n ) P r e f ec t u r e - Y e a r . . . . S t ill b i r t h r a t e ( s t ill b i r t h s p e r , b i r t h ) P r e f ec t u r e - Y e a r . . . . S e x r a t i o ( M e n W o m e n ) P r e f ec t u r e - Y e a r . . . . F e m a l e l a b o r f o r ce p a r t i c i p a t i o n ( w o r k e r s p e r w o m e n ) P r e f ec t u r e - Y e a r . . . . P a n e l C : M a t e r n a l A g e - Y e a r - L e v e l S t a t i s t i c s U n i t O b s e r v a t i o n s M e a nS t d . D e v . M i n i m u m M a x i m u m O u t - o f - w e d l o c k b i r t h s h a r e ( li v e b i r t h s p e r li v e b i r t h s ) M a t e r n a l A g e - Y e a r . . . . O u t - o f - w e d l o c k b i r t h r a t e ( li v e b i r t h s p e r , w o m e n ) M a t e r n a l A g e - Y e a r . . . . S e x R a t i o ( A d j u s t e d : S a m e a s i n P a n e l A ) M a t e r n a l A g e - Y e a r . . . . N o t e s : N o t e s : P a n e l A r e p o r t s t h e s u mm a r y s t a t i s t i c s f o r t h e p r e f ec t u r e - y e a r - ag e - l e v e l p o pu l a t i o n ce n s u ss t a t i s t i c s o f nd . P a n e l B r e p o r t s t h e s u mm a r y s t a t i s t i c s f o r t h e p r e f ec t u r e - y e a r - l e v e l v i t a l a ndp o pu l a t i o n ce n s u ss t a t i s t i c s o f , , a nd . P a n e l C r e p o r t s t h e s u mm a r y s t a t i s t i c s f o r t h e m a t e r n a l ag e - y e a r - l e v e l p o pu l a t i o n ce n s u ss t a t i s t i c s o f nd . T h e nu m b e r o f p r e f ec t u r e s i s . M e a s u r e d y e a r s i n P a n e l s A a nd C a r e nd . M e a s u r e d y e a r s i n P a n e l B a r e , , a nd . A g e r a n g e i ss e tt o b e i n P a n e l s A a nd C . S o u r ce s : D a t a p r e s e n t e d i n P a n e l s A a nd C a r e f r o m B u r e a u o f S t a t i s t i c s , O ffi ce o f t h e P r i m e M i n i s t e r ( b ) ; B u r e a u o f S t a t i s t i c s , O ffi ce o f t h e P r i m e M i n i s t e r ( b ) , B u r e a u o f S t a t i s t i c s , O ffi ce o f t h e P r i m e M i n i s t e r ( ) , a nd B u r e a u o f S t a t i s t i c s , O ffi ce o f t h e P r i m e M i n i s t e r ( ) . D a t a p r e s e n t e d i n P a n e l B a r e f r o m S t a t i s t i c s B u r e a u o f t h e C a b i n e t( b ) , S t a t i s t i c s B u r e a u o f t h e C a b i n e t( ) , S t a t i s t i c s B u r e a u o f t h e C a b i n e t( ) , B u r e a u o f S t a t i s t i c s , O ffi ce o f t h e P r i m e M i n i s t e r ( b ) , B u r e a u o f S t a t i s t i c s , O ffi ce o f t h e P r i m e M i n i s t e r ( ) , D i v i s i o n o f H e a l t h a nd W e l f a r e S t a t i s t i c s , W e l f a r e M i n i s t e r ’ s S ec r e t r a i a t( ) , B u r e a u o f S t a t i s t i c s , O ffi ce o f t h e P r i m e M i n i s t e r ( b ) , B u r e a u o f S t a t i s t i c s , O ffi ce o f t h e P r i m e M i n i s t e r ( ) , a nd D i v i s i o n o f H e a l t h a nd W e l f a r e S t a t i s t i c s , W e l f a r e M i n i s t e r ’ s S ec r e t r a i a t( ) . ohorts for Analysis M a l e s / F e m a l e s Age measured in 1950 (a) 1950 Population Census
Cohorts for Analysis M a l e s / F e m a l e s Age measured in 1955 (b) 1955 Population Census
Figure 4: Sex Ratios by Prefecture Measured in the 1950 and 1955Population Censuses
Notes: The sex ratio is defined as the number of men divided by the number of women. All those rates arethe national averages based on the 1950 and 1955 Population Censuses. Source: Created by the authorsusing Bureau of Statistics, Office of the Prime Minister (1951a) and Bureau of Statistics, Office of the PrimeMinister (1956a).
Marriage Status
To better understand the potential influence of the wartime losses of men on the marriagemarket after the war, we use several demographic variables measured in the populationcensuses. As discussed in Section 2.2, those include the proportion of singles, marriedpeople, divorced people, and widowhoods. Since the population census captures thepeople’s status in October in the census years, these variables reflect marriage statusat the survey points. This enables us to mitigate the potential influence of internalmigration because we match the data on the sex ratio to the outcome variables measuredat the same survey points. Figure 3 illustrates the national averages of these outcomevariables. Panel A of Table 1 presents the summary statistics. Online Appendix B.1describes the sources of the documents in detail.
Sex Ratio
We aim to measure changes in the sex ratio due to the wartime losses of men as thesex ratio in each prefecture-year-age cell. Figure 4 decomposes the national average sexratio illustrated in Figure 1 into all 46 prefectures. We focus on people aged 17–50 ineach census year, meaning that those born between 1900 and 1939 are included in the In other words, the number of singles, married people, divorced people, and widowhoods measuredin the population censuses are the stock rather than the flow of these measures. Figures 4a and 4b both confirm that all the prefectures experience declinesin the sex ratio and that those shocks persist, as shown in Figure 1. While there aresome variations in the degrees of the reductions in the sex ratios over the prefectures,unobservable factors that might be correlated with the sex ratio can be controlled forin our model presented in the next subsection. In the regression analysis, we use theadjusted sex ratio in the spirit of Brainerd (2017) to account for gender differences in ageat marriage. The sex ratio in prefecture ( i )-year ( t )-age ( a ) cell is defined as follows: SR ita = (cid:80) j = − MALE i,t,a + j (cid:80) j = − FEMALE i,t,a + j (1)where MALE and
FEMALE are the number of men and women, respectively. Figure B.2in Online Appendix B.2 illustrates the adjusted sex ratio by prefecture and age. Finally,Figure 1c shows a relatively clear boom in the late 10s and 20s in one prefecture, whichindicates the influx of younger male workers to Tokyo. We confirm that our main resultsare robust to that potential influx in section 6.
To improve the identification, we employ a quasi-experimental design that uses changesin the sex ratio due to the substantial wartime losses of men as an exogenous shock on themarriage market after the war. We consider a bilateral-specific fixed effect model definedas follows: y ita = π + δ SR ita + φ ia + ν t + (cid:15) ita , (2)where i indicates the prefecture, t indicates the measured census year, and a indicatesage ranges from 17 to 50. The variables y and SR are the outcome variable and sex ratiodefined in equation 1 that are measured at each prefecture-year-age cell, respectively. φ isthe prefecture-by-age fixed effect, ν is the year fixed effect, and (cid:15) is a random error term.As introduced in Section 3.1, our panel data have a three-dimensional panel structurewith respect to prefecture, year, and age and we assign fixed effects to the prefecture-by-age cells ( φ ia ). This flexible setting of the error component enables us to control for This age range is slightly wider than that used in previous studies. For instance, Brainerd (2017)focuses on people aged 18–44 in 1959. We use a wider age range given that while the number of singlesin their 40s was stable, the proportion of marriages, divorces, and widowhoods still changed in their 40s,as shown in Figure 3. We confirm that our main results are not sensitive to slight changes in the age binsin the sex ratio (Section 6). Hence, the majority ofthe prefecture-specific trends in all the dependent variables with respect to age are con-trolled for using bilateral-specific fixed effects (Davis 2002). In addition, the unobservabletime trends and macroeconomic shocks are captured using the year fixed effect, ν . Theremaining variations used for the identification are then the within variations in eachprefecture-by-age cell excluding the parallel shifts over the cells.The key identifying assumption is that the variations in the sex ratio are exogenous toany socioeconomic conditions that might affect the outcome variables. This assumption isplausible given that the draft was conducted randomly during the war (Watanabe 2014a;2014b) and that the losses of men were not concentrated in provincial urban areas butdistributed equally over the prefectures (Ogasawara and Komura 2021). Potential threatsthat may be correlated with the sex ratio and marriage market outcomes in this settingare internal migration and regional economic losses (Brainerd 2017). First, although somepeople must have crossed prefecture borders from rural to urban regions after the war,this migration did not change the overall distributions of the sex ratios in Japan between1950 and 1955, as shown in Figure 4. This means that internal migration only occurredin a few large prefectures such as Tokyo and Osaka, as suggested in Section 3.1, ratherthan in all prefectures. Indeed, (Okazaki and Suda 1969, p. 54) reveal that cross-bordermigration over prefectures was limited throughout the 1950s. Moreover, as discussed earlier, Figure 4 indicates that these systematic flows of tem-porary workers from rural areas to Tokyo only comprise men in their early 20s. Since ourmodel in equation 2 uses the within variations across years, the most important variationsfor the identification are the dramatic changes in the sex ratios after the late 20s. Thismeans that our results should not be influenced by such migration. In other words, the age effects are allowed to vary across prefectures because we account for theinteraction between the prefecture and age fixed effects. Brainerd (2017) is the first study to use aregion-by-age panel in a single census year of 1959 to analyze the impacts of the wartime losses of men inRussia. The cohorts are identical to the measured ages in 1959 in her two-way region-by-age panel, whichcan identify the cohort effects of the losses. In this light, we expand her technique to a three-dimensionalpanel by controlling for age fixed effects to identify the cohort effects using prefecture-by-age fixed effects.This means that we use the within variations across years in each prefecture-by-age cell to disentanglecohort effects from age effects. The average cross-border migration rate (i.e., migration per 100 people) was less than 3% in 1955and roughly 30% of that migration occurred from non-metropolitan to metropolitan areas (Okazaki andSuda 1969, p. 55). Comparing Figure 4a with 4b suggests this point: specifically, the model may use both the dramaticimprovements in the sex ratios between, say, 26 and 31 from 1950 and 1955 and the substantial declinesin the ratios between, say, 32 and around 40 from 1950 and 1955. Despite this, in Section 6, we show the robustness of our main results by including an indicatorvariable for observations aged less than 30 in Tokyo and the interaction term with respect to the 1955
Overall Effects
Panel A of Table 2 presents the results from the specification of equation 2. Columns(1)–(4) and (5)–(8) show the results for women and men, respectively. First, column (1)indicates that the sex ratio is positively associated with the proportion of single women.The estimate suggests that a one standard deviation decrease in the sex ratio decreasesthe number of single women by 44 per 1 ,
000 women. Correspondingly, the estimatedcoefficient on the sex ratio is negative and statistically significant (column (2)). Theestimate indicates that a one standard deviation decrease in the sex ratio increases thenumber of married women by 38 women. The results for men are similar to those forwomen (columns (5) and (6)). The estimates reported in both columns indicate that aone standard deviation decrease in the sex ratio decreases the number of single (married)men by 40 (41) per 1 ,
000 men.Column (3) indicates that the coefficient on the sex ratio is estimated to be − .
57 andstatistically significant. This estimate suggests that a one standard deviation decrease inthe sex ratio increases the proportion of divorced women by 8 . ,
000 women. We finda much smaller estimate for men (column (7)). The estimated coefficient is − .
35 andstatistically significant, suggesting that a one standard deviation decrease in the sex ratioincreases the proportion of divorced men by 3 per 1 ,
000 men. We find a similar resultfor widowhoods. The estimate in column (4) is weakly statistically significantly positive, dummy in our baseline specification. . .
0) per 1 ,
000 women(men), respectively.Overall, relative male scarcity due to the war has similar impacts on both gendersin terms of the proportion of singles, marriages, and divorces. The women and menin prefectures with higher relative male scarcity are more likely to marry and divorce.This result may simply reflect the fact that younger people who could not marry duringwartime in those prefectures started to be matched as Figure 2a indicates. The greaternumber of marriages may have systematically induced the greater likelihood of divorce(Section 2.2). The differences in the magnitude of the estimates between women and men,however, might suggest the higher bargaining position of men in the marriage market.Indeed, the clear gender differences in the estimates for divorces (columns (3) and (7))and widowhoods (columns (4) and (8)) suggest better outside options for men underrelative male scarcity. In other words, men might be more likely to remarry than women,who faced worse outside options after the war. This finding is consistent with the lowerproportion of divorced and widowed men than women (Figure 3).It is also useful to discuss some of the differences between our results and those ofrelated previous studies in terms of the cost of divorce. In post-World War I France,the greater wartime losses of men decreased the proportion of divorced women and men,implying that the women facing relative male scarcity were more likely to stay single andless likely to ever marry (Abramitzky et al. 2011, p. 136). Our result suggests that theopposite movement occurred in post-World War II Japan within the same mechanisms.In the case of Russia after World War II, the men in regions with greater male lossesdue to the war were less likely to marry. Brainerd (2017) explains that the stronglypronatalist Family Code of 1944, which led to the high cost of divorce as well as nearlycostless non-marital sexual relations, decreased the probability of male marriage. In thislight, Japanese men experienced a similar institutional change after the war. Duringthe changes in the democratization policies in Japan, the Family Laws enshrined in theCivil Code were overhauled in 1947. The Civil Code allowed the divorce and division ofproperties at divorce, which might have increased the cost of divorce for men. However, A newspaper in 1946 states: “people were encouraged to get married to fend themselves, so theyheld joint wedding ceremonies” (Shinbun 1946). Another article describes the tendency to hold cheaperwedding ceremonies than in the prewar period, which may have increased the number of marriages afterthe war (Shinbun 1949) Under the Family Laws within the Meiji Civil Code dating from 1898, although bigamy was forbidden for both husband and wife, adultery committed by a wife was recognized grounds for divorce, whileadultery on the husband’s part could only be a reason for divorce if he were found guilty of the crime ofillicit intercourse. The new Civil Code abolished this institution and established the equality of men andwomen upon divorce. See Online Appendix A.1 for the details of this institutional background. a b l e : E ff ec t s o f t h e G e nd e r I m b a l a n ce o n M a rr i ag e M a r k e t O u t c o m e s : nd P o pu l a t i o n C e n s u s D a t a W o m e n M e n ( )( )( )( )( )( )( )( ) S i n g l e M a rr i e d D i v o r ce d W i d o w h oo dS i n g l e M a rr i e d D i v o r ce d W i d o w h oo d P a n e l A : O v e r a ll E ff ec t s S e x R a t i o ( S R ) . − . − . . . − . − . . ( . )( . )( . )( . )( . )( . )( . )( . ) P a n e l B : H e t e r og e n e o u s E ff ec t s S e x R a t i o ( S R ) . − . − . − . . − . − . . ( . )( . )( . )( . )( . )( . )( . )( . ) S e x R a t i o ( S R ) × I ( Y e a r = ) − . − . . . − . . . − . ( . )( . )( . )( . )( . )( . )( . )( . ) *** , ** , a nd * r e p r e s e n t s t a t i s t i c a l s i g n i fi c a n ce a tt h e % , % , a nd % l e v e l s , r e s p ec t i v e l y . S t a nd a r d e rr o r s f r o m t h ec l u s t e r - r o bu s t v a r i a n cee s t i m a t i o n r e p o r t e d i np a r e n t h e s e s a r ec l u s t e r e d a tt h e - p r e f ec t u r e l e v e l. N o t e s : T h e d e p e nd e n t v a r i a b l e s u s e d i n c o l u m n s ( ) – ( ) a r e t h e p r o p o r t i o n o f s i n g l e m e n , m a rr i e d m e n , d i v o r ce d m e n , a nd w i d o w e d m e np e r , m e n , r e s p ec t i v e l y . T h e d e p e nd e n t v a r i a b l e s u s e d i n c o l u m n s ( ) – ( ) a r e t h e p r o p o r t i o n o f s i n g l e w o m e n , m a rr i e d w o m e n , d i v o r ce d w o m e n , a nd w i d o w e d w o m e np e r , w o m e n , r e s p ec t i v e l y . I ( Y e a r = ) i nd i c a t e s a n i nd i c a t o r v a r i a b l e t h a tt a k e s o n e i f t h e y e a r i s nd ze r o i f t h e y e a r i s ,i. e ., a1955 y e a r du mm y t h a t d e p e nd s o n l y o n t . T h e nu m b e r o f o b s e r v a t i o n s i s , ( p r e f ec t u r e s × - ag e r a n g e × ce n s u s y e a r s ) i n a ll t h e r e g r e ss i o n s . A ll t h e r e g r e ss i o n s i n c l ud e p r e f ec t u r e - ag e - s p ec i fi c fi x e d e ff ec t s a nd y e a r fi x e d e ff ec t s . A ll t h e r e g r e ss i o n s a r e w e i g h t e db y t h e a v e r ag e nu m b e r o f p e o p l e ( w o m e n i n c o l u m n s ( ) – ( ) ; m e n i n c o l u m n s ( ) – ( )) i n e a c hp r e f ec t u r e - ag ece ll. eterogeneous Effects To assess the adjustment mechanism in the marriage market, we consider a flexible speci-fication that includes an interaction term between the sex ratio and an indicator variablethat takes one for the later census year, 1955. We expect that the combination of theestimated coefficients on the sex ratio and its interaction term can be used to analyze thepersistence of relative male scarcity in the marriage market.Panel B of Table 2 presents the results from our flexible specification following thesame layout as Panel A. First, we assess the results for women (columns (1)–(4)). Column(1) indicates that the impacts of the wartime losses of men on the proportion of singlewomen decrease from 1950 to 1955. The marginal effects are estimated to be 480 .
32 in1950 and 387 .
82 (480 . − .
50) in 1955. Column (2) indicates that the main effect ofthe sex ratio is statistically significantly negative and that its interaction effect is alsonegative and weakly statistically significant. We find a similar result for the proportion ofdivorced women in column (3): the interaction effect is weakly statistically significantlypositive but small. Both results imply that the impacts of relative male scarcity on theproportion of married and divorced women persists in 1955. Indeed, the marginal effectsin 1955 are estimated to be − .
72 in column (2) and − .
78 in column (3), which arenot far from the overall effect (reported in columns (2) and (3) of Panel A).We find an interesting change in the impacts of relative male scarcity on the proportionof widowed women. The estimates in column (4) indicate that the marginal effects of thesex ratio are − .
84 in 1950 and 109 .
71 ( − .
84 + 148 .
55) in 1955. This implies thata one standard deviation decrease in the sex ratio is associated with an increase in theproportion of widowed women by 3 . ,
000 women in 1950, but a decrease in theproportion of widowed women by 11 per 1 ,
000 people in 1955. Both social norms andinstitutional reasons may explain this result for widowed women. In the initial stage by1950 (i.e., soon after the end of the massive repatriation), the higher male scarcity dueto the wartime losses of husbands can simply reflect the greater number of widowhoods.Kawaguch (2003) explains that widowed women preferred not to remarry as a “ eirei notsuma (wife of spirits of war dead).” However, by the early 1950s, widowed women mighthave faced serious economic hardship. Indeed, the administration encouraged widowedwomen to remarry to cope with economic hardship, especially when the military pensionwas partly abolished in 1946 by the General Headquarters of the Allied Powers. The A popular magazine titled
Syufu-no-Tomo (housewife’s friend) for women argued: “For widows,remarry first, and [seek] public assistance second.”. This describes the clear trend that the remarriage ofwidowed women was regarded as a strategy to be financially independent at that time(Kawaguch 2003, .
80 in 1950 and 325 .
98 (467 . − .
82) in 1955.We find a corresponding result for the proportion of married people in column (6): themain effect of the sex ratio is statistically significantly negative and its interaction effectis statistically significantly positive. This result suggests that the impacts of relative malescarcity on the proportion of married men decreases over time. This trend is consideredto be consistent with the rightward shift of the distribution of the sex ratio shown inFigure 1. Indeed, the sex ratios for 20s improve from 1950 to 1955 (Figure 1b comparedwith 1c), which might attenuate the better outside options in the marriage market formen (note that we fix the age bin (i.e., 17–50) in the analyses). The results for divorcedand widowed men can support this interpretation. Column (7) indicates that the marginaleffect of the sex ratio on the proportion of divorced men attenuated in 1955. A similarattenuation can be found in the proportion of widowed men (column (8)). The estimatedmarginal effects in 1955 are 24 .
61 ( − .
01 + 10 .
40) in column (7) and 22 .
92 (54 . − . p.116–123). Additional Analyses
In this section, we investigate the impacts of the wartime losses of men on age at firstmarriage and birth outcomes. First, we test the influence of the age of grooms and brides,marital fertility, and stillbirths using the prefecture-year-level panel dataset obtained fromboth census reports and vital statistics records. Second, we analyze out-of-wedlock birthsusing the age-year-level panel dataset from census reports.
Although the prefecture-year-age-level panel data on grooms and brides’ ages are notavailable, we can obtain prefecture-year-level information on average age at first marriagefrom the official reports of the censuses. To prepare the difference-in-differences setting,we digitize the statistics not only from the postwar censuses of 1950 and 1955 but alsofrom the prewar census of 1935 (Bureau of Statistics, Office of the Prime Minister 1951b,1956b; Statistics Bureau of the Cabinet 1939b). In addition, we investigate the potentialquality/quantity adjustment at birth due to the gender imbalance caused by the war. Todo so, we digitize the maternal fertility and stillbirth rates using both the censuses andthe vital statistics of Japan (Division of Health and Welfare Statistics, Welfare Minister’sSecretraiat 1953, 1957; Statistics Bureau of the Cabinet 1936). Panel B of Table 1presents the summary statistics. Online Appendix B.3 shows the summary statistics bymeasured year.For prefecture i and measured year t , we specify the regression as follows: h it = ω + γ SR it + µ i + λ t + e it , (3)where h is average age (and the age gap) at first marriage, the marital fertility rate,or the stillbirth rate. SR is the sex ratio modified based on equation 1, which can besimply expressed as the number of men aged 17–50 divided by the number of women aged15–40. µ is the prefecture fixed effect, λ is the year fixed effect, and e is a random We use the marital fertility rate (i.e., the number of live births per 1,000 married women) ratherthan the general fertility rate (i.e., the number of live births per 1,000 women aged 15—44) becauseout-of-wedlock births were rare in Japan at that time (Section 5.2). We modify the adjusted sex ratio of equation 1 to merge the prefecture-year-level data as SR it = (cid:80) a =17 MALE i,t,a / (cid:80) a =15 FEMALE i,t,a , where
MALE i,t,a and
FEMALE i,t,a are the number of men andwomen in the i - t - a cell, respectively. The prefecture-year-age-level data on the number of men and women Age at First Marriage
First, we examine the impacts of relative male scarcity on assortative matching in termsof age. Table 3 presents the results for the age of grooms and brides and their age gap.Columns (1)–(4) show the results for the regressions using equation 3, whereas columns(5)–(8) show the results for the regressions using the same equation but including thefemale labor force participation rate as a control variable. Panel A presents the resultsfor the 1935 and 1950 samples, whereas Panel B presents those for the 1935 and 1955samples.Column (1) of Panel A indicates that the estimated coefficient on the sex ratio ispositive and statistically significant. The estimate suggests that a one standard deviationdecrease in the sex ratio is associated with a decrease in the average age of brides by 0 . . in 1930 and 1935 are obtained from the 46 volumes of the official reports of the 1930 and 1935 PopulationCensuses. This means that we use 92 volumes in total to construct the data on the prefecture-year-leveladjusted sex ratio. For simplicity, however, we note the citations as Statistics Bureau of the Cabinet(1933) and Statistics Bureau of the Cabinet (1939a). Another way to address heterogeneous treatment effects over time is to use the specification includingthe interaction term between the sex ratio and year dummies with the pooled panel data from 1935 to1955. However, such a specification has to assume that prefecture fixed effects do not vary between 1950and 1955. An advantage of this specification using three census years at the same time is that it caninclude the prefecture-specific line time trends in the model. However, given that we are using a shortpanel, it considerably reduces efficiency because more than 30% of the observations are used to estimatethe trends. Considering this, we use the 1950 and 1955 census data separately in this analysis. a b l e : E ff ec t s o f t h e G e nd e r I m b a l a n ce o n A g e a t F i r s t M a rr i ag e : , , a nd P o pu l a t i o n C e n s u s D a t a A v e r ag e ag e a t fi r s t m a rr i ag e ( )( )( )( )( )( ) W o m e n M e n D i ff e r e n ce W o m e n M e n D i ff e r e n ce P a n e l A : C e n s u s D a t a S e x R a t i o ( S R ) . . − . . . − . ( . )( . )( . )( . )( . )( . ) F e m a l e L a b o r F o r ce P a r t i c i p a t i o n R a t e N o N o N o Y e s Y e s Y e s P a n e l B : C e n s u s D a t a S e x R a t i o ( S R ) . . − . . . − . ( . )( . )( . )( . )( . )( . ) F e m a l e L a b o r F o r ce P a r t i c i p a t i o n R a t e N o N o N o Y e s Y e s Y e s *** , ** , a nd * r e p r e s e n t s t a t i s t i c a l s i g n i fi c a n ce a tt h e % , % , a nd % l e v e l s , r e s p ec t i v e l y . S t a nd a r d e rr o r s f r o m t h ec l u s t e r - r o bu s t v a r i a n cee s t i m a t i o n r e p o r t e d i np a r e n t h e s e s a r ec l u s t e r e d a tt h e - p r e f ec t u r e l e v e l. N o t e s : T h e d e p e nd e n t v a r i a b l e s u s e d i n c o l u m n s ( ) a nd ( ) a r e a v e r ag e ag e a t fi r s t m a rr i ag e f o r w o m e n a nd m e n , r e s p ec t i v e l y . T h e d e p e nd e n t v a r i a b l e u s e d i n c o l u m n ( ) i s t h e d i ff e r e n ce i n a v e r ag e ag e a t fi r s t m a rr i ag e b e t w ee n w o m e n a nd m e n ( w o m e n m i nu s m e n ) . A ll t h e r e g r e ss i o n s i n c l ud e p r e f ec t u r e a nd y e a r fi x e d e ff ec t s . C o l u m n s ( ) – ( ) i n c l ud e t h e f e m a l e l a b o r f o r ce p a r t i c i p a t i o n r a t e , d e fin e d a s t h e nu m b e r o ff e m a l e w o r k e r s p e r w o m e n . T h e nu m b e r o f o b s e r v a t i o n s i s ( p r e f ec t u r e s × ce n s u s y e a r s ) i n a ll t h e r e g r e ss i o n s . A ll t h e r e g r e ss i o n s a r e w e i g h t e db y t h e a v e r ag e nu m b e r o f m a rr i ag e s i n e a c hp r e f ec t u r ece ll. arital Fertility and Stillbirth Rates Next, we examine the impacts of relative male scarcity on birth outcomes after marriage.Table 4 presents the results for marital fertility and stillbirth rates. Columns (1) and (2)show the results for the maternal fertility rate, whereas columns (3) and (4) show thosefor the stillbirth rate. The regressions in columns (2) and (4) include the female laborforce participation rate as a control variable. Panel A (B) presents the results for the1935 and 1950 (1955) samples.Column (1) of Panel A indicates that the estimated coefficient is statistically signifi-cantly negative and that this result is unchanged if we control for the potential opportunitycost of giving birth (column (2) of Panel A). The estimate in column (1) of Panel A im-plies that a one standard deviation decrease in the sex ratio is associated with an increasein live births by 13 per 1 ,
000 married women. In the case of post-World War II Russia,a lower sex ratio resulting from relative male scarcity resulted in fewer marital births,implying that men preferred fewer children than women and had a greater influence onfertility choices (Brainerd 2017, p. 237). In this light, a potential explanation for our re-sult is that men preferring more children than women had a stronger bargaining positionin the fertility decision (Ogasawara and Komura 2021). Columns (1) and (2) of PanelB suggest that this effect on marital fertility becomes statistically insignificant in 1955,implying that improvements in the gender imbalance might have attenuated the relativeadvantage of men.Columns (3) and (4) of Panel A suggest that the estimated effects are positive butstatistically insignificant in 1950. In columns (3) and (4) of Panel B, however, we finda negative relationship between the sex ratio and stillbirth rate in 1955. The estimatein column (3) of Panel B indicates that a one standard deviation decrease in the sexratio increases the number of stillbirths by 7 . ,
000 births. As explained before, animportant fact here is that the number of stillbirths in the postwar period includes thelarge number of deaths before birth due to abortion (see Section 2.3). The statisticallyinsignificant influence of relative male scarcity on marital fertility in columns (1) and (2)of Panel B can be partly explained by the rise in the number of abortions.Overall, our result suggests that men, who had a stronger bargaining position thanwomen, might have decided to abort, especially when the pregnancy was unexpected. Asdescribed in Section 2.3, an official survey of 1964 indicates that a large number of marredwomen who experienced abortions had two or more children beforehand. This is consistentwith the findings of Norgren (2008) that people lacked knowledge of contraception at that26able 4: Effects of the Gender Imbalance on Marital Fertility and Stillbirth:1935, 1950, and 1955 Population Census and Vital Statistics (VS) Data
Marital Fertility Rate Stillbirth Rate(1) (2) (3) (4)Panel A: 1935–1950 Census and VS DataSex Ratio ( SR ) − . − . .
286 20 . . . . . SR ) − . − . − . − . . . . . ***, **, and * represent statistical significance at the 1%, 5%, and 10% levels, respectively. Standard errors fromthe cluster-robust variance estimation reported in parentheses are clustered at the 46-prefecture level.Notes: The dependent variable used in columns (1) and (2) is the number of live births per 1 ,
000 married women.The dependent variable used in columns (3) and (4) is the number of stillbirths per 1 ,
000 births. All the regressionsinclude prefecture and year fixed effects. Columns (3) and (4) include the female labor force participation rate,defined as the number of female workers per 100 women. The number of observations is 92 (46 prefectures × time and thus relied heavily on the use of artificial abortions. Out-of-wedlock Births
Existing evidence suggests that exogenous male losses due to wars raise the share of out-of-wedlock births. In France after World War I, out-of-wedlock births were positivelycorrelated with military mortality rates (Abramitzky et al. 2011). Bethmann and Kvas-nicka (2012) reveal that the wartime losses of men increased the share of out-of-wedlockbirths among total births after the Second World War in Bavaria. Brainerd (2017) alsofinds that the decline in the sex ratio is predicted to increase the share of out-of-wedlockbirths in urban areas of Russia after the war. The weight of evidence thus indicates apositive causal effect of a decline in the sex ratio on the share of out-of-wedlock births,which is consistent with the theoretical implications of intra-household bargaining (Becker1991; Willis 1999).The current study builds on this evidence in the literature. However, we seek to addempirical evidence on Japan, a country that experienced a different trend of out-of-wedlockfertility. Indeed, while out-of-wedlock birth rates were increasing after the Second WorldWar in Germany and Russia, those rates were decreasing in postwar Japan. Indeed, the27hares of out-of-wedlock births were 14.2% in 1947 in Germany (Bethmann and Kvasnicka2012, p. 181) and 15.6% in 1959 in Russia (Brainerd 2017, p. 233), whereas the mean ratein 1950s Japan was 3 .
3% (Table 1). Thus, this study is the first to provide empiricalevidence on an economy that experienced the opposite trend of out-of-wedlock fertility. Despite the small proportion of illegitimate births at that time, the 1950 and 1955Population Censuses do document the number of out-of-wedlock births by maternal age(Bureau of Statistics, Office of the Prime Minister 1951a, 1956a). We can then calculatetwo measures of the probability of having out-of-wedlock births: the share of out-of-wedlock births and out-of-wedlock birth rate. The share of out-of-wedlock births is definedas the number of out-of-wedlock live births per 100 live births and the out-of-wedlock birthrate is the number of out-of-wedlock life births per 1,000 women. Panel C of Table 1presents the summary statistics. Figures B.3 and B.4 in Online Appendix B.3 show thoserates by maternal age.The regression specification is as follows: w at = α + β SR at + θ a + κ t + u at , (4)where a indicates the maternal age (from 17 to 50 years) and t indicates the measuredcensus year. The variable w is either the share of out-of-wedlock births or the out-of-wedlock birth rate and SR is the adjusted sex ratio based on equation 1. θ is theage fixed effect, κ is the year fixed effect, and u is a random error term. We find clearU-shaped and inverse U-shaped trends in the share of out-of-wedlock births and out-of-wedlock birth rate, respectively (Figures B.4 and B.3 in Online Appendix B.3). The agefixed effect can thus control for these systematic trends in both rates. Similarly, the yearfixed effect captures the decreasing trend in both rates. As these fixed effects capture thesystematic trends in both rates in terms of age and year, we use the within variation ineach maternal age cell to estimate the parameter of interest, β . Specifically, 3 .
6% in 1950 and 3 .
1% in 1955. If we focus on the rates for the 20–40 age range, theywere roughly 1.0–2.0% (Figure B.4 in Online Appendix B.3). Although we cannot directly compare thefigures because of differences in the sample periods, the proportion of out-of-wedlock births was 7.4% inpost-World War I France Abramitzky et al. (2011, p. 150). While this difference may be an interesting topic for future studies, the potential reasons for such alow share of out-of-wedlock births include the divorce law in postwar Japan, which prohibited other part-ners outside marriage, and the inheritance law, which was unfavorable for out-of-wedlock-born children(Ogasawara and Komura 2021). To merge the adjusted sex ratio of equation 1 into the maternal age-year-level data on the share ofout-of-wedlock births, we modify equation 1 as SR at = (cid:80) j = − (cid:80) i =1 MALE i,t,a + j (cid:80) j = − (cid:80) i =1 FEMALE i,t,a + j . .2.2 Results Table 5 presents the results. Columns (1) and (2) show the results for the share of out-of-wedlock births. Column (1) suggests that a one standard deviation decrease in thesex ratio increases out-of-wedlock births by 0 . . × .
83) per 100 live births. We alsoconsider the specification that allows the marginal effects of the sex ratio to vary overthe measured years in column (2). The estimated coefficient on the interaction term isclose to zero and statistically insignificant, suggesting that the marginal effects in 1955are similar to those in 1950. In other words, the impacts of the wartime losses of men onthe share of out-of-wedlock births persisted throughout the 1950s.Columns (3) and (4) show the results for the out-of-wedlock birth rate. Column (3)indicates that a one standard deviation decrease in the sex ratio is associated with anincrease in out-of-wedlock births of 1 . .
28 and statistically significant, whichwipes out a large part of this negative effects of the sex ratio. As a result, the marginaleffect of the sex ratio is estimated to be − . . First, we test the potential influence on the influx of younger men to Tokyo in 1955. Asdiscussed earlier, there must have been an influx of male workers aged in their 10s–20sto 1955 Tokyo. Table 6 presents the results from the specification including an indicatorvariable for observations aged 17–29 in 1955 Tokyo in equation 2. In this specification, wecontrol for unobserved factors that might be correlated with changes in the sex ratio forthose aged 17–29 in Tokyo from 1950 to 1955. As shown, the results are largely unchanged29able 5: Effects of the Gender Imbalance on Out-of-Wedlock Fertility:1950 and 1955 Population Census Data
Share of Out-of-Wedlock Births Out-of-Wedlock Birth Rate(1) (2) (3) (4)Sex Ratio ( SR ) − . − . − . − . .
90) (1 .
35) (1 .
79) (0 . SR ) × I(Year=1955) .
91 13 . .
70) (0 . ***, **, and * represent statistical significance at the 1%, 5%, and 10% levels, respectively. Standard errors from thecluster-robust variance estimation reported in parentheses are clustered at the 34-maternal age level.Notes: The dependent variable used in columns (1) and (2) is the proportion of out-of-wedlock live births per 100live births. The dependent variable used in columns (3) and (4) is the number of out-of-wedlock live births per 1,000women. I(Year=1955) indicates an indicator variable that takes one if the year is 1955 and zero if the year is 1950, i.e.,a 1955 year dummy that depends only on t . The number of observations is 68 (34 maternal ages × if we include the indicator variable for the potential influx of men into the metropolitanarea. 30 a b l e : R o bu s t n e ss C h ec k s : E ff ec t s o f t h e G e nd e r I m b a l a n ce o n M a rr i ag e M a r k e t O u t c o m e s C o n tr o lli n g f o r a n I nd i c a t o r V a r i a b l e f o r T o ky o–1955– A g e d C e ll s W o m e n M e n ( )( )( )( )( )( )( )( ) S i n g l e M a rr i e d D i v o r ce d W i d o w h oo dS i n g l e M a rr i e d D i v o r ce d W i d o w h oo d P a n e l A : O v e r a ll E ff ec t s S e x R a t i o ( S R ) . − . − . . . − . − . . ( . )( . )( . )( . )( . )( . )( . )( . ) P a n e l B : H e t e r og e n e o u s E ff ec t s S e x R a t i o ( S R ) . − . − . − . . − . − . . ( . )( . )( . )( . )( . )( . )( . )( . ) S e x R a t i o ( S R ) × I ( Y e a r = ) − . − . . . − . . . − . ( . )( . )( . )( . )( . )( . )( . )( . ) *** , ** , a nd * r e p r e s e n t s t a t i s t i c a l s i g n i fi c a n ce a tt h e % , % , a nd % l e v e l s , r e s p ec t i v e l y . S t a nd a r d e rr o r s f r o m t h ec l u s t e r - r o bu s t v a r i a n cee s t i m a t i o n r e p o r t e d i np a r e n t h e s e s a r ec l u s t e r e d a tt h e - p r e f ec t u r e l e v e l. N o t e s : T h e d e p e nd e n t v a r i a b l e s u s e d i n c o l u m n s ( ) – ( ) a r e t h e p r o p o r t i o n o f s i n g l e m e n , m a rr i e d m e n , d i v o r ce d m e n , a nd w i d o w e d m e np e r , m e n , r e s p ec t i v e l y . T h e d e p e nd e n t v a r i a b l e s u s e d i n c o l u m n s ( ) – ( ) a r e t h e p r o p o r t i o n o f s i n g l e w o m e n , m a rr i e d w o m e n , d i v o r ce d w o m e n , a nd w i d o w e d w o m e np e r , w o m e n , r e s p ec t i v e l y . I ( Y e a r = ) i nd i c a t e s a n i nd i c a t o r v a r i a b l e t h a tt a k e s o n e i f t h e y e a r i s nd ze r o i f t h e y e a r i s ,i. e ., a1955 y e a r du mm y t h a t d e p e nd s o n l y o n t . T h e nu m b e r o f o b s e r v a t i o n s i s , ( p r e f ec t u r e s × - ag e r a n g e × ce n s u s y e a r s ) i n a ll t h e r e g r e ss i o n s . A ll t h e r e g r e ss i o n s i n c l ud e a n i nd i c a t o r v a r i a b l e t h a tt a k e s o n e f o r t h e o b s e r v a t i o n s ag e d i n T o ky o , p r e f ec t u r e - ag e - s p ec i fi c fi x e d e ff ec t s , a nd y e a r fi x e d e ff ec t s . A ll t h e r e g r e ss i o n s a r e w e i g h t e db y t h e a v e r ag e nu m b e r o f p e o p l e ( w o m e n i n c o l u m n s ( ) – ( ) ; m e n i n c o l u m n s ( ) – ( )) i n e a c hp r e f ec t u r e - ag ece ll. sensitive to changes in the age window ofthe sex ratio. This is because it uses the within variation in each prefecture-age cellfor the identification, meaning that a substantially narrow age window undercuts theuseful information in the exogenous changes in the sex ratio due to the war, whereasa substantially broader age window attenuates the estimates. Note that a one age lossin the sex ratio leads to 33 age-cell losses in calculating the sex ratio in a given censusyear in our definition (equation 1), meaning a substantial change in the target marriagemarket. Indeed, we confirm that our baseline estimates tend to be sensitive to such achange if we alter the age window by ± ± Sex Ratio (Narrow) indicates thenumber of men from − Sex Ratio (Broad) indicates the numberof men from − a b l e : R o bu s t n e ss C h ec k s : E ff ec t s o f t h e G e nd e r I m b a l a n ce o n M a rr i ag e M a r k e t O u t c o m e s C h a n g e s i n t h e A g e W i nd o w s ( ± Y e a r Sh i f t s f r o m t h e B a s e li n e D e fin i t i o n ) W o m e n M e n ( )( )( )( )( )( )( )( ) S i n g l e M a rr i e d D i v o r ce d W i d o w h oo dS i n g l e M a rr i e d D i v o r ce d W i d o w h oo d P a n e l A : O v e r a ll E ff ec t s S e x R a t i o ( N a rr o w ) . − . − . . . − . − . . ( . )( . )( . )( . )( . )( . )( . )( . ) S e x R a t i o ( B r oa d ) . − . − . . . − . − . . ( . )( . )( . )( . )( . )( . )( . )( . ) P a n e l B : H e t e r og e n e o u s E ff ec t s S e x R a t i o ( N a rr o w ) . − . − . − . . − . − . . ( . )( . )( . )( . )( . )( . )( . )( . ) S e x R a t i o ( N a rr o w ) × I ( Y e a r = ) − . − . . . − . . . − . ( . )( . )( . )( . )( . )( . )( . )( . ) S e x R a t i o ( B r oa d ) . − . − . − . . − . − . . ( . )( . )( . )( . )( . )( . )( . )( . ) S e x R a t i o ( B r oa d ) × I ( Y e a r = ) − . − . . . − . . . − . ( . )( . )( . )( . )( . )( . )( . )( . ) *** , ** , a nd * r e p r e s e n t s t a t i s t i c a l s i g n i fi c a n ce a tt h e % , % , a nd % l e v e l s , r e s p ec t i v e l y . S t a nd a r d e rr o r s f r o m t h ec l u s t e r - r o bu s t v a r i a n cee s t i m a t i o n r e p o r t e d i np a r e n t h e s e s a r ec l u s t e r e d a tt h e - p r e f ec t u r e l e v e l. N o t e s : S e x R a t i o ( N a rr o w ) i s t h e nu m b e r o f m e n f r o m − t o9 y e a r s o l d e r t h a n a w o m a n o f ag i v e n ag e d i v i d e db y t h e nu m b e r o f w o m e n i n t h e s a m e ag e r a n g e . S e x R a t i o ( B r oa d ) i s t h e nu m b e r o f m e n f r o m − t o11 y e a r s o l d e r t h a n a w o m a n o f ag i v e n ag e d i v i d e db y t h e nu m b e r o f w o m e n i n t h e s a m e ag e r a n g e . T h e d e p e nd e n t v a r i a b l e s u s e d i n c o l u m n s ( ) – ( ) a r e t h e p r o p o r t i o n o f s i n g l e m e n , m a rr i e d m e n , d i v o r ce d m e n , a nd w i d o w e d m e np e r , m e n , r e s p ec t i v e l y . T h e d e p e nd e n t v a r i a b l e s u s e d i n c o l u m n s ( ) – ( ) a r e t h e p r o p o r t i o n o f s i n g l e w o m e n , m a rr i e d w o m e n , d i v o r ce d w o m e n , a nd w i d o w e d w o m e np e r , w o m e n , r e s p ec t i v e l y . I ( Y e a r = ) i nd i c a t e s a n i nd i c a t o r v a r i a b l e t h a tt a k e s o n e i f t h e y e a r i s nd ze r o i f t h e y e a r i s ,i. e ., a1955 y e a r du mm y t h a t d e p e nd s o n l y o n t . T h e nu m b e r o f o b s e r v a t i o n s i s , ( p r e f ec t u r e s × - ag e r a n g e × ce n s u s y e a r s ) i n a ll t h e r e g r e ss i o n s . A ll t h e r e g r e ss i o n s i n c l ud e p r e f ec t u r e - ag e - s p ec i fi c fi x e d e ff ec t s a nd y e a r fi x e d e ff ec t s . A ll t h e r e g r e ss i o n s a r e w e i g h t e db y t h e a v e r ag e nu m b e r o f p e o p l e ( w o m e n i n c o l u m n s ( ) – ( ) ; m e n i n c o l u m n s ( ) – ( )) i n e a c hp r e f ec t u r e - ag ece ll. Conclusion
This study used relative male scarcity due to the casualties of World War II to analyzethe impacts of the gender imbalance on marriage market and birth outcomes. We foundthat people who faced relative male scarcity were more likely to marry and divorce. Thegender difference in the estimates suggests that men had a stronger bargaining positionin the marriage market than women. Indeed, widowed women were less likely to remarrythan widowed men, who had better outside options. The overall situation of women inthe marriage market did not improve throughout the 1950s. Regarding birth outcomes,marital fertility rates increased in the areas that suffered relative male scarcity, althoughthis boom disappeared by 1955. Correspondingly, stillbirth rates increased substantiallyin the areas that faced relative male scarcity in the mid-1950s, which reflects the impactof the enactment of the Abortion Act. The wartime losses of men also increased the shareof out-of-wedlock births in the 1950s.Our evidence from post-World War II Japan is not without its limitations. Whilewe investigate the middle-run (i.e., 10 years after the war) effects of the gender imbal-ance on demographic outcomes, we provide no evidence on the long-term impacts of theunbalanced sex ratio because of the unavailability of systematic data on outcomes after1960. The unavailability of systematic data on assortative matching also made it difficultfor us to analyze the impacts of the gender imbalance on assortative matching after thewar. Despite these limitations, this study newly digitized a comprehensive census-baseddataset that has information on marriage market and birth outcomes at two survey pointsand exploited the plausibly exogenous variations in the wartime losses of men. This studyis thus the first to provide suggestive evidence on the dynamic relationships between thegender imbalance and demographic changes as well as the gender-based differences in itseffects. Finally, our result on out-of-wedlock births indicates that the theoretical pre-diction of intra-household bargaining is considered to be robust in an economy in whichmarital fertility is dominant. 35able 8: Robustness Checks: Effects of the Gender Imbalance on Age at First Marriage,Marital Fertility, Stillbirth, and Out-of-Wedlock FertilitySlight Changes in the Age Windows ( ± Average age at first marriage1935–1950 Census Data 1935–1955 Census Data(1) (2) (3) (4) (5) (6)Panel A Women Men Difference Women Men DifferenceSex Ratio (Narrow) 5 . . − .
165 1 .
232 0 . − . . . . . . . . . − .
242 0 .
565 0 . − . . . . . . . − . − . − . − . . . . . − . . − . − . . . . . − . − . − . − . .
87) (1 .
52) (0 .
92) (2 . − . − . − . − . .
35) (0 .
89) (1 .
36) (0 . × I(Year=1955) .
29 12 . .
29 14 . .
73) (0 .
93) (1 .
70) (1 . − − I(Year=1955) indicates an indicator variable that takes one if the year is 1955 and zero if the year is 1950, i.e., a1955 year dummy that depends only on t .In Panels A and B, the number of observations is 92 (46 prefectures × × eferences Ran Abramitzky, Adeline Delavande, and Luis Vasconcelos. Marrying Up: The Role ofSex Ratio in Assortative Matching.
American Economic Journal: Applied Economics ,3(3):124–157, July 2011.Daron Acemoglu and David H Autor. Women, War, and Wages: The Effect of FemaleLabor Supply on the Wage Structure at Midcentury Daron Acemoglu and David H.Autor.
Journal of Political Economy , 112(3):497–551, June 2004.Josh Angrist. How do sex ratios affect marriage and labor markets? Evidence fromAmerica’s second generation.
Quarterly Journal of Economics , 117(3):997–1038, August2002.Gary S. Becker. A theory of marriage: Part I.
Journal of Political Economy , 81(4):813–846, 1973.Gary S. Becker. A theory of marriage: Part II.
Journal of Political Economy , 82(2):S11–26, 1974.Gary S. Becker.
A treatise on the family . Cambridge, 1991.Marianne Bertrand, Esther Duflo, and Sendhil Mullainathan. How much should we trustdifferences-in-differences estimates?
Quarterly Journal of Economics , 119(1):249–275,February 2004.Dirk Bethmann and Michael Kvasnicka. World War II, Missing Men and Out of WedlockChildbearing.
The Economic Journal , 123(567):162–194, August 2012.Marianne P Bitler and Lucie Schmidt. Birth Rates and the Vietnam Draft.
AmericanEconomic Review , 102(3):566–569, May 2012.Elizabeth Brainerd. The Lasting Effect of Sex Ratio Imbalance on Marriage and Family:Evidence from World War II in Russia.
Review of Economics and Statistics , 99(2):229–242, May 2017.Bureau of Statistics, Office of the Prime Minister.
Sy¯owanijy¯ugonen Kokuseich¯osah¯okoku,dainanakan, fukenhen (1955 Population Census of Japan, prefecture part (volume 7)).[in Japanese] . Bureau of Statistics, Office of the Prime Minister, Tokyo, 1951a.37ureau of Statistics, Office of the Prime Minister.
Sy¯owanijy¯ugonen Kokuseich¯osah¯okoku,dainanakan, zenkokuhen (1950 Population Census of Japan). [in Japanese] . Bureau ofStatistics, Office of the Prime Minister, Tokyo, 1951b.Bureau of Statistics, Office of the Prime Minister.
Sy¯owasanjy¯unen Kokuseich¯osah¯okoku,daigokan, fukenhen (1955 Population Census of Japan, prefecture part (volume 5)). [inJapanese] . Bureau of Statistics, Office of the Prime Minister, Tokyo, 1956a.Bureau of Statistics, Office of the Prime Minister.
Sy¯owasanjy¯unen Kokuseich¯osah¯okoku,daigokan, zenkokuhen (1955 Population Census of Japan). [in Japanese] . Bureau ofStatistics, Office of the Prime Minister, Tokyo, 1956b.Bureau of Statistics Office of the Prime Minister.
Sh¯owajy¯ugonen Kokuseich¯osah¯okoku,daiikkan (1940 Population Census of Japan, volume 1). [in Japanese] . Bureau of Statis-tics Office of the Prime Minister, Tokyo, 1961.Kerwin Kofi Charles and Ming Ching Luoh. Male incarceration, the marriage market,and female outcomes.
The Review of Economics and Statistics , 92(3):614–627, August2010.Pierre Andr´e Chiappori, Bernard Fortin, and Guy Lacroix. Marriage Market, DivorceLegislation, and Household Labor Supply.
Journal of Political Economy , 110(1):37–72,February 2002.Peter Davis. Estimating multi-way error components models with unbalanced data struc-tures.
Journal of Econometrics , 106:67–95, 2002.Division of Health and Welfare Statistics, Welfare Minister’s Secretraiat.
Sy¯owanijy¯ugonenJink¯odo¯tait¯okei, Daiichibunsatu (1950 Vital Statistics, volume 1). [in Japanese] . Divi-sion of Health and Welfare Statistics, Welfare Minister’s Secretraiat, Tokyo, 1953.Division of Health and Welfare Statistics, Welfare Minister’s Secretraiat.
Sy¯owasanjy¯unenJink¯odo¯tait¯okei, Jy¯okan (1955 Vital Statistics, volume 1). [in Japanese] . Division ofHealth and Welfare Statistics, Welfare Minister’s Secretraiat, Tokyo, 1957.Lena Edlund, Hongbin Li, Junjian Yi, and Junsen Zhang. Sex Ratios and Crime: Evidencefrom China.
Review of Economics and Statistics , 95(5):1520–1534, December 2013.38arald Fuess. Meijiki no konin to rikon ibunkatekishiten de mita kindai syakai no seni,(Marriage and Divorce in the Meiji Era: Transitions in Modern Society from a Cross-Cultural Perspective). [in Japanese]. In Satomi Kurosu, editor,
Rekishijink¯ogaku karamita kekkon, rikon, saikon, (Marriage, Divorce, and Remarriage from the Perspectiveof the Historial Demography). [in Japanese] , pages 171–203. Reitaku University Press,Chiba, 2012.Fumio Hayashi and Edward C Prescott. The depressing effect of agricultural institutionson the prewar Japanese economy.
Journal of Political Economy , 116(4):573–632, August2008.Tomoko Kanzaki.
Sengo nihon jyosei seiakushi, (A History of post-war Japanese Women’sPolicy). [in Japanese] . Akashi Shoten, Tokyo, 2009.Emiko Kawaguch.
Sens¯o mib¯ojin higaito to kagai no hazamade, (War Widow: BetweenVictim and Offender). [in Japanese] . Domesu Publishers inc., Tokyo, 2003.Noriko Kurushima, Hiroko Nagano, and Shizue Osa.
Jend¯a kara mita nihonshi, (JapaneseHistory from a Gender Perspective). [in Japanese] . Otsuki Shoten, Tokyo, 2015.Jeanne Lafortune. Making Yourself Attractive: Pre-Marital Investments and the Returnsto Education in the Marriage Market.
American Economic Journal: Applied Economics ,5(2):151–178, January 2013.Kiyoshi Nakagawa.
Nihon toshi no seikatu hend¯o, (Life Changes in Urban Japan). [inJapanese] . Keiso Shobo, Tokyo, 2000.Takafusa Nakamura and Masayasu Miyazaki.
Shiry¯o Taiheiy¯osens¯ohigaich¯osah¯okoku(Damage Survey Report of the Second World War). [in Japanese] . University of TokyoPress, Tokyo, 1995.Tiana Norgren. cyu¯zetu to hinin no seijigaku sengo nihon no ripurodakusyon seisaku,(Abortion Before Birth Control: The Politics of Reproduction in Postwar Japan). [inJapanese] . Aoki Shoten Publishing Co., Tokyo, 2008.Kota Ogasawara and Mizuki Komura. Consequences of war: Japan’s demographic tran-sition and the marriage market.
Journal of Population Economics , forthcoming, 2021.Yoichi Okazaki and Tomi Suda. Recent trends of migration in Japan. [in Japanese].
TheJournal of Population Problems , 109:53–64, January 1969.39aria Porter. How do sex ratios in China influence marriage decisions and intra-householdresource allocation?
Review of Economics of the Household , pages 1–35, April 2015.Laura Salisbury. Women’s Income and Marriage Markets in the United States: Evidencefrom the Civil War Pension.
The Journal of Economic History , 77(1):1–38, February2017.Asahi Shinbun.
Ch¯okan, (Morning Newspaper) . November 1946.Asahi Shinbun.
Ch¯okan, (Morning Newspaper) . September 1949.Statistics Bureau of the Cabinet.
Nihonteikoku jink¯od¯otait¯okei (The vital statistics ofempire Japan, 1930 edition). [in Japanese] . Statistics Bureau of the Cabinet, Tokyo,1931.Statistics Bureau of the Cabinet.
Sh¯owagonen Kokuseich¯osah¯okoku, fukenhen (1930 Popu-lation Census of Japan, prefecture part). [in Japanese] . Statistics Bureau of the Cabinet,Tokyo, 1933.Statistics Bureau of the Cabinet.
Sh¯owagonen Kokuseich¯osah¯okoku, daiikkan (1930 Pop-ulation Census of Japan, volume 1). [in Japanese] . Statistics Bureau of the Cabinet,Tokyo, 1935.Statistics Bureau of the Cabinet.
Nihonteikoku jink¯od¯otait¯okei (The vital statistics ofempire Japan, 1935 edition). [in Japanese] . Statistics Bureau of the Cabinet, Tokyo,1936.Statistics Bureau of the Cabinet.
Sh¯owaj¯unen Kokuseich¯osah¯okoku, fukenhen (1935 Popu-lation Census of Japan, prefecture part). [in Japanese] . Statistics Bureau of the Cabinet,Tokyo, 1939a.Statistics Bureau of the Cabinet.
Sh¯owajy¯unen Kokuseich¯osah¯okoku, daiikkan (1935 Pop-ulation Census of Japan, volume 1). [in Japanese] . Statistics Bureau of the Cabinet,Tokyo, 1939b.Statistics Bureau of the Prime Minister’s Office.
Sh¯owanijy¯uninen Rinjikokuseich¯osakekkah¯okoku, sonoichi (1947 Extraordinary Population Census of Japan, volume 1).[in Japanese] . Statistics Bureau of the Prime Minister’s Office, Tokyo, 1948.40obuyoshi Toshitani. Senjitaisei to kazoku kokkas¯od¯oinh¯o ni okeru kazokuseisaku tokazokuh¯o, (Wartime Rejime and Family: Family Policy and Family Law in the NationalMobilization System). [in Japanese]. In Masao Fukushima, editor,
Kazoku seisaku toh¯o, vol.6, (Family: Policy and Law). [in Japanese] , pages 255–362. The University ofTokyo Press, Tokyo, 1984.Haruko Wakita, Reiko Hayashi, and Kazuko Nagahara.
Nihon jyosei shi, (JapaneseWomen’s History). [in Japanese] . Yoshikawa Kobunkan, Tokyo, 2011.Shang-Jin Wei and Xiaobo Zhang. The Competitive Saving Motive: Evidence from RisingSex Ratios and Savings Rates in China.
Journal of Political Economy , 119(3):511–564,June 2011.Robert J. Willis. A theory of out-of-wedlock childbearing.
Journal of Political Economy ,107(6):S33–64, December 1999.Yoshihiko Yuzawa.
Sengo kazoku hend¯o no t¯okeiteki k¯osatu, (A Statistical Study of Post-war Family Change). [in Japanese] . The University of Tokyo Press, Tokyo, 1977.Yoshihiko Yuzawa.
Kateinai jend¯a no genten meiji no kekkon meiji no rikon, (The originof Gender in the Home: Marriage and Divorce in Meiji Era). [in Japanese] . KadokawaGakugei Co., Tokyo, 2005. 41 ppendices ppendix A Background Appendix
A.1 Institutional Background
With the fall of the Japanese Tokugawa Shogunate in 1868, a modernization policy wasimplemented by the new Meiji Government. Nevertheless, the position of Japanese womenbefore World War II was still lower than that of men. Japan’s postwar move towarddemocracy led by the General Headquarters of the Allied Powers saw women gain theright to vote in 1945 and the promulgation of a new Japanese Constitution in 1946, whichled to the equality of the sexes. The transitions in Japanese civil law, civil rights, andschool education from the end of the 19th century to the mid-20th century are reviewedbelow.Established in 1870 and a forerunner to the Penal Code of Japan, the Outline ofthe New Criminal Code (
Shinritsu K¯ory¯o ) placed the wife and mistress of a man onessentially the same legal footing. Moreover, crimes by a wife or mistress against a manwere punished more severely than those against a wife or mistress committed by a manin that Code (see Wakita et al. 2011, p. 193). The Family Registration Law ( koseki-h¯o )passed in 1871 sought to establish control over the nation by establishing the home ( ie )in which one resided as the fundamental social unit. This law established the systematicdomination of men over women: with the head of the family at the top of the registry,direct ancestors, direct descendants, and male family members were positioned abovelineal descendants, collateral relatives, and women (The Research Society for Women’sHistory 1990, p. 4). In 1873, Edict No. 162 of the Grand Council of State ( daj¯okan ) gavewives access to courts to seek a divorce, with the condition that they be “accompaniedby [a] father, brother or relative” (Wakita et al. 2011, p. 194). Nonetheless, men werenot obliged to support their ex-wives, nor were women awarded custody of their children.Divorce was thus an event that disadvantaged women (Fuess 2012, p. 179).The Family Laws ( Mibun h¯o ) within the Meiji Civil Code dating from 1898 similarlyprovided for male dominance over women within the family. The Laws instituted as legalstandards rights accruing to the head of a household ( kosyu ken ), rights of succession tofamily headship ( katoku s¯ozoku ), and the family system based on the subordination offemale family members ( ie seido ). Although bigamy was forbidden for both husband andwife, adultery committed by a wife was recognized grounds for divorce, while adultery onthe husband’s part could only be a reason for divorce if he were found guilty of the crimeof illicit intercourse. The Meiji Civil Code also established the household head as the1uperior authority within the family, giving him the right to determine the residence of afamily member and the right to decide marriages and adoptions. Accordingly, as headsof households, husbands and fathers were legally permitted to remove members fromthe family register if they became married or adopted a child without their permission.Parental authority in principle rested solely with the father, only to be exercised bythe mother when a father was unable to do so. When mothers carried out financialmanagement or other legal acts related to property on behalf of a child, the agreement ofa family council ( shinzoku kai ; family members selected by the court) was required (seeWakita et al. (2011, pp. 200–202); Kurushima et al. (2015, pp. 170–171)).The family system ( ie seido ) was maintained as prescribed by the former civil code,and women did not achieve the right to vote, even in the interwar period. The rights ofhousehold heads were exceptionally strong in prewar Japan because of a family systemlegitimized by law. Therefore, the position of women remained low in the lead-up toWorld War II.On August 14, 1945, Japan accepted the Potsdam Declaration and its defeat in WorldWar II, after which large-scale democratization policies were set in motion by the GeneralHeadquarters of the Allied Powers. Through these policies, Japanese women achievedequal status to men under law, the right to vote, and educational opportunities equalto those of men. This section summarizes the rights gained by women as a result ofJapan’s democratization. On October 11, 1945, General MacArthur issued a directive tothe cabinet of Kijiro Shidehara to implement the “Five Great Reforms.” These pertainedto (1) The liberation of women, (2) the right of workers to organize, (3) the liberalizationof education, (4) the abolition of autocratic governance, and (5) the democratization ofthe economy. The granting of woman’s suffrage received particularly strong attention andbecame one of the earliest rights achieved by women as a result of the reforms (Kanzaki2009, p. 19). Indeed, a Revised General Election Law implemented in 1945 enfranchisedall citizens above the age of 20. The 22nd general election for the House of Representativesheld on April 10, 1946 was the first election in which women exercised their right to vote,resulting in the election of 39 women to the Diet (Wakita 2011, p. 275). The new Constitution of Japan was promulgated on November 3, 1946 and came intoeffect on May 3, 1947. The Constitution included provisions for the dignity of the individ-ual (Article 13), equality under law (Article 14), the essential equality of men and women Immediately after the war, Fusae Ichikawa, who had been active in the prewar women’s suffragemovement, founded the Women’s Committee on Post-war Policy. With around 70 female members, thecommittee articulated to the government their demands for women’s suffrage (see Kanzaki (2009, pp.19–22)). Specifically, Article 14 prohibits discrimination in“political, economic or social relations because of sex.” Article 24, establishing the prin-ciple of individual dignity and the essential equality of the sexes within the family, thesmallest unit of society, clearly states that marriage “shall be based only on the mutualconsent of both sexes and . . . maintained through mutual cooperation with the equalrights of husband and wife as a basis.” This stands in stark contrast to the Meiji era civilcode, which constrained the rights of a woman compared with her husband (see Yuzawa2012, p. 48).Based on the fundamental principles of the new Constitution, in December 1947, theFamily Laws enshrined in Part 4 (Relatives) and Part 5 (Inheritance) of the Civil Codewere completely overhauled. As a result, the old Japanese family system ( ie seido ) andthe rights of householders ( kosyu ken ) were abolished, and inheritance by new heads ofhouseholds was replaced in favor of the equal distribution of inheritance. Patriarchy asa family system was thus eliminated and the position of women in relation to marriage,family relations, and inheritance was raised (see Wakita et al. (2011, pp. 276–277); Ku-rushima et al. (2015, pp. 232–233)). Furthermore, the Fundamental Law of Educationimplemented in March 1947 provided for equal educational opportunity without discrimi-nation on the basis of sex or social status and the principle of co-education. The enactmentof the Labor Standards Law in April 1947 also prohibited the payment of lower wagesto women than to men on the basis of their gender. Thus, the revised Civil Code’s Nonetheless, gender biases remaining in current laws should be noted. For instance, Article 731 ofthe Civil Code establishing marriageable age states that “a man who has attained 18 years of age, and awoman who has attained 16 years of age may enter into marriage.” Article 733 establishing a period ofprohibition of remarriage: “A woman may not remarry unless six months have passed since the day ofdissolution or rescission of her previous marriage.” Article 177 of the Penal Code dealing with the crimeof rape states that “a person who, through assault or intimidation, forcibly commits sexual intercoursewith a female of not less than 13 years of age commits the crime of rape and shall be punished byimprisonment with work for a definite term of not less than three years. The same shall apply to a personwho commits sexual intercourse with a female under 13 years of age.” (See Kurushima et al. (2015, pp.232–233)). The specific articles in the Labor Standards Law that raised the position of female workers were Article3 (Equal Treatment), Article 4 (Principle of Equal Wages for Men and Women), Article 60 (WorkingHours and Days Off for Girls), Article 63 (Night Work and Restrictions on Dangerous and HarmfulJobs), Article 64 (Ban on Belowground Labor), Article 65 (Before and After Childbirth), Article 66(Time for Child Care), Article 67 (Menstrual Leave), and Article 68 (Traveling Expenses for ReturningHome). Article 4 most clearly expresses the principle of gender equality in the workplace, prohibiting thepayment of lower wages to women because of their gender when employed in the same type of occupationand with the same abilities as men. For more about the Labor Standards Law, see Kanzaki (2009, pp.71–103)
Appendix B Data Appendix
B.1 Marriage Status
The prefecture-year-age-level data on the number of single, married, divorced, and wid-owed people are from the official reports of the 1950 and 1955 Population Censuses (Bu-reau of Statistics, Office of the Prime Minister 1951a, 1956a). For the data on the numberof male divorces and widowhoods, we replace the few hyphened observations of ages lessthan 20 in Yamanashi with zero because they are considered to be typos. The prefecture-year-age-level data on the number of women and men (denominator of the proportion ofsingle, married, divorced, or widowed people) are also from the 1950 and 1955 PopulationCensuses (Bureau of Statistics, Office of the Prime Minister 1951a, 1956a).
B.2 Sex Ratio
The prefecture-year-age-level data on the number of women and men are from the 1950and 1955 Population Censuses (Bureau of Statistics, Office of the Prime Minister 1951a,1956a). Figure 1 shows the sex ratio in the census years. To check the impacts of thewartime losses of men on the sex ratio, we also digitize the 1930 and 1935 PopulationCensuses (Statistics Bureau of the Cabinet 1933, 1939a). Figure B.1 illustrates the sexratios in 1930 and 1935 by age, confirming that there were no dramatic declines in the sexratio in either year. In the empirical analyses, we use the adjusted sex ratio calculatedfrom equation 1. Figure B.2 presents the adjusted sex ratio by prefecture, year, and age.
B.3 Age at First Marriage, Marital Fertility, and Stillbirth
Prefecture-year-level data on average age at first marriage are from the official reportsof the 1935, 1950, and 1955 Population Censuses (Bureau of Statistics, Office of thePrime Minister 1951b, 1956b; Statistics Bureau of the Cabinet 1939b). Prefecture-year-level data on the number of live births and stillbirths are from the official reports of1935, 1950, and 1955 Vital Statistics of Japan (Division of Health and Welfare Statistics,Welfare Minister’s Secretraiat 1953, 1957; Statistics Bureau of the Cabinet 1936). Thedenominator of the marital fertility rate (number of married women) is obtained from4 .700.720.740.760.780.800.820.840.860.880.900.920.940.960.981.001.021.041.061.081.10 M a l e s / F e m a l e s Age measured in 1930 (a) 1930 Population Census M a l e s / F e m a l e s Age measured in 1935 (b) 1935 Population Census
Figure B.1: Sex Ratios by Age Measured in 1930 and 1935Population Censuses
Notes: The sex ratio is defined as the number of men divided by the number of women. All those rates arethe national averages based on the 1930 and 1935 Population Censuses. Source: Created by the authorsusing Statistics Bureau of the Cabinet (1933) and Statistics Bureau of the Cabinet (1939a). A d j u s t ed S e x R a t i o Age measured in 1950 (a) 1950 Population Census A d j u s t ed S e x R a t i o Age measured in 1955 (b) 1955 Population Census
Figure B.2: Adjusted Sex Ratios by Prefecture Measured in the 1950 and 1955Population Censuses
Notes: The adjusted sex ratio is defined as in equation 1. All those rates are the national averages based onthe 1950 and 1955 Population Censuses. Source: Created by the authors using Bureau of Statistics, Officeof the Prime Minister (1951a) and Bureau of Statistics, Office of the Prime Minister (1956a). a b l e B . : Su mm a r y S t a t i s t i c s : P r e f ec t u r e - Y e a r - L e v e l P a n e l D a t a s e t f r o m t h e , , a nd P o pu l a t i o n C e n s u s S t a t i s t i c s P a n e l A : D e p e nd e n t V a r i a b l e s Y e a r O b s e r v a t i o n s M e a nS t d . D e v . M i n i m u m M a x i m u m A v e r ag e A g e a t F i r s t M a rr i ag e W o m e n . . . . . . . . . . . . M e n . . . . . . . . . . . . M e n - W o m e n . . . . . . . . . . . . M a r i t a l F e r t ili t y a ndS t ill b i r t h R a t e s M a r i t a l f e r t ili t y r a t e ( li v e b i r t h s p e r , m a rr i e d w o m e n ) . . . . . . . . . . . . S t ill b i r t h r a t e ( s t ill b i r t h s p e r , b i r t h ) . . . . . . . . . . . . P a n e l B : I n t e n s i t y V a r i a b l e Y e a r O b s e r v a t i o n s M e a nS t d . D e v . M i n i m u m M a x i m u m S e x r a t i o ( M e n W o m e n ) . . . . . . . . . . . . P a n e l C : C o n t r o l V a r i a b l e Y e a r O b s e r v a t i o n s M e a nS t d . D e v . M i n i m u m M a x i m u m F e m a l e l a b o r f o r ce p a r t i c i p a t i o n ( w o r k e r s p e r w o m e n ) . . . . . . . . . . . . N o t e s : T h i s t a b l e r e p o r t s t h e s u mm a r y s t a t i s t i c s f o r t h e p r e f ec t u r e - y e a r - l e v e l ce n s u s a nd v i t a l s t a t i s t i c s o f , , a nd . P a n e l s A , B , a nd C s h o w t h e s u mm a r y s t a t i s t i c s f o r t h e d e p e nd e n t v a r i a b l e s ,i n t e n s i t yv a r i a b l e , a nd c o n t r o l v a r i a b l e , r e s p ec t i v e l y . S o u r ce s : S t a t i s t i c s B u r e a u o f t h e C a b i n e t( b ) , S t a t i s t i c s B u r e a u o f t h e C a b i n e t( ) , B u r e a u o f S t a t i s t i c s , O ffi ce o f t h e P r i m e M i n i s t e r ( b ) , B u r e a u o f S t a t i s t i c s , O ffi ce o f t h e P r i m e M i n i s t e r ( ) , B u r e a u o f S t a t i s t i c s , O ffi ce o f t h e P r i m e M i n i s t e r ( b ) , B u r e a u o f S t a t i s t i c s , O ffi ce o f t h e P r i m e M i n i s t e r ( ) , S t a t i s t i c s B u r e a u o f t h e C a b i n e t( ) , S t a t i s t i c s B u r e a u o f t h e C a b i n e t( ) , D i v i s i o n o f H e a l t h a nd W e l f a r e S t a t i s t i c s , W e l f a r e M i n i s t e r ’ s S ec r e t r a i a t( ) , a nd D i v i s i o n o f H e a l t h a nd W e l f a r e S t a t i s t i c s , W e l f a r e M i n i s t e r ’ s S ec r e t r a i a t( ) . N u m be r o f li v e b i r t h s pe r f e m a l e s Maternal age measured in 1950
LegitimateOut-of-wedlock (a) 1950 Population Census N u m be r o f li v e b i r t h s pe r f e m a l e s Maternal age measured in 1955
LegitimateOut-of-wedlock (b) 1955 Population Census
Figure B.3: Legitimate and out-of-wedlock birth rates by maternal agemeasured in the 1950 and 1955 Population Censuses
Notes: The legitimate birth rate is the number of legitimate live births per 1,000 women. The out-of-wedlock birth rate isthe number of out-of-wedlock live births per 1,000 women. All the rates are the national averages based on the 1950 and1955 Population Censuses. Prefecture-year-age-level statistics are not available for either legitimate or out-of-wedlock livebirth rates. Source: Created by the authors using Bureau of Statistics, Office of the Prime Minister (1951a) and Bureauof Statistics, Office of the Prime Minister (1956a).
B.4 Out-of-Wedlock Births
The age-year-level data on the number of out-of-wedlock live births are from the 1950and 1955 Population Censuses (Bureau of Statistics, Office of the Prime Minister 1951b,1956b). The data on the denominator for the share of out-of-wedlock births (numberof live births) and out-of-wedlock birth rate (number of women) are also from the 1950and 1955 Population Censuses. Figure B.3 shows the legitimate birth and out-of-wedlockbirth rates in the census years. Figure B.4 shows the out-of-wedlock birth share in thecensus years, confirming that most live births are within a marriage.
B.5 Female Labor Force Participation
The female labor force participation rate is the number of women working in any sectorper 100 women. The data on the number of female workers in 1950 and 1955 are obtainedfrom the censuses (Bureau of Statistics, Office of the Prime Minister 1951b, 1956b). Thedata on the number of female workers in 1935 are linearly interpolated using the 1930and 1940 Population Censuses (Bureau of Statistics Office of the Prime Minister 1961;Statistics Bureau of the Cabinet 1935). Panel C of Table B.1 shows the summary statisticsby year. 8 .001.002.003.004.005.006.007.008.009.0010.0011.0012.0013.0014.0015.00 O u t - o f - w ed l o ck li v e b i r t h s pe r li v e b i r t h s Maternal age measured in 1950 (a) 1950 Population Census O u t - o f - w ed l o ck li v e b i r t h s pe r li v e b i r t h s Maternal age measured in 1955 (b) 1955 Population Census
Figure B.4: Out-of-wedlock birth share by maternal agemeasured in the 1950 and 1955 Population Censuses
Notes: The out-of-wedlock birth share is defined as the number of out-of-wedlock live births per 100 live births. All therates are the national averages based on the 1950 and 1955 Population Censuses. Prefecture-year-age-level statistics arenot available for the out-of-wedlock live birth share. Source: Created by the authors using Bureau of Statistics, Office ofthe Prime Minister (1951a) and Bureau of Statistics, Office of the Prime Minister (1956a).
Table B.2: Summary Statistics: Number of Deaths and Missing Homefront People (mea-sured in 1948)
Year Observations Mean Std. Dev. Minimum MaximumNumber of deaths and missing people 1948 46 7 , . , . , Notes: Note: This table reports the summary statistics for the prefecture-level data on the number of deaths and missinghomefront people measured in 1948. Source: Nakamura and Miyazaki (1995).
B.6 Deaths and Missing People in the Homefront Population
We digitize the statistics on the number of deaths and missing people in the homefrontpopulation as well as the number of people in 1944 using Nakamura and Miyazaki (1995,pp. 279–281). The data do not include the number of injured people. The number ofdeaths and missing people were concentrated in Tokyo, Hiroshima, and Nagasaki prefec-tures, which experienced substantial air attacks and atomic bombings. Table B.2 presentsthe summary statistics.
B.7 The Asahi Shimbun (Newspaper)
The Asahi Newspaper (shinbun) is one of the most popular newspapers in Japan andis read many people. Its past issues have been digitally archived and released online(https://database.asahi.com/index.shtml [in Japanese]).9 ppendix C Empirical Analysis Appendix
C.1 Wartime Losses of Homefront People
Table C.1 presents the results of the specification that includes the number of deathsand missing people in the homefront population (see Online Appendix B.6) in equation 2.Since the losses of homefront people are cross-sectional data measured in 1948, we interactthis variable with the indicator variable that takes one if the year is 1955 to create thewithin variation. Table C.2 presents the results of the specification that includes thenumber of deaths and missing people in the homefront population in equation 3. Thewartime losses of homefront people are interacted with the 1955 dummy in a similar way.The estimates are largely unchanged if we include the female labor force participationrate in all the regressions (not reported). We could not conduct a similar exercise to theage-year-level panel data on out-of-wedlock births because the data on the wartime lossesof homefront people are unavailable by age. Tables C.1 and C.2 show similar results toour main results in Tables 2, 3, and 4. 10 a b l e C . : R o bu s t n e ss C h ec k s : E ff ec t s o f t h e G e nd e r I m b a l a n ce o n M a rr i ag e M a r k e t O u t c o m e s : I n c l ud i n g t h e N u m b e r o f D e a t h s a nd M i ss i n g P e o p l e i n t h e H o m e f r o n t P o pu l a t i o n W o m e n M e n ( )( )( )( )( )( )( )( ) S i n g l e M a rr i e d D i v o r ce d W i d o w h oo dS i n g l e M a rr i e d D i v o r ce d W i d o w h oo d P a n e l A : O v e r a ll E ff ec t s S e x R a t i o ( S R ) . − . − . . . − . − . . ( . )( . )( . )( . )( . )( . )( . )( . ) P a n e l B : H e t e r og e n e o u s E ff ec t s S e x R a t i o ( S R ) . − . − . − . . − . − . . ( . )( . )( . )( . )( . )( . )( . )( . ) S e x R a t i o ( S R ) × I ( Y e a r = ) − . − . . . − . . . − . ( . )( . )( . )( . )( . )( . )( . )( . ) *** , ** , a nd * r e p r e s e n t s t a t i s t i c a l s i g n i fi c a n ce a tt h e % , % , a nd % l e v e l s , r e s p ec t i v e l y . S t a nd a r d e rr o r s f r o m t h ec l u s t e r - r o bu s t v a r i a n cee s t i m a t i o n r e p o r t e d i np a r e n t h e s e s a r ec l u s t e r e d a tt h e - p r e f ec t u r e l e v e l. N o t e s : T h e d e p e nd e n t v a r i a b l e s u s e d i n c o l u m n s ( ) – ( ) a r e t h e p r o p o r t i o n o f s i n g l e m e n , m a rr i e d m e n , d i v o r ce d m e n , a nd w i d o w e d m e np e r , m e n , r e s p ec t i v e l y . T h e d e p e nd e n t v a r i a b l e s u s e d i n c o l u m n s ( ) – ( ) a r e t h e p r o p o r t i o n o f s i n g l e w o m e n , m a rr i e d w o m e n , d i v o r ce d w o m e n , a nd w i d o w e d w o m e np e r , w o m e n , r e s p ec t i v e l y . I ( Y e a r = ) i nd i c a t e s a n i nd i c a t o r v a r i a b l e t h a tt a k e s o n e i f t h e y e a r i s nd ze r o i f t h e y e a r i s ,i. e ., a1955 y e a r du mm y t h a t d e p e nd s o n l y o n t . T h e nu m b e r o f o b s e r v a t i o n s i s , ( p r e f ec t u r e s × - ag e r a n g e × ce n s u s y e a r s ) i n a ll t h e r e g r e ss i o n s . A ll t h e r e g r e ss i o n s i n c l ud e t h e nu m b e r o f d e a t h s a nd m i ss i n g p e o p l e i n t h e h o m e f r o n t p o pu l a t i o n ( m e a s u r e d i n ) i n t e r a c t e d w i t h t h e du mm y , p r e f ec t u r e - ag e - s p ec i fi c fi x e d e ff ec t s , a nd y e a r fi x e d e ff ec t s . A ll t h e r e g r e ss i o n s a r e w e i g h t e db y t h e a v e r ag e nu m b e r o f p e o p l e ( w o m e n i n c o l u m n s ( ) – ( ) ; m e n i n c o l u m n s ( ) – ( )) i n e a c hp r e f ec t u r e - ag ece ll. Average age at first marriage1935–1950 Census Data 1935–1955 Census Data(1) (2) (3) (4) (5) (6)Panel A Women Men Difference Women Men DifferenceSex Ratio 5 . . − .
196 1 .
318 1 . − . . . . . . . − . − . − . − . . . . . ***, **, and * represent statistical significance at the 1%, 5%, and 10% levels, respectively. Standard errors fromthe cluster-robust variance estimation reported in parentheses are clustered at the 46-prefecture level.Notes: The sex ratio is the number of men aged 17–50 divided by the number of women aged 15–40. In Panel A,the dependent variable used in columns (1) and (4) is average age at first marriage for women, whereas that incolumns (2) and (5) is average age at first marriage for men. In Panel A, the dependent variable used in columns(3) and (6) is the difference in average age at first marriage between women and men (women minus men). InPanel B, the dependent variable used in columns (1) and (3) is the number of live births per 1,000 married women,whereas that in columns (2) and (4) is the number of stillbirths per 1,000 births. All the regressions include thenumber of deaths and missing people in the homefront population (measured in 1948) interacted with the 1955dummy, prefecture fixed effects, and year fixed effects. I(Year=1955) indicates an indicator variable that takesone if the year is 1955 and zero if the year is 1950, i.e., a 1955 year dummy that depends only on t .In Panels A and B, the number of observations is 92 (46 prefectures × ×2 census years) in all the regressions.