Japan in 1995: Delayed Marriage Is the Core of Falling Births
- Main cause
- The AI sees postponed and fewer marriages as the main cause of Japan's falling births.E-06E-07E-02E-11
The AI's own summary of its diagnosis and policy. Figures are checked against the evidence.
Live births (Japan) 1973 → 1994 −41%
The AI was given only the data that had been published by the end of 1995 (annual statistics up to 1994). It may nonetheless know about later events and outcomes from its training data. Please read the forecast part with that in mind.
In the latest annual data published by the end of 1995 (for 1994), Japan had about 1.24 million births and a total fertility rate of 1.50. Births were about 41% fewer than in 1973. The rate was slightly higher than in 1993, the year before.
Marriages fell about 37% between 1972 and 1987, and births fell by almost as much between 1973 and 1987. By contrast, between 1987 and 1994 marriages rose about 12% while births fell about 8%. After marriages rose in 1993 on the year before, births also rose in the following year, 1994.
＋6
In 1994 Japan's total fertility rate was 1.50, below Sweden, the United States, France and South Korea and above Germany (World Bank figures for countries other than Japan). Female labour-force participation that year (ILO estimates published by the World Bank) was about 50% in Japan; it was higher in Sweden and the United States and slightly lower in France, South Korea and Germany. This comparison alone does not establish cause and effect.
＋9
Sweden's total fertility rate rose sharply from 1983 to 1990 and then fell through 1994. During that fall, unemployment rose from about 3% in 1991 to about 10% in 1994. Germany's rate fell sharply in the early 1970s and over the 20 years from 1975 to 1994 stayed at a low level, averaging 1.38 (all World Bank figures).
＋34
The main measures in place by the end of 1995 were as follows (ministry presumed for the childbirth grant and the childcare-leave benefit). The child allowance for children under age 3 paid 5,000 yen a month per child for a first or second child (under an amendment in force from January 1992). The childbirth lump-sum grant was created in October 1994. In December 1994 an agreement among the Ministers of Health and Welfare, Finance and Home Affairs set up the emergency five-year childcare programme, with the goal of raising childcare places for children under age 3 to 600,000 by FY1999. In a separate decision that same December, the Ministries of Education, Health and Welfare, Labour and Construction drew up the Angel Plan. In April 1995 the employment-insurance childcare-leave benefit began, paying 25% of wages in total, including the part paid after a return to work.
Marriages fell about 37% between 1972 and 1987, and births fell by almost as much between 1973 and 1987. After marriages rose in 1993 on the year before, births also rose in the following year, 1994. Working on the premise that most children in Japan are born to married couples (the materials contain no statistics on births outside marriage, so this is the AI's premise), the AI judges that the number and timing of marriages drive the total number of births. Yet between 1987 and 1994 marriages rose about 12% while births fell about 8%, and the ratio of births to marriages also fell. Because the evidence is only the co-movement of totals, the AI rates its confidence as medium.
In 1994 female labour-force participation (ILO estimates) was about 50% in Japan; Sweden and the United States, whose fertility rates were higher than Japan's, also had higher participation. Germany, whose fertility rate was lower than Japan's, also had lower participation. Among these six countries there is no pattern in which higher female employment goes with lower fertility. In Japan the right to childcare leave took effect only in April 1992, and fixed-term and day workers remain excluded. A benefit during leave (25% of wages, including the part paid after a return to work) began only in April 1995. The AI believes that the remaining either-or choice between staying in work and having children is delaying marriage and childbirth. But the evidence is a single-year comparison of six countries, and the materials contain no statistics on Japanese women leaving work at childbirth, so the AI rates its confidence as weak.
As of the end of 1995 the child allowance covers only children under age 3, at 5,000 yen a month for a first or second child. The 300,000 yen childbirth lump-sum grant was set with reference to items such as the average delivery fee at national hospitals and the cost of check-ups around birth, so it is aimed at the cost of childbirth itself. The national cash benefits in the materials stop at age 3; no scheme in the materials is aimed at the costs of raising a child from age 3 onward, so households carry them. The AI believes this weighs on decisions about second and third children, but rates its confidence as weak because the materials contain no statistics on household child-rearing costs.
Real GDP growth turned negative in 1993, and unemployment rose between 1991 and 1994. Yet even while real growth was high in 1985–1989, the fertility rate fell from 1.76 in 1985 to 1.57 in 1989. The AI sees economic and job insecurity as at most a factor that slightly deepened the decline of the 1990s.
The 18 billion yen of national money for the emergency five-year childcare programme in FY1995 is only just over one-thousandth of the Ministry of Health and Welfare's budget for that year (including supplementary budgets). The AI judges that expanding childcare is the right direction, but that at this scale it will not change the course of births.
Over the 20 years from 1975 to 1994, Germany's fertility rate stayed at a low level, averaging 1.38. Assume that Japan's rate settles around that 1.38 by around 2025. Births also depend on the size of the parent generation: people born in 1990–1994 number only about 0.7 times those born in the 1960s, the generation that was of parenting age in 1994. Assuming those born from 1995 onward match the 1990–1994 level and allowing for this, births by around 2025 would be about 810,000. The AI considers Scenario A the most likely, because nothing has yet been done to stop the postponement of marriage, and Germany has already spent 20 years at a low level.
If later marriage turns out to be mere postponement and births in women's thirties catch up, the fertility rate would stop falling at around 1.51, the 1990–1994 average. In that case births by around 2025, estimated the same way as in Scenario A, would be about 890,000. Even if the rate stops falling, births are expected to decline because the parent generation shrinks.
Sweden's fertility rate rose by a net 0.27 between 1983 and 1994. Assuming Japan regains that much through a large expansion of childcare and childcare leave, the rate would be about 1.77 by around 2025, and births, estimated the same way as in Scenario A, would be about 1.04 million. However, Sweden's fall after 1990 coincided with rising unemployment, which suggests that a recovery can be undone by the economy.
Japan's fertility rate fell by 0.31 in the ten years from 1984 to 1994 (with an uptick in 1994 from the year before). Assuming a fall of the same size happens over one more decade and then stops, the rate would be about 1.19 by around 2005 and would stay around that level by around 2025. Births, estimated the same way as in Scenario A, would be about 700,000.
THOUGHT EXPERIMENT
Show the calculation
Public money in each of FY1996 and FY1997, double that of FY1995: 36 billion yen × 2 = 72 billion yen. National share at half: 72 billion yen ÷ 2 = 36 billion yen. Extra national money: 36 billion yen − 18 billion yen = 18 billion yen. Share of the ministry budget: 18 billion yen ÷ 14.9556 trillion yen × 100 = about 0.12
Finish the five-year childcare programme two years early
- Who, how much, how long
- Bring forward the targets of the emergency five-year childcare programme (FY1995–FY1999) by two years, so that the goal of raising childcare places for children under age 3 to 600,000 is met by FY1997. If FY1995 proceeds as planned, doing the rest in the two years FY1996 and FY1997 requires twice the planned pace (estimate). Assuming costs rise in proportion to the pace of expansion, public money in FY1996 and FY1997 would be double that of FY1995, at 72 billion yen a year, with the national share at half of that, 36 billion yen a year (estimate).
- Funding
- Paid from the general account. National money rises to 36 billion yen a year, and the extra 18 billion yen a year is equal to about 0.12% of the Ministry of Health and Welfare's FY1995 budget (including supplementary budgets) (estimate). It is covered by slowing the growth of other spending within that ministry's budget, with no new tax. The cost falls on taxpayers and on those who receive the spending whose growth is slowed.
- Ministry
- Ministry of Health and Welfare (childcare), Ministry of Finance (budget), Ministry of Home Affairs (local finance): the three ministries whose ministers agreed the emergency five-year childcare programme
- Target indicator
- 600,000 childcare places for children under age 3 by FY1997. Keep the total fertility rate from falling below 1.51, the 1990–1994 average, until around 2000.
- Side effects / who pays
- Costs need not rise in proportion to the pace of expansion. They would rise per site if land or childcare workers ran short, and fall if existing facilities could be used, so the estimate could be off in either direction. Expanding places quickly may outpace the supply of childcare workers and lower quality. Taxpayers bear the cost, and families that do not use childcare gain nothing directly.
- Evidence
- E-62E-65E-79E-80E-81E-05E-03
Show the calculation
Public money in each of FY1996 and FY1997, double that of FY1995: 36 billion yen × 2 = 72 billion yen. National share at half: 72 billion yen ÷ 2 = 36 billion yen. Extra national money: 36 billion yen − 18 billion yen = 18 billion yen. Share of the ministry budget: 18 billion yen ÷ 14.9556 trillion yen × 100 = about 0.12
Childcare leave for fixed-term workers, with the benefit paid during leave
- Who, how much, how long
- Extend childcare leave and the childcare-leave benefit to fixed-term and day workers, who are now excluded from the right to leave (keeping the condition of at least 12 months of insurance coverage in the 2 years before leave). At the same time, pay in advance during leave the return-to-work benefit (5% of wages) that is now paid only after 6 months or more back at work, so that the amount received during leave rises from 20% to 25% of wages, the same as the current total. The benefit rate itself does not exceed the current total. The law is amended in FY1996 and the change starts in April 1997.
- Funding
- Paid from the existing funding of the employment-insurance childcare-leave benefit (premiums from workers and employers, plus a national treasury share). The cost cannot be computed, because the materials contain no data on how many more people would be covered or on their wages. The cost falls on the workers and employers who pay premiums, and on taxpayers for the treasury share.
- Ministry
- Ministry of Labour (childcare leave and employment insurance; responsibility for the current childcare-leave benefit is presumed in the materials)
- Target indicator
- Without letting female labour-force participation fall below its 1994 level of 50.1%, bring the total fertility rate by around 2010 close to 1.73, France's 1994 level, a country whose female participation was close to Japan's (47.5% in 1994).
- Side effects / who pays
- Paying in advance means the 5% also goes to people who do not return to work, weakening the incentive to return. Small employers may struggle to cover for fixed-term staff on leave. The self-employed and others outside employment insurance are not reached. Whether premiums must rise cannot be decided, because the cost cannot be computed.
- Evidence
- E-89E-88E-64E-18E-21E-15E-03
Put one year's growth in the health and welfare budget into accounts for newborns
- Who, how much, how long
- As a thought experiment, for 10 years from FY1996, an amount equal to the growth of the Ministry of Health and Welfare's budget in FY1995 (about 1.1 trillion yen) is taken each year out of that budget's growth and paid into a "child account" for every child born that year. Divided by the number of births in 1994, this comes to about 880,000 yen per child (estimate). The account cannot be drawn on until age 18 and can be used for education and housing.
- Funding
- No new tax: the use of the budget's growth is reassigned. If a year's growth falls short of this amount, pensions, health care and other spending stay at the previous year's level. The cost falls on the people whose benefits grow more slowly as a result.
- Ministry
- Ministry of Health and Welfare (the scheme), Ministry of Finance (budget and management of the accounts)
- Target indicator
- Raise the total fertility rate to 1.88, Sweden's 1994 level, by around 2025.
- Side effects / who pays
- The share of the population aged 65 and over rose from 7.1% in 1970 to 14.1% in 1994, so any year in which benefits for older people do not grow would concentrate the burden on older people and patients. It is unfair to children born in 1995 or earlier. Because the money arrives 18 years later, it may do little to encourage a decision to have a child, and it may merely shift the timing of births.
- Evidence
- E-65E-99E-01E-13E-100E-101
Show the calculation
Growth of the FY1995 Ministry of Health and Welfare budget (incl. supplementary) on the previous year, 1.0841 trillion yen ÷ births in 1994, 1,238,328 = about 875,000 yen
Amounts and targets are either calculated from the data or marked as estimates, conjectures or design values. The bold option is a thought experiment that puts ideas before feasibility.
The diagnosis that marriage is the main cause faces a strong objection from the data itself. Between 1987 and 1994 marriages rose about 12%, yet births fell about 8%, and the ratio of births to marriages also fell. If the number of marriages determined births, births should have risen during those seven years of more marriages. If married couples have begun having fewer children, the real problem is not marriage but the work-family wall and the costs after marriage. Sweden, a country with high female labour-force participation, raised its fertility rate sharply between 1983 and 1990. The AI thinks the time lag between marriage and childbirth explains part of this gap, but without statistics on births to married couples in the materials, it cannot fully dismiss this objection.
Check (fertility rate): the actual rate in 2025 was 1.14, below every scenario. Scenario A (about 1.38), which the AI had judged most likely, missed by a wide margin. The path also matched none of the scenarios: the rate fell up to 2005, recovered for a while up to 2015 and then fell again by 2025. Germany, the model for Scenario A, and Sweden, the model for Scenario C, both had a rate of about 1.4 in 2024 (World Bank figures).
＋3
Check (births): births in 2025 were about 0.67 million, about 46% fewer than in 1994. The closest was Scenario D (about 700,000), while A, B and C were too high. The parent-generation assumption was close to the actual births of 1995–2000. Applying the actual 2025 rate in the same way as in Scenario A gives about 0.67 million, close to the actual figure. Most of the miss came from the fertility-rate assumption.
＋7
Checking the diagnosis: marriages fell about 38% from 1994 to 2025, so the decline in marriage that the AI ranked first went much further. But births fell about 46% over the same period, more than marriages, and the ratio of births to marriages also fell. The AI believes its view that falling marriage was central was broadly not wrong, but that the decline in births after marriage, raised by the counter-view, was also under way at the same time. On the second-ranked cause, the work-family wall, female labour-force participation rose between 1995 and 2025 while the fertility rate fell to 1.14, so more women in work did not automatically bring a recovery in births.
Checking the policies: Japan in fact raised the childcare-leave benefit rate in stages, and in 2014 set it at 67% for the first 6 months of leave. The child allowance was widened several times in coverage and amount, and in 2024 became 30,000 yen a month for a third or later child. The childbirth lump-sum grant was also raised in stages. In 2001 the Cabinet adopted a policy on supporting work and child-rearing that included a zero-waiting-list strategy for childcare, and in 2023 the Children's Future Strategy was drawn up. On the other hand, the policy timeline in the materials shows neither a decision to bring forward the targets of the emergency five-year childcare programme, the AI's realistic plan, nor any scheme like the bold plan's redirection of growth in the health and welfare budget into child accounts. Over this period the fertility rate came to 1.14 in 2025.
Would it have worked? The AI judges that the realistic plan's faster childcare expansion alone would not have changed the course. In reality, benefit-rate increases and childcare measures were piled up, and the fertility rate recovered to 1.45 between 2005 and 2015, but it stood at 1.14 in 2025. Meanwhile marriages fell between 2015 and 2025, so work-family support did not stop the decline in marriage. Whether the ambitious plan (extending leave to fixed-term workers and paying the benefit in advance) would have worked cannot be checked, because the materials contain no statistics that measure such an effect. As for the bold plan, given that the share of the population aged 65 and over rose to 30% by 2025, the AI believes that continuing to redirect growth in the health and welfare budget to children would have clashed sharply with benefits for older people.
How it is made: claude-opus-5-5 · prompt f4af36acb40a