Billari et al 2006 — Pushing the Limit Long-term Trends in Late Fertility in Sweden

Summary

Using over a century of Swedish vital statistics (1900–2001) and linked population register data, Billari, Kohler, Andersson, and Lundström document a dramatic U-shaped trajectory in births to women aged 40+ (very late fertility), with a trough around 1977–1980 followed by a partial recovery that remains far below historical levels. Using occurrence-exposure rates to isolate first-birth trends among childless women, they show that first-birth rates doubled at ages 40–41 between 1971 and 2001, but that a behavioral wall at ages 44–45 remained essentially flat across three decades of birth cohorts. Analysis of physiological and assisted reproductive technology (ART) constraints establishes that the apparent recovery reflects behavioral change and compositional shifts, and that ART's compensatory potential is modest (adding roughly $7$ percentage points (pp) at age 40), with egg quality as the primary barrier.

Key Claims

• Swedish late fertility (40+) followed a U-shape from a peak of $\approx 115.5$ births per $1{,}000$ women in 1900 to a trough of $8.6$ per $1{,}000$ in 1977, recovering to $26.8$ per $1{,}000$ by 2001 — still far below historical levels.
• Latest-late fertility (45+): trough of $31$ births in 1980, recovered to $89$ by 2001; the share of 45+ within all 40+ births declined from $\approx 10\%$ historically to $\approx 3$–$4\%$ by 2001.
• First-birth occurrence-exposure rates (childless women only) doubled at ages 40–41 from 1971 to 2001 ($0.78 \to 1.84$ per $1{,}000$ childless women); this parity-controlled measure rules out compositional explanations.
• A behavioral wall exists at ages 44–45: first-birth occurrence-exposure rates are essentially flat above age 44 across $30$ birth cohorts (1925–1955), with the absolute probability of first birth between ages 45 and 50 only $0.23\%$ in 2002.
• The absolute probability of a first birth between ages 40 and 45 for a childless 40-year-old was $7.48\%$ in 2002, up from $3.26\%$ in 1970 — a large relative increase, but tiny absolute probability.
• Physiological constraints are stable: female sterility is $\approx 24$–$46\%$ at age 40; ART (in vitro fertilization (IVF) with own eggs) adds only $\approx 7$ pp of cumulative probability at age 40 (Leridon 2004); egg donation maintains $40$–$50\%$ success rates, showing egg quality as the primary barrier, not uterine capacity.
• Rectangularization evidence is weak: period interquartile range (IQR) of the first-birth age schedule ($\approx 6$ years) has been stable; cohort IQR peaked at $7.5$ years for the 1956 cohort and declined only $10.5\%$ to $6.8$ years for the 1961 cohort.
• Compositional factors partly drive the raw 40–44 fertility increase: parity shift ($>50\%$ of 40–44 births are first or second children in Sweden by 2001, up from $\approx 37\%$ in the late 1960s) and changing age structure of women 15–49.
• Extreme age at birth was stable at $\approx 51$ until WWII, declined through the 1970s, then restored post-1980; post-1990 upward push (oldest birth age $57$ in 1993; age $\geq 53$ occurred $4$ times in 1986–2000 vs. only $2$ times in the prior 90 years).
• Denmark had the highest ART utilization in Europe at the time of writing ($>4.2\%$ of births from ART), providing cross-national context on the ceiling of technological fertility compensation.

Concepts Introduced or Extended

• Late and Latest-Late Fertility
• Demographic Transition
• Fertility-Employment Correlation
• Nonmarital Fertility

Entities Mentioned

• Francesco Billari
• Hans-Peter Kohler
• Gunnar Andersson
• Hans Lundström

Quotes

"The overall development of birth rates for women aged 40 years and older has followed a U-shape in Sweden during the 20th century."

"The probability of having a first birth between exact ages 40 and 45 among childless women has almost doubled from 3.26 to 7.48 percent between 1970 and 2002 … [yet] the probability of having a birth at age 45 or above among childless 45-year-old women is extremely low at 0.23 percent in 2002."

"The biological limits to fertility are largely not a result of a decreasing ability to achieve pregnancies through assisted reproduction but rather caused by the declining quality of eggs."

My Take

The paper's central contribution is the occurrence-exposure rate decomposition, which cleanly separates parity-compositional effects from genuine behavioral change at late ages. The behavioral wall at 44–45 — flat across 30 cohorts — is one of the most striking empirical regularities in the paper: regardless of social transformations in fertility timing, the physiological ceiling does not move. The ART evidence (Leridon 2004) is used efficiently to show that the $\approx 7$ pp compensatory ceiling at age 40 is binding well before the behavioral wall. The weakness noted in the paper itself is that rectangularization evidence is ambiguous given the short cohort IQR series available. The paper is a 2006 working paper version; check for the published version (Population and Development Review) for final figures.