Bongaarts 2002 — The End of the Fertility Transition in the Developing World

Summary

Bongaarts examines fertility transitions in 143 developing countries (United Nations (UN) 1950–2000 data) to identify regularities that can guide projections. The central finding is that the pace of fertility decline decelerates systematically as countries progress through the transition, and that reaching near-replacement fertility requires high levels of human development (life expectancy $\approx 75$ + literacy $\approx 95\%$) that most developing countries will not achieve by 2025. The paper challenges the UN's assumption of uniform convergence to total fertility rate (TFR) $= 2.1$, predicting substantial cross-country variation and a "nontrivial" number of stalling transitions.

Key Claims

• Pace deceleration is the dominant empirical regularity. Annual pace of TFR decline falls from $\approx 0.15$ at TFR $4$–$6$ to $\approx 0.04$ at TFR $2.0$–$2.5$; countries entering the transition in the 1960s saw pace drop from $0.15 \to 0.06$ over three decades.
• Later entrants start at lower development levels. Life expectancy at transition onset fell from $57.7$ (1960s cohort) to $48.7$ (1990s cohort); literacy from $61.6\%$ to $41.3\%$. Because higher development at onset predicts faster initial pace, later-entering transitions are doubly slower: earlier onset and lower starting pace.
• Late-transition fertility is driven by life expectancy and literacy only. Ordinary least squares (OLS) regression ($R^2 = 0.60$) of TFR on five development indicators, years of observation, and regional dummies: only life expectancy ($-0.104$, $p < 0.001$) and literacy ($-0.015$, $p = 0.003$) are significant. Gross domestic product (GDP) per capita, percent urban, and percent in agriculture are not significant.
• Replacement requires high human development. Only $20$ of $137$ countries had life expectancy $>70$ and literacy $>90\%$ in 1995–2000; this group averaged TFR $2.24$. Countries with below-replacement TFR $\leq 2.1$ ($n = 21$, avg TFR $1.76$) have similar development profiles but are a selected group — their experience is not a reliable guide for others.
• Fertility stalling is a real possibility. A nontrivial number of developing countries will plateau above replacement for decades, as Argentina and Uruguay did since the 1950s (TFR still $\approx 3$ in 1995–2000). Demographic and Health Surveys (DHS) data from 9 countries with two surveys in the 1990s show average pace of $0.048$/yr vs. UN estimate of $0.073$/yr for the same countries; $5$ of $9$ show pace $\leq 0.03$/yr.
• Three-phase causal structure. Pretransition: natural fertility, unresponsive to development. Early transition: diffusion and social interaction release pent-up demand for family limitation — faster the more developed a country is at onset. Mid/late transition: fertility closely tied to development (principally life expectancy and literacy), with social interaction playing a diminished independent role.
• Tempo effects acknowledged but deferred. Rising age at first marriage and first birth in developing countries will depress period TFR below completed fertility; full analysis beyond the scope of this paper. The implication is that observed period TFRs may overstate quantum decline.
• UN projections broadly correct on average, with offsetting errors. The UN's linear-interpolation method will over-project for countries that stall above replacement and under-project for countries that fall below 2.1 — the errors roughly offset, making the 2020–25 unweighted developing-world average ($\approx 2.8$) approximately right, but the country-level distribution will be far wider than projected.

Concepts Introduced or Extended

• Fertility Stalling
• Demographic Transition (three-phase causal structure: natural fertility → diffusion-driven → development-driven)
• Tempo Effects in Fertility (developing-world application: rising age at first birth depresses period TFR)
• Stochastic Fertility Forecasting (systematic deceleration bias not captured by linear interpolation)

Entities Mentioned

• John Bongaarts

Quotes

"It would not be surprising if fertility in a nontrivial number of developing countries were to stall above replacement for a few decades."

"Reaching near-replacement fertility apparently requires high levels of human development in most countries."

"Although the preceding analysis has identified factors that will make these projections too high for some countries and too low for others, the average trend for the next quarter-century seems about right."

My Take

The paper's empirical contributions — pace deceleration, development-at-onset trends, and the regression identifying life expectancy and literacy as the only significant determinants of late-transition fertility — are robust and influential. The three-phase causal taxonomy (natural fertility → diffusion-driven → development-driven) is a useful conceptual organizing framework. The limitation is that the analysis ends at 1995–2000, before the 2000s wave of stalling transitions in sub-Saharan Africa became widely documented. The causal mechanism behind pace deceleration (diffusion exhaustion vs. development slowdown) is identified but not formally distinguished. The tempo-effect caveat is flagged but deferred — a gap, since period TFR may overstate the quantum decline particularly in countries with rising age at first birth.