Summary
Heuveline (2002) analyzes cause-specific mortality at ages 15–34 (suicide, homicide, motor vehicle injuries, HIV/AIDS) across 26 countries from 1950 to 1996, using the World Health Organization (WHO) Mortality Database and age-sex standardization. The paper documents that overall young adult mortality declined by one-third between the late 1950s and early 1990s, while the combined share attributable to the three violent causes rose from 26% to 43%. A principal components analysis (PCA) on the full age-sex-cause matrix summarizes international variation in two interpretable dimensions — aggregate violent mortality level ("control") and the mix between intentional and unintentional causes ("motivation") — and reveals the United States as a dramatic outlier among industrialized nations.
Key Claims
- Overall standardized mortality at ages 15–34 fell ~33% from the late 1950s to the early 1990s; the three violent causes combined increased 9% over the same period, driving their share of young adult deaths from 26% to 43%.
- Motor vehicle injury: dominant cause at ages 15–34 in most countries (except Mexico); inverted U-shaped trend — peaked 1965–74 in 20 of 26 countries; declined sharply thereafter; lowest rates in Nordic countries and UK.
- Homicide: standardized rates nearly doubled from the late 1950s to the early 1990s across the full sample; enormous cross-national variance (US rate 14.06/100k vs. Japan's 1.2/100k and Finland's 2.88/100k as the highest European rate); Latin American countries (Mexico 30.85, Venezuela 19.39) far exceed all others; sex ratios highest in Latin America (Mexico M/F = 13.1).
- Suicide: non-monotonic trend (down 1960s, up 1970s, down 1980s–90s); high country variance; male/female sex ratio rose from 2.0 (1955–64) to 3.8 (1985–94) as female rates declined and male rates increased except at ages 20–24; Catholic countries have lowest rates.
- Sex ratios diverge by cause: motor vehicle M/F ratio fell from 6.5 to 4.1 (1955–64 to 1985–94) while suicide M/F ratio rose from 2.0 to 3.8 — convergence only in the sense that the highest and lowest ratios moved toward an intermediate range.
- Catholic underreporting: negative correlation between suicide and undetermined-cause mortality is detected across the full sample, but is almost entirely explained by the three Latin American countries. European Catholic countries (Ireland, Italy, Spain, Portugal) do not show elevated rates of undetermined or other unintentional injury deaths — their low suicide rates appear real, not artifacts of misclassification.
- HIV/AIDS: annualized HIV incidence ratio approaches or exceeds three-cause violent mortality rates in Spain (40.68 vs. 28.18), Portugal (43.28 vs. 41.83), Italy (23.23 vs. 27.74), US (48.29 vs. 59.29), and Switzerland (25.52 vs. 43.45); Russia has highest three-cause combined rate of all 30 countries post-1980 (74.52/100k), with homicide rising +115% 1980s to early 1990s.
- PCA structure: 32-dimensional mortality space (4 age groups × 2 sexes × 4 causes) collapses to 5 components explaining 95% of variance; PC1 (35%) separates suicide from homicide/other unintentional; PC2 (24%) aggregates violent causes vs. genuinely accidental. PCA temporal trend: worst 1970–1981 (baby boom cohort entering 15–35 window) — consistent with Easterlin's relative cohort size argument.
- U.S. position: clear outlier on both PCA dimensions — extremely high combined violent mortality and heavy homicide loading; below Ukraine and Russia post-1980 only when the Russian Federation/Ukraine data are included.
Method
- Data: WHO Mortality Database; 26 countries with ≥35 years of usable data 1950–1996 (1,128 country-year observations); International Classification of Diseases (ICD)-6 through ICD-10 tracked with concordance table.
- Standardization: direct age-sex standardization using the average sex-age distribution in the database as the standard.
- HIV: WHO cause-specific data not used (quality concerns); Joint United Nations Programme on HIV/AIDS (UNAIDS) end-1997 estimates annualized over 18 years.
- PCA: on the 32-variable (age × sex × cause) rate matrix; first 5 components retain 95% of variance; factors interpreted by loadings.
Concepts Introduced or Extended
- Young Adult Mortality Hump — precursor analysis using fixed age range (15–34) and sex-age standardization; documents that violent causes constitute the majority of hump-level excess mortality; predates the Sum of Smooth Exponentials (SSE) decomposition method; illustrates the exact measurement limitation (fixed age range) that Remund et al. later resolved.
- Sociology of Suicide — cross-national evidence on the Catholic underreporting controversy (Latin American but not European Catholic artifact), sex ratio trends (rising male/female ratio since 1960s), and the relative cohort size interpretation of the 1970s suicide peak.
Entities Mentioned
Quotes
"The contribution of these three causes of death to overall mortality between ages fifteen and thirty-five has risen from 26 percent in the late 1950s to 43 percent from the late 1970s on."
"The distribution of mortality rates from homicide is much more skewed. [...] three American countries (the United States, Venezuela, and Mexico) clearly stand out for their mortality rates from homicide, five, seven, and eleven times higher, respectively, than the highest rate in a European country (Finland)."
"The correlation [between suicide and undetermined-cause mortality] was almost entirely due, however, to the three Latin American countries. [...] low mortality rates from suicide in European Catholic countries seem to reflect a real difference in the incidence of fatal suicide rather than a statistical artifact."
My Take
This is a thorough descriptive-comparative paper that served as an important reference point for cross-national young adult mortality research before the SSE method (Remund et al. 2018) made it possible to isolate true excess hump mortality from senescent overlap. Heuveline's use of fixed age ranges (15–34) and standardized rates is well-executed within its own frame but carries the conceptual limitation that Remund et al. later critiqued. The PCA is a genuinely innovative contribution for the period, producing the "control" and "motivation" axes that remain interpretively compelling. The U.S. outlier finding is strikingly sharp. The HIV section is creative but the annualized-ratio construction is a rough approximation. The relative cohort size argument is suggestive but the timing overlap (baby boomers entering the 15–35 window 1964–1981) is asserted rather than tested with a formal Easterlin-type model.