Cutler and Meara 2001 — Changes in the Age Distribution of Mortality Over the 20th Century

Summary

Cutler and Meara document the shift in the age distribution of US mortality over the 20th century using three sub-period decompositions: 1900–40 (public health and nutrition), 1940–60 (antibiotics), and 1960–90 (cardiovascular treatment). Each era was dominated by mortality improvements at a different age range — shifting progressively toward older ages — and driven by a different proximate mechanism. The paper provides difference-in-differences (DiD) evidence for the antibiotic contribution (comparing diseases sulfa drugs treated versus those they did not) and attributes the post-1960 cardiovascular disease (CVD) decline almost entirely to medical care rather than behavior change.

Key Claims

• US age-adjusted mortality fell ~67% from 1900 to 1998 (crude rate: 1 in 42 Americans/year → 1 in 125); life expectancy (LE) at birth rose from ~48 to ~80 years.
• Era 1 (1900–40, +15.9 LE years): Dominated by infant and child mortality declines, driven by public health infrastructure (clean water, food safety) and improved nutrition. Table 1: infants contributed +4.7 yrs, children 1–14 contributed +4.4 yrs, young adults 15–34 contributed +3.6 yrs. Elderly contributed only +0.4 yrs.
• Era 2 (1940–60, +6.4 LE years): Dominated by infections in young adults and infants, driven by sulfa drugs (late 1930s) and penicillin (late 1940s). Difference-in-differences: comparing pneumonia/influenza (antibiotic-treatable) against dysentery/diarrheal diseases (not treatable), the net antibiotic effect is approximately −3.8%/year (Table 4). Elderly contributed +1.2 yrs — their first substantial era contribution.
• Era 3 (1960–90, +5.1 LE years): Dominated by elderly mortality, driven by cardiovascular disease treatment. CVD accounted for ~98% of the 1960–90 mortality reduction; CVD death rates fell ~2%/year from the mid-1960s peak, with ~two-thirds of the cumulative decline occurring by 1995. Elderly contributed +1.8 yrs; older adults (35–64) contributed +1.4 yrs.
• Annual mortality decline rates for the elderly: 0.3%/year (1900–40) → 1.1%/year (1940–60) → 1.1%/year (1960–90). The acceleration from Era 1 to Era 2 is entirely driven by antibiotics; the high pace was then sustained into Era 3 by cardiovascular treatment.
• The "medicalization of death": post-1960, medical care displaced public health and nutrition as the dominant mechanism of mortality decline. This is qualitatively distinct from the earlier eras in that it required ongoing healthcare delivery and spending rather than one-time infrastructure improvements.
• Behavioral factors (smoking cessation) contributed to CVD decline but the authors argue medical care — statins, antihypertensives, coronary care units, coronary artery bypass grafting (CABG) — is the primary driver of the post-1960 improvement.
• The age distribution shift (from young-age to old-age dominance) is a direct historical explanation for the $b_x$ tilt in Lee-Carter mortality models: as elderly mortality became the locus of improvement, the age-specific sensitivity weights $b_x$ tilted toward older ages.

Concepts Introduced or Extended

• Period Mortality — three-era age decomposition of US 20th-century LE gains; historical context for the $b_x$ tilt
• Lee-Carter Model — the $b_x$ tilt pattern has a direct historical cause in the era-by-era shift of dominant age groups
• Compression of Morbidity — period of improvement at old ages supports compression narrative but paper focuses on mortality not morbidity
• Medicalization of Death — coined or formalized here as the post-1960 shift to medical care as primary mortality mechanism

Entities Mentioned

• David Cutler
• Ellen Meara

Quotes

"The shift of mortality from the young to the old, combined with the increasing importance of medical care as a mechanism for reducing mortality, has transformed the healthcare system."

"Much of the improvement [post-1960] is thus a result of medical care, not behavioral change."

My Take

The three-era narrative is compelling and the DiD for antibiotics (Table 4, −3.8%/yr net effect) is well-identified. The attribution of post-1960 CVD decline to medical care rather than behavior is less definitive — the authors acknowledge smoking cessation contributed and rely partly on the timing argument (CVD decline began before widespread exercise/diet changes). The paper's core value for this wiki is the quantitative table of LE contributions by age group and era (Table 1), which is the cleanest decomposition of 20th-century US mortality history available, and the historical rationale it provides for the $b_x$ tilt in Lee-Carter models. The "medicalization of death" framing is important context for understanding why medical spending on the elderly grew so dramatically post-1960.