Ronald Lee

Overview

Ronald Lee is an economist and demographer at the University of California, Berkeley. He is best known as co-developer (with Lawrence Carter) of the Lee-Carter (LC) model — the most widely used statistical method for long-run mortality forecasting. His research spans demographic forecasting, population aging, intergenerational transfers, and Social Security solvency. The LC model is the foundation of most SSA, Census Bureau, and academic long-run US mortality projections.

Key Contributions

• Lee-Carter model (Lee and Carter 1992) — "Modeling and Forecasting U.S. Mortality," JASA 87(419): 659–671. Decomposes log age-specific mortality as $a_x + b_x \cdot k_t + \varepsilon$, where $k_t$ is a single latent mortality index trending linearly downward, $a_x$ is the age profile of log mortality, and $b_x$ is each age group's sensitivity to the overall trend. Enables long-run LE projections with well-defined uncertainty intervals. Standard reference model for US Social Security actuarial mortality forecasting.
• LC 8-year retrospective (Lee 2000) — "The Lee-Carter Method for Forecasting Mortality, with Various Extensions and Applications," North American Actuarial Journal 4(1): 80–93. Survey and critical self-assessment of the LC method after 8 years. Key findings: (1) out-of-sample validation 1989–1997: LC forecast +1.30yr $e_0$ gain vs. actual +1.43yr; SSA forecast implied only +0.73yr — LC ≈ 2× more accurate; (2) recommends jump-off fix using most recent observed rates as initialization; (3) catalogs six shortcomings (linear k not structural, $b_x$ instability, CI too narrow, sex differentials diverge, single-factor, parameter uncertainty excluded); (4) surveys extensions including sex disaggregation, cause disaggregation (always higher mortality), and Wilmoth one-stage weighted SVD; (5) Social Security finance application — CBO adopted LC–Tuljapurkar stochastic framework 1996–1998. Juha Alho discussant confirms LC as association model and recommends Poisson ML.
• LC forecasting evaluation (Lee and Miller 2001) — "Evaluating the Performance of the Lee-Carter Method for Forecasting Mortality," Demography 38(4): 537–549. Shows that LC forecasts systematically underestimated ex-post realized LE gains in several countries, suggesting the model's trend-rate estimate is conservative.
• b_x tilt and linear e0 trend (Lee 2003) — Resolves the puzzle of why period e0 grows linearly despite sub-linear implied gains from any single age group's rate trajectory under constant-proportional ASDR decline. The age pattern of mortality improvement (the $b_x$ weights in the LC model) shifted from young-age dominance (early 20th century, infectious disease control) to old-age dominance (post-WWII, cardiovascular treatment). Vaupel (1986) showed that improvement at older ages generates more life-years per unit ASDR decline; as b_x tilted toward older ages, this compensated for diminishing returns at young ages and maintained linear e0 growth. See Period Mortality.
• SSA forecasting gap (Lee 2003) — Documents a ~3.8-year discrepancy between extrapolation-based forecasts (White 2002, Oeppen-Vaupel 2002) and SSA's 2002 intermediate projection for US $e_0$ in 2030, directly foreshadowing Soneji and King (2012). LC itself projects ~2 years above SSA; White and OV extrapolations project ~3.8 years above. See SSA Mortality Forecasting.
• International convergence (Lee 2003) — Estimates that national e0 converges toward the best-practice frontier at $\alpha \approx 0.06$–$0.08$ per year (half-life ≈10 years), with slightly accelerating convergence for nations far below the frontier.
• Demographic transition survey (Lee 2003b) — "The Demographic Transition: Three Centuries of Fundamental Change," Journal of Economic Perspectives 17(4): 167–190. Comprehensive survey of the global demographic transition 1800–2100. Key analytical contributions: (1) the three-sub-phase age distribution sequence, especially the counter-intuitive Phase 3a where mortality decline initially makes populations younger; (2) fertility-driven vs. mortality-driven aging as economically distinct problems — the former is an inherent resource cost, the latter is an institutional constraint on retirement ages; (3) quantification of tempo distortion (period TFR depressed 10–40% below cohort completed fertility by rising mean age of childbearing); (4) India's demographic bonus (1970–2015): +22% per capita income from dependency ratio change alone. See Demographic Transition and Demographic Dividend.
• Baby boomers and the federal budget (Lee and Skinner 1999): JEP 13(1) survey. Documents SSA mortality projections as too pessimistic — projected US rates at ages 60–79 are one-third to one-half of observed international rates in peer countries. Pay-as-you-go payroll tax trajectories by 2070: SSA Middle (LE=81) → 20%; Lee-Carter (LE=87) → 24%; LE=90 → 27%; LE=100 → 32%. Lee-Tuljapurkar stochastic model: even a 2.2 pp immediate payroll tax increase leaves a 75% probability of trust fund exhaustion before 2070. Longer life ≈ +2% Medicare costs (end-of-life costs are merely postponed). Medical technology identified as the dominant Medicare risk.
• Infinite horizon SS imbalance (Lee, Miller, and Anderson 2004): NBER WP 10917. 500-year stochastic projection yields infinite horizon actuarial imbalance of −5.15% of payroll vs. 2004 Trustees Report's −3.5%; difference driven by higher mortality improvement assumptions. Demonstrates 75-year AB₇₅ is not a sustainability measure; identifies the "Unstable measure" as the best 75-year approximation for the infinite horizon central estimate. 95% PI: −1.3% to −10.5%. Also shows stochastic immigration makes negligible difference to OADR uncertainty.
• Fuller stochastic trust fund model (Lee, Anderson, and Tuljapurkar 2003) — Extended the 1998 AER P&P approach into a complete treatment of OASDI income and cost rate mechanics under stochastic inputs. Key finding: SSA intermediate projection sits below the stochastic median (official projections are optimistic relative to the model's central tendency); fertility dominates 75-year uncertainty while SSA ranks it last. Bridges the 1998 brief application and the 2004 infinite-horizon work. See Lee Anderson and Tuljapurkar 2003 — Stochastic Forecasts of the Social Security Trust Fund, Michael Anderson, Shripad Tuljapurkar.
• Stochastic Social Security finance projections (Lee and Tuljapurkar 1998) — AER Papers and Proceedings 88(2): 237–241. First fully probabilistic forecast of OASDI trust fund finances, combining LC mortality with stochastic fertility and AR(1) models for productivity growth and interest rates (750 sample paths, 1995–2070). Key results: mean trust fund exhaustion in 2026 (vs. SSA's 2029); mean LTAB = −3.3 pp (vs. SSA's −2.2 pp); median payroll tax to maintain year-ahead balance rises to 21% by 2070 (P97.5 = 34%). Uncertainty decomposition: fertility > productivity > interest > mortality at 75-year horizon — the reverse of SSA's ranking. See Long-Term Actuarial Balance and Shripad Tuljapurkar.

Related

• Lawrence Carter
• Timothy Miller
• Michael Anderson
• Jonathan Skinner
• Shripad Tuljapurkar
• Lee-Carter Model
• Long-Term Actuarial Balance
• Period Mortality
• SSA Mortality Forecasting
• Jim Oeppen
• James W. Vaupel
• Samir Soneji
• Gary King
• Office of the Chief Actuary
• Demographic Transition
• Demographic Dividend

Sources

• Lee Anderson and Tuljapurkar 2003 — Stochastic Forecasts of the Social Security Trust Fund
• Lee and Carter 1992 — Modeling and Forecasting U.S. Mortality
• Lee 2000 — The Lee-Carter Method for Forecasting Mortality with Various Extensions and Applications
• Lee 2003 — Mortality Forecasts and Linear Life Expectancy Trends
• Lee 2003b — The Demographic Transition Three Centuries of Fundamental Change
• Lee and Skinner 1999 — Will Aging Baby Boomers Bust the Federal Budget
• Lee and Tuljapurkar 1998 — Uncertain Demographic Futures and Social Security Finances
• Lee, Miller, and Anderson 2004 — Stochastic Infinite Horizon Forecasts for Social Security and Related Studies
• Lee and Tuljapurkar 1994 — Stochastic Population Forecasts for the United States Beyond High Medium and Low
• Lee and Tuljapurkar 1999 — Population Forecasting for Fiscal Planning Issues and Innovations
• Bell 1997 — Comparing and Assessing Time Series Methods for Forecasting Age-Specific Fertility and Mortality Rates
• Booth 2006 — Demographic Forecasting 1980 to 2005 in Review
• Booth Maindonald and Smith 2002 — Age-Time Interactions in Mortality Projections Applying Lee-Carter to Australia
• Booth Tickle and Smith 2005 — Evaluation of the Variants of the Lee-Carter Method of Forecasting Mortality A Multi-Country Comparison
• Carter and Lee 1986 — Joint Forecasts of US Marital Fertility, Nuptiality, Births, and Marriages
• Carter and Prskawetz 2001 — Examining Structural Shifts in Mortality Using the Lee-Carter Method
• Denton Feaver and Spencer 2005 — Time Series Analysis and Stochastic Forecasting
• Meseguer 2008 — The Out-of-Sample Performance of Stochastic Methods in Forecasting Age-Specific Mortality Rates
• Meseguer 2010 — Forecasting Vital Rates and Population with Bayesian Vector Autoregressions
• Pedroza and King 2002 — Revisiting Demographic Methods
• Tuljapurkar, Li and Boe 2000 — A Universal Pattern of Mortality Decline in the G7 Countries
• Tuljapurkar and Boe 1999 — Validation, Probability-Weighted Priors, and Information in Stochastic Forecasts
• Thompson et al. 1989 — Multivariate Time Series Projections of Parameterized Age-Specific Fertility Rates
• Wilmoth 2005 — Some Methodological Issues in Mortality Projection