Javier Meseguer

Overview

Javier Meseguer is an economist at the Social Security Administration (SSA) who applies Bayesian econometric methods to long-horizon demographic forecasting. His 2010 paper adapts Villani's (2008) mean-adjusted BVAR to jointly forecast mortality and fertility rates across 38 countries, using the cohort-component projection identity to produce full distributional forecasts of age-structured population, and demonstrates that the dominant demographic standard (Lee-Carter) dramatically understates long-run forecast uncertainty.

Key Contributions

• Bayesian VAR for demographic forecasting (Meseguer 2010): Applied Villani's mean-adjusted BVAR (Normal-diffuse prior, three-block Gibbs sampler) to 22-age-group mortality BVARs and 7-age-group fertility BVARs. Introduced a correlation-adaptive Minnesota $\lambda_2$: $\lambda_2(i,j) = 0.8 \times \mathrm{Corr}[\Delta\log y_i, \Delta\log y_j]$, allowing adjacent age groups to share information without imposing uniform shrinkage.
• Stationarity constraint: Enforced stationarity by rejecting Gibbs draws where the companion matrix spectral radius exceeds 1 — a simple indicator-function constraint effective in practice.
• Lee-Carter vs. BVAR comparison: Showed Lee-Carter severely understates long-run uncertainty; actual 65+ female population is already outside the Lee-Carter 90% credible interval by 2009. The informative BVAR matches the UN Medium Variant closely; the diffuse BVAR overshoots (TFR 2.13 vs. target 2.0; total population 601.5M vs. Alt 2's 514.8M).

Related

• Steady State VAR
• Lee-Carter Model
• Cohort Component Method
• Sims-Zha Prior
• Mattias Villani