Does Population Aging Lower Aggregate Saving? A Systematic Review and Meta-Analysis

Panel regressions and calibrated life-cycle models have studied the effect of population aging on aggregate saving along separate lines, on partly overlapping samples. This systematic review and meta-analysis pools both under a single protocol. A search of five bibliographic databases identified 36 eligible studies, 33 of which yield estimates that can be placed on a common scale, each counted once. Random-effects pooling gives a decline of 0.56 percentage points in the saving rate per percentage point of age dependency (95 percent confidence interval 0.33 to 0.80). Still, heterogeneity is high, and the prediction interval for a new study includes zero. A robust Bayesian model average that allows for publication bias reduces the estimate to 0.09 points, with a credible interval spanning zero, and precision-weighted corrections reverse sign when one unusually precise study is removed. Reduced-form and structural estimates are similar but too imprecise to establish agreement, and estimates based on the elderly share are about twice as large as those based on the old-age dependency ratio. The evidence supports the direction of the effect but not its magnitude.