A harmonised twelve-country dataset of net capital and FDI stocks constructed under a bell-shaped retirement profile, 1960-2022

Net fixed-capital and inward-FDI stocks for twelve economies, built from investment flows by the perpetual inventory method under one uniform retirement profile. 755 rows, 1960-2022 (US from 1961; German stocks start 1970). Capital is complete over 1990-2022: n=396. The profile is 21 weights over ages 0-20; annual write-off peaks at 6.6% at age 12, close to a Weibull with k=2.16. The main series renormalises the terminal share to zero; a robustness series keeps the original 7.75% recursively. Both are supplied: levels differ 3.5-16.5% in the window, growth rates correlate at 0.997. Capital is constant 2015 US$ per capita, seeded at 2x GDP in each country's first year with GDP data. Because the main profile is zero from age 20, the seed enters with coefficient exactly zero from t0+20: every main-variant stock in the window is an exact function of the flows, given the schedule. That is algebra. FDI 1990-2022 is UNCTAD for eight economies, verified against World Investment Report 2024 at 0.41% median deviation (World Bank: 43.7%); China, India, Korea, Thailand are World Bank. 45 pre-1990 rows could not be deflated, and stocks cumulating them stay mixed-basis into the window - Brazil 1990-2007, 81% of its 1990 stock. Flags mark every affected row. PWT 11.0 is the benchmark (constant 2021 national prices, not the 2015 US$ basis of the constructed series, so growth rates only): correlations 0.952 and 0.960. That says little - no retirement at all reaches 0.967. The placebo test is reported with its limits: 45% of random weights are not rejected.