Abstract
International investment agreements and investment-promotion policies assume that foreign direct investment (FDI) adds to host-country capital. We examine this premise by estimating the long-run pass-through of FDI into gross fixed capital formation for 155 economies over 1997–2023, using dynamic fixed-effects panels, system GMM, 2SLS and common-correlated-effects estimators, together with a new measure of investment-treaty exposure constructed from 4125 treaties. The average long-run pass-through is about 0.23, well below one-for-one, which is consistent with substantial crowding-out. Pass-through declines steeply with development: it is close to one at the 10th percentile of GDP per capita, about 0.5 at the median and statistically zero in upper-middle- and high-income economies. The gradient is robust to alternative lag structures, cross-sectional dependence, the exclusion of pass-through FDI economies, sectoral composition and bias-corrected estimation, but it is imprecise when FDI is instrumented only by its own deep lags, so the estimates are best read as robust conditional associations rather than definitive causal effects. Denser investment-treaty networks appear to weaken pass-through, but the association disappears once development stage is controlled for; capital-account openness, regulatory quality and product-market regulation likewise play no robust independent role. The capital-formation dividend of FDI appears to be concentrated in low-income economies.