Why does capital flow from poor to rich countries?

Lucas’ classic paper (1990) highlighted the paucity of capital flows from rich to poor countries, in contrast with predictions of standard theory. He suggested four explanations – but they cannot explain the copious capital flows from certain emerging relatively low-income countries, notably China, to the USA. Empirical studies confirm Lucas’ observation. Additionally, Prasad et al (2007) showed that developing countries with less reliance on foreign capital grow faster. Gourinchas & Jeanne (2009) added a second puzzle, the “allocation puzzle”: flows are not only too low, they are directed toward countries with lower productivity growth and lower investment, i.e. the “wrong” ones; they attribute this to a distortion in savings, e.g. financial repression. This paper traces the causal processes in post-reform China. Starting around 1978, reforms allowed rural households to keep their own surpluses, facilitated “township-village enterprises”, established enterprise areas open to FDI (e.g. Shenzhen), and began making state-owned enterprises (SOEs) more efficient. High productivity at comparatively very low cost was gradually achieved, and openness to trade allowed Chinese goods to conquer the world. Statistical analysis shows that the generated profits led to high levels of capital accumulation, both corporate and public sector. Together with high household savings, channelled by state banks into (primarily) SOEs, this provided ample capital for reinvestment, as well as a large surplus. No capital inflows were required, except for FDI which had a vital technology-importation role. In particular, the massive export success generated hard currency, allowing large-scale purchase of overseas assets, e.g. US Treasury Bonds. China is not unique: previous East Asian economies had parallel experiences on a smaller scale (cf. also Buera & Shin (2011)). Thus, a relatively simple explanation of both puzzles is possible. More sophisticated interpretations, e.g. Kalemli-Ozcan et al (2010), Sandri (2010), Reinhardt et al (2013), are discussed from a methodological viewpoint.