Time-variation in the effects of push and pull factors on portfolio flows : evidence from a Bayesian dynamic factor model

The extent to which push and pull factors affect international capital flows is widely debated. We contribute to this strand of literature by estimating the relative importance of push and pull factors for portfolio flows over a time span, encompassing the global financial crisis, the European sovereign debt crisis as well as the beginning of the Covid-19 pandemic. To do so, we extract common and country-specific components from fund flow data using Bayesian dynamic factor models with time-varying coefficients and stochastic volatility. Assuming that the common component represents push factors and the country-specific component pull factors, we show that (i) time-variation matters and (ii) there is a substantial amount of heterogeneity in the importance of factors across regions (advanced versus emerging market economies) and asset classes (equity versus bonds). We find that the relative importance of push factors for flows into advanced economies has on average increased over time, particularly for EU countries. With respect to flows into emerging market economies, we find very heterogeneous results between individual countries. Moreover, we identify risk measures, US stock market returns, US real interest rates, the US real effective exchange rate and the oil price as important push factors. Pull factors seem to covary with domestic stock market returns, in particular.