Analyzing ETF Flows through the lens of factor models

Abstract

I investigate the relationship between ETF flows and a number of explanatory variables, focusing on smart-beta funds. Among these, I make a sample of funds following Momentum, Quality, Volatility, Growth, and Value factors, aiming to explain how different factors affect flows into these very different strategies. Flow is defined as the change in number of shares outstanding and variables such as returns, fees, turnover, and more are used as explanatory variables. I draw on the research of Kalaycioglu (2004), Ivanov (2014), and Clifford (2014) to find relevant variables to regress against and build on their models by assessing the significance of the variables they used in the context of the 5 groups of ETFs outlined above. I document a steady positive flow into ETFs and find that many of the funds studied are more passive than they would claim. Through a number of panel regressions on fund data, I find evidence of return chasing, with past returns having the greatest impact on momentum funds and the least impact on growth funds. I find that this is likely a result of investors’ search for skilled managers as funds which follow their specified factor closely get higher flows in the long term. Lagged returns appear more significant in explaining flows than current returns. In performing a factor analysis, I use 3, 4, and 5-factor models and find that the lagged factor loadings have a higher impact on flows than current-month loadings. In addition, all the funds considered do seem to be following the factors they claim, with significant results across the board when using returns as the dependent variable in the regressions. A negative relationship between fund flows and fees is documented on longer time scales such as 3 and 5 years but is harder to see over shorter periods. All of the funds in my sample turn out to be more passive than expected.