I've been working on a promising line of research that identifies the proportion of volume (intraday and daily) that is attributable to directional trading vs. arbitrage/program trading. It turns out that historical expectations are different when volume is elevated due to an increase in directional trading vs. an increase in arb.
Along the way doing that research, I began to look at the proportion of ETF volume to total stock exchange volume. Specifically, following a line of research that I originally heard about from Jason Goepfert, I looked at the proportion of daily SPY volume to total NYSE volume.
Yesterday, for example, we declined over 1% in SPY on an above-average proportion of SPY:NYSE volume. Since 2003 (N = 844 trading days), we've had 81 days in which SPY has declined by 1% or more. When the proportion of SPY:NYSE volume is above average (N = 40), the next two days in SPY average a gain of .44% (22 up, 18 down). When the proportion of SPY:NYSE volume is below average (N = 41), the next two days in SPY average a gain of only .13% (19 up, 22 down).
Similarly, when SPY has risen by 1% or more (N = 81), the proportion of SPY:NYSE volume has made a difference. When SPY has been strong and SPY volume is high relative to NYSE (N = 40), the next four days in SPY average a gain of .55% (26 up, 14 down). When SPY has been strong and SPY volume is low relative to NYSE (N = 41), the next four days in SPY average a gain of only .03% (22 up, 19 down).
This will be a promising line of research for other ETFs and their underlying indices.