Equity Valuation and Portfolio Management

EQUITY CORRELATIONS DECAY AT HIGH FREQUENCIES

Both the variable intertrade duration and the bid-ask bounce have been held responsible for yet another phenomenon in high-frequency data: the decaying correlation of equity returns. This property of high-frequency data is known as the Epps effect.¹³

The Epps effect can be vividly illustrated on any liquid pair of financial instruments. Exhibits 20.12 and 20.13 show the Epps effect on the returns of a pair of ETFs: SPY and EFA. The two securities were chosen mostly due to their high liquidity, which in turns enables granular high-frequency analysis. SPY, offered by the State Street Global Advisors, is an ETF designed to mimic S&P 500 and boasts a market capitalization of over US$75 billion at the time this Chapter is written. EFA is a Morgan Stanley-run ETF that tracks MSCI EAFE Index of large foreign equities. EFA presently has a market capitalization of over US$30 billion. Both securities trade on the New York Stock Exchange (NYSE) alongside common equities.

Exhibits 20.12 and 20.13 compare correlations of the returns of two securities sampled on November 9, 2009, at 45-second, 15-second, 1-second, and 200-millisecond intervals. The returns were computed from two types of quotes: (1) “last tick” data at the end of each time interval; and (2) interpolated and sampled at the end of each time interval. Furthermore, the quotes were computed from trade data, simple bid-ask average and size-weighted bid-ask average, as discussed earlier ...