ABSTRACT
The paper pioneers research on high frequency (HF) quoting noise in electronically traded agricultural futures markets. HF quoting – quickly cancelling posted limit orders and replacing them with new ones – emerges as a strategy for liquidity-providing traders. HF quoting can generate noise in price quotes which adds uncertainty to order execution and impairs the informational value of bid and ask prices. It can also lead to the perception that markets cannot be trusted for commercial transactions. Using intraday Best Bid Offer data for 2008–2013 and wavelet-based measures of volatility, we investigate the excess variance and co-movement discrepancies in the bid and ask prices. We find excess HF quoting variance exists. It is the highest at 250-ms scale – 90% higher than the variance implied by a random walk – but declines quickly to 7% at the 32 s scale. But its economic magnitude is negligibly small. Bid and ask price co-movements show a low degree of discrepancy with average correlations at 0.67 at 250 ms and reaching 0.95 at 8 s. All measures indicate that HF quoting noise has declined through the period. Overall, HF quoting has not caused excess variance during the transition to electronic trading in the liquid corn futures market.
Disclosure statement
No potential conflict of interest was reported by the authors.
Correction Statement
This article has been republished with minor changes. These changes do not impact the academic content of the article.
Notes
1 ‘CFTC Moves to Rein In High-Speed Traders’, Wall Street Journal, August 22nd, 2013.
2 For example, Clark-Joseph (Citation2013) identifies the existence of ‘exploratory trading’ in the S&P500 futures market.
3 Consider the December 2009 contract. To avoid maturity and overlapping data problems, the data range from September 1 to November 30. Other contracts are treated similarly.
4 Since May 2012, CME has changed corn trading hours several times (See Lehecka, Wang, and Garcia Citation2014). We use the 9:45 am–1:00 pm window in the full sample for consistency.
5 As a robustness check, we compare variance ratios using two alternative upper limits of 17.07 (J = 13) and 68.27 min (J = 15) to check if estimates stabilize as expected. The comparison suggests that the excess variance ratios at the 250 ms change only modestly. Importantly, the variance ratios stabilize at the 34.13 min upper limit. Thus the 34.13 min upper limit (J = 14) is used.
6 Tick size is selected for comparison because order execution price changes are often viewed in this context, and because it is constant through the sample, making temporal comparisons easier.
7 These results are not presented but are available from the authors.
8 We also estimate the model including the first and last 15 min of data. As expected, fundamental price variance represented by the 34.13 min scale is higher because of accumulated overnight information. As a result, estimated variance ratios are lower. For the 2010 March contract for instance, the wavelet variance ratio at the 250-ms scale drops from 2.27 to 2.18. The change suggests the excluded time exhibits higher fundamental volatility.