Psychological barriers in oil futures markets

a r t i c l e i n f o Available online xxxx JEL classification: G15 O13 Q43 Keywords: Psychological barriers Clustering WTI Brent Multiple hypothesis testing WTI and Brent futures are tested for the presence of psychological barriers around $10 price levels, applying a multiple hypothesis testing approach for statistical robustness. Psychological barriers are found to be present in Brent prices but not in WTI prices, which is argued to be due to the more prominent role that Brent plays as a global benchmark and, based on recent behavioural finance research, the greater complexity inherent in Brent fundamental value determination. Brent particularly displays evidence that when breaching a $10 barrier level from below with rising prices, the trend is for prices to fall on average subsequently. Similar behavioural-based patterns are evidenced at the $1 barrier level for the WTI–Brent spread. We show that psychological barriers only appear to influence prices in the pre-credit crisis period of 1990–2006, with such effects dissipating during the crisis and as markets reverted back to wider economy focused fundamentals. A range of reaction windows are applied with the main finding being that the trading potential around such psychological barrier levels is primarily in the immediate 1–5 days following a breach. The research contributes to the scant existing research on psychological influences on energy market traders, and suggests strong potential for further application of behavioural finance theories to improving understanding of energy markets price dynamics. Recent research by Narayan et al. (2011) investigates price clustering in oil futures and finds significant evidence of clustering in these contracts particularly around whole dollar amounts (i.e. prices with digits ending in .00). Further recent research by Bharati et al. (2012) finds clustering in whole dollar amounts ending in the 9 digit in NYMEX oil contracts. These findings contradict the notion of pricing efficiency, which would suggest that prices evolve in a manner where the likelihood of any given price change is approximately equal. This would in turn result in the distribution of " trailing " digits (the last digit of a price) following either a uniform distribution of digits or a distribution following Benford's Law (see Bharati et al., 2012) and thus there would be no systemic clusters of prices around digits. In contrast to this theory, a number of alternative price dynamic theories have been suggested, particularly a price attention and attractiveness hypothesis, developed based on …