The investment universe consists of 28 commodity futures contracts covering six sectors, namely, industrial metals, energy, grains, livestock, precious metals, and softs. The complete list of underlying commodities is shown in Table 1. First, each month, obtain all tweets matching commodity keywords outlined in Table 1 via the Twitter academic API. Second, take the following procedures to standardize and remove noise from the collected tweets.
ASSET CLASS: futures | REGION: Global | FREQUENCY:
Monthly | MARKET: commodities | KEYWORD: Sentiment Factor, Commodity Futures
I. STRATEGY IN A NUTSHELL
The strategy trades 28 commodity futures across six sectors using monthly sentiment extracted from Twitter. Commodities with higher sentiment changes are bought, while those with lower changes are sold, forming an equally-weighted long-short portfolio rebalanced monthly.
II. ECONOMIC RATIONALE
The strategy exploits behavioral biases explained by the appraisal-tendency framework, where emotions from prior events affect traders’ decisions, causing systematic mispricing. Correcting these biases generates a return spread between high and low sentiment-change commodities.
III. SOURCE PAPER
Wisdom of Crowds and Commodity Pricing [Click to Open PDF]
John Hua Fan, Griffith University – Department of Accounting, Finance and Economics, Griffith University, Australia; Sebastian Binnewies, Griffith University – Griffith Business School; Sanuri De Silva, Griffith University – Griffith Business School
<Abstract>
We extract commodity-level sentiment from the Twittersphere in 2009-2020. A long-short systematic strategy based on sentiment shifts more than doubles the Sharpe ratio of extant commodity factors. The sentiment premium is unrelated to fundamentals but is exposed negatively to basis risk and is more pronounced during periods of macro contraction and deteriorating funding liquidity. Sentiment-induced mispricing is asymmetric, i.e., commodities with low (high) sentiment shifts tend to be overvalued (undervalued) when the aggregate market is in backwardation (contangoed). Furthermore, the observed premium arises almost entirely from commodities with the most retweet activities, while retweets and likes themselves do not exhibit stronger predictive ability compared to non-influential tweets.
IV. BACKTEST PERFORMANCE
| Annualised Return | 7.26% |
| Volatility | 9.6% |
| Beta | N/A |
| Sharpe Ratio | 0.75 |
| Sortino Ratio | N/A |
| Maximum Drawdown | -12.2% |
| Win Rate | 71% |
