The Natural Gas Announcement Day Puzzle

Abstract

This paper studies natural gas futures returns on EIA storage announcement days.More than 50% of the annual return is earned on these days. We find a significant difference between announcement and non-announcement day returns, which cannot be explained by the announcement surprise or other control variables. At the intraday level, the return splits half into a pre- and post-announcement part. The pre-announcement return is entirely generated on days when storage levels exceed analysts’ expectations casting doubt on explanations based on informed trading.After transaction and funding cost, a simple trading strategy yields substantial returns.Gas Markets, Announcement Effect, Storage News, Intraday

Temperature, storage, and natural gas futures prices

“…Apparently, the storage announcement leads traders to reassess the implications of other information thereby producing much larger price movements than on other days. Some prior literature (e.g., Chiou-Wei et al, 2014;Prokopczuk et al, 2021) also suggested that RNG W should differ on announcement days, implying that I 0 should be significant in the mean equation, but we found only I 3 significant. We do not have a good explanation for this result.…”

“…Some prior literature (e.g., Chiou‐Wei et al, 2014; Prokopczuk et al, 2021) also suggested that RNG W should differ on announcement days, implying that I 0 should be significant in the mean equation, but we found only I 3 significant. We do not have a good explanation for this result.…”

“…A voluminous literature has examined determinants of NG spot (e.g., Brown & Yucel, 2008; Hartley & Medlock, 2014; Hartley et al, 2008; Rubaszek & Uddin, 2020) and futures (e.g., Linn & Zhu, 2004; Mu, 2007; Prokopczuk et al, 2021) prices. Some authors (Brigida, 2014; Brown & Yucel, 2008; Caporin & Fontini, 2017; Ji et al, 2018) document a prominent role for crude oil prices in determining NG prices.…”

“…Gay et al (2009) and Halova et al (2014) find that NG prices react more to inventory shocks during the injection season from April to October than during the withdrawal season from November to March. Prokopczuk et al (2021) document a significant difference in mean NG returns on EIA announcement versus nonannouncement days while controlling for the storage surprises, supply and demand factors, and factors specific to NG futures. They show that opening and closing a short position before and after the announcement yields a mean annual return of 12% after transaction and funding costs.…”

“…5 Forecasted temperatures can affect inventory demand, while unanticipated temperatures could affect final demand. As mentioned in Section 1, most studies use daily HDD or CDD measures provided by NOAA (Mu, 2007), Bloomberg (Wang et al, 2019), or the AGA (Prokopczuk et al, 2021). Exceptions include Hu et al (2014), which uses weather derivatives data.…”