An evaluation of EIA gasoline price predictions

ABSTRACT

This study evaluates the accuracy of the US Energy Information Administration (EIA) monthly forecasts of gasoline prices for 2005–2021. Our findings indicate that the 1-, 3-, 6-, 9-, and 12-month-ahead EIA forecasts do not, on average, significantly over- or under-predict gasoline prices, and they generally outperform the random walk and the nowcast-based benchmarks in terms of both the mean squared forecast error and the predictive information content. We further show that the shorter-horizon (longer-horizon) EIA forecasts predict directional change under asymmetric (symmetric) loss. Additional results indicate that the 1-, 3-, 6-, 9-, and 12-month-ahead EIA forecast errors are all orthogonal to consumer expected change in gasoline prices, actual change in core consumer price index, consumer vehicle-buying attitudes, and consumer sentiments. However, the 1-month-ahead (1- through 12-month-ahead) EIA forecast errors fail to be orthogonal to crude oil price changes (consumer expected inflation), meaning that these indicators can potentially help improve the accuracy of the EIA forecasts of gasoline prices.

KEYWORDS:

• Energy prices
• random walk
• directional accuracy
• asymmetric loss
• consumer survey data

JEL CLASSIFICATION:

• Q43
• C53

Acknowledgement

The authors would like to thank three reviewers for their helpful comments and suggestions.

Disclosure statement

This is to acknowledge that NO financial interest or benefit has arisen from the direct applications of this research.

Notes

¹ There are studies that make use of exchange rate movements to predict energy prices. Pincheira-Brown et al. (2022), for instance, show that the Chilean exchange rate contains useful information for predicting crude oil, gasoline, propane, and heating oil prices.

² Crude oil prices are also useful in predicting natural gas prices. For instance, Luo, Guo, and Li (2022) examine whether one-way return connectedness indices and net pairwise directional connectedness indices (NPDC) from the grey energy market to the natural gas market can help predict natural gas returns. They show that ‘one-way return connectedness index from WTI crude oil to natural gas performs better in short-term return forecasting, while the NPDC indices from WTI crude oil to natural gas perform better in long-term return forecasting (p. 1)’. Wang et al. (2022) focus on uncertainty indices and economic conditions. They shows that such indicators contain useful information for predicting the volatility of clean energy stock and natural gas prices.

³ See Baghestani (2005), who shows that survey forecasts of inflation and growth help improve the accuracy of survey forecasts of US 3-month Treasury bill rates.

⁴ Replacing (A_t+f – PF_t+f) in the test Equation 4(4) $\left(A_{t+f}-P{F}_{t+f}\right)\hspace{0.17em}={\beta }_0+{\beta }_1{X}_{t-1}+u_{t+f}$(4) by log(A_t+f) – log(PF_t+f) yields similar p-values to those in rows 2 and 3.

⁵ See Baghestani and Palmer (2017) for analysing the dynamic relationship between ICC and ICE.