Gasoline Price Trends According to Futures

From the WSJ yesterday:

After a sizzling start to the year, gasoline futures prices are sliding, easing pressures on drivers and the U.S. economy and raising the prospect that prices at the pump could be headed lower still.

Gasoline futures, a key yardstick for wholesale prices, are down 6.3% from their high for the year reached on March 26, as the price of crude oil that gets refined into gasoline has dropped a similar amount amid easing tension over Iran, the world’s fourth-largest oil producer.

Joanne Shore, a senior analyst of the EIA, says retail prices typically lag behind those of gasoline futures by a few weeks. About half of the change in futures gets passed through within a week, and the rest tends to show up over the next several weeks, according to Ms. Shore. who follows gasoline prices.

In addition to lower tension over Iran’s disputed nuclear program, two major east-coast refineries that had been earmarked for closure now appear to be getting a new lease on life. The owners of the refineries, ConocoPhillips Co. and Sunoco Inc., are in talks to sell the facilities rather than close them.

Meanwhile, refineries in Europe and the U.S. that had shut down for maintenance are due to come back on line soon, increasing available supplies, Lawrence Eagles, an analyst at J.P. Morgan Chase, JPM said in a recent research note.

Figure 1 depicts the relationship between RBOB futures for the front month, and the price of gasoline (in logs) over the 2005M10-2012M04 period (that’s the period I could get the futures data).

Figure 1: Log gasoline futures (end of month for front month contract, RBOB NYMEX) (blue, left axis) and log gasoline price per gallon (all, average of month) (red, right axis). Observation for April is for 4/16 (red square). NBER defined recession dates shaded gray. Source:, St. Louis Fed FRED, NBER, author’s calculations.

The relationship appears to be pretty strong; a simple bivariate regression using the lagged futures price yields the following estimates:

(1) pgast = 0.514 + 0.719 × fgast-1 + ut

Adj-R2 = 0.97; SER = 0.034; DW = 1.90; smpl = 2005M11-2012M03; n = 77

Using this relationship, and the futures prices as of April 19, one can infer the path of future gasoline prices. I show the result in Figure 2:

Figure 2: Actual price of gasoline, all formulations, $/gallon (dark blue), and gasoline prices implied by futures as of 4/19 and equation (1). April observation is for 4/16 (dark blue square). NBER defined recession dates shaded gray. Source:, St. Louis Fed FRED, NBER, author’s calculations (see text).

One issue is whether futures are an unbiased predictor (or even a particularly good predictor) of future commodity prices. In fact, this was the subject of some research conducted at the CEA back in 2001. The academic version of this memo, updated and expanded, was reported in Chinn and Coibion (Tables 1 and 2) [post; An energy-focused earlier version, Chinn, Coibion, Leblanc (2005); described in this post]. The energy excerpt from Table 1 is presented below.

Table 1 from Chinn and Coibion (2010). Note: The table presents estimated results by OLS of equation (2) in the text for different commodities and futures prices horizons. For gasoline, only the old version (“HU”) of futures are used in the regressions. Statistical significance at the 10%, 5%, and 1% level are denoted by *, **, and *** respectively. For β, the null is that β=1. Newey-West HAC standard errors, using truncation equal to the futures horizon minus one.

We find that gas futures at the 3, 6 and 12 month horizon are essentially unbiased predictors of future spot prices (for the earlier series, NYMEX futures ticker “HU”). While the proportion of variation explained is not particularly high (17%), futures outperform a no-drift random walk (although not with statistical significance), along the RMSE and MAE dimensions. In contrast, an estimated (rolling) ARIMA(1,1,1) does worse than a no-drift random walk (and hence obviously, futures).

Hence, market expectations (as proxied by futures, which appear to be unbiased predictors for gasoline prices) predict a downward drift in gasoline prices, to about $3.50 by year’s end.

This post originally appeared at Econbrowser and is posted with permission.