Explaining oil price movement difficult with available data

Anyone has to struggle, even with the data from the 1980s, to make this case (Fig 1). Maybe the period from 1984 to 1990 might be construed to support that point of view, but the rest of the time the relationship seems flaccid.

Some analysts have turned to the relationship between crude oil prices and the crack spread to allege the opposite, that crude oil prices and refinery margins move together. In Fig. 2, it appears that the two are correlated, especially in the last 10 years. Crude oil prices, of course, represent the return to the upstream. The crack spread is a surrogate for a return to refining.

In Fig. 2 the crack spread is a cocktail of two parts gasoline price and one part distillate price less three parts crude oil price. It ignores, of course, the prices of oil products that represent 30-40% of the average refinery outturn.

As for the crude oil component, published or traded crack spreads are crude-type specific. Prices of the various crude types—light, heavy, sweet, sour, etc.—do not move up and down together, but crack spreads don't capture this vagary.

Even so, as Fig. 3 shows, crack spreads and refinery margins (from the Energy Information Administration, the source of all the data in these charts) move in lock step. And they follow, more or less, the crude oil price from 1987 to 2003. So is the crack spread-crude oil price correlation in Fig. 2 a coincidence?

The supply of refined products and the supply of crude oil seem independent.

So an underlying logic might be that demand-pull alone is driving the supply-demand balance for both the crack spread and crude oil prices. Maybe sometimes, but hardly anyone believes it true for the whole 18-year period.

Utilization rates

If the data were available, the utilization rates of crude oil producing capacity and refining capacity might provide a clue.

Unfortunately, world crude oil producing capacity is in the eye of the beholder, and not a very reliable denominator for capacity utilization.

Further, year-to-year changes in crude oil production, the numerator, come from a large number of relatively small producers plus just a handful of large producers in the Organization of Petroleum Exporting Countries. Price-determining production is, so far, still in the hands of the latter. Most of the time, their self-interest has more to do with crude oil prices than anything else. Not much quantitatively analyzable data exist there.

As for refinery capacity and utilization, the only data readily available for, say, the US comes in the form of distilling capacity utilization. That seems to have little correlation with either crack spreads or refining margins (Fig. 3).

Most analysts believe that's true because the bottlenecks in refining—the capacity limitations—are usually in refinery conversion units, not distilling. By conversion capacity they mean catalytic cracking, coking, and maybe hydrocracking, the units that convert the heavy end of the barrel to light products.

Unfortunately, again, no data are available to capture conversion capacity utilization. The capacities of individual process units are reasonably well- known because they are published annually in Oil & Gas Journal. But the throughputs of these units are not. Some companies simulate the throughputs by building models of refinery flows, but their results are tightly held.

A surrogate for conversion capacity utilization is the light-heavy (L-H) differential, the difference between the prices for heavy fuel oil and a mix of light products (Figs. 3 and 4). The L-H differential is the value accruing to conversion capacity.

Some analysts are tempted to explain the obvious swings in the L-H differential by the rate at which refiners build new capacity and the growth of light product demand.

However, the availability of conversion feed is importantly affected by the mix of crude oil grades around the world. For example, as the supply of Venezuelan heavy crudes was disrupted by civil unrest in the last few years, the L-H differential sagged due to shortfalls in pitch availability to feed cokers.

Someone might suggest that distilling utilization should have gone up as refiners scrambled to run more light crudes to make up the light products deficit, but that answer can't be found in the available statistics.

What to conclude?

Given this mélange of unruly data, what does one conclude?

Capacity utilization data are interesting but either not available or sometimes misleading.

The surrogates, crude oil prices, crack spreads, and L-H differentials need large doses of inference to tell what's physically happening in the upstream and downstream.

The recent correlation between crude prices and crack spreads is indeed a coincidence.

While some analysts focus mainly on security issues, most are ultimately interested in money—the returns to the upstream or downstream or the prices that can be used to predict them. The key to forecasting them would be the future underlying flows and capacities in the upstream and downstream.

But with notable (and even then, sporadic) exceptions, few have been able to use the available but inadequate flow data to anticipate the extent of the extraordinary upstream and downstream changes in prices and margins that have happened in 2004.

The author

William L. Leffler of Venus Consulting spent most of the last two decades of his 36 year career with a major oil company analyzing and forecasting oil prices and margins. He is now a writer of books in PennWell Corp.'s nontechnical series, a lecturer on their subjects, and a consultant.