Oil inventories lose their influence on prices
Changes in reported U.S. oil inventories play only a limited role in driving futures prices on the NYMEX. Their influence is far less than the cross-over effects from other asset classes such as equities.
The chart below shows correlation between the change in U.S. inventories as reported by the Department of Energy (DOE) each week and movement in front-month oil prices on the publication date.
Price correlations with total inventories (excluding the Strategic Petroleum Reserve) and around the NYMEX delivery point at Cushing, Oklahoma are shown separately.
On average both correlations are negative, which is what theory would predict: a drawdown in inventories should be associated with a rise in nearby futures prices. But the correlations are very weak and too unstable to have much predictive value.
Since 2005, the correlation between total U.S. inventories and NYMEX spot prices has ranged from -0.48 (strongly negative) to +0.13 (perversely if mildly positive). The correlation between Cushing inventories and NYMEX spot has ranged from -0.45 and +0.18.
The two correlations are not even stable in relation to one another. Correlations with total inventories have been strong when correlations with Cushing stocks are weak, and vice versa. If NYMEX prices are driven by inventory levels, sometimes the market is more influenced by the general level of stocks, sometimes more by stocks close to the delivery point.
There are some caveats with this analysis:
(1) The data are not as “clean” as they might be. The chart shows correlations between the stock change on Friday (which is always the reporting date) and the next Wednesday (almost always the date of publication). It ignores a small number of days each year on which the DOE postpones publication until Thursday.
But this does not usually occur on more than six weeks each year. It is too few to affect the overall results. Cleaning up the data further might strengthen the correlation, but not by much.
(2) The data could be corrected for intra-day price movements. At present, the charts compare stock changes with the close-to-close movement in NYMEX front month prices. The relevant price movement is from the publication time (normally 1030 EST) to the close.
Studying intra-day price movements might strengthen the correlations, perhaps significantly, though there is no way of knowing without re-running the analysis with price movements at the tick-level.
(3) U.S. inventories may or may not be relevant for NYMEX pricing. The NYMEX contract is something of a hydra. Sometimes analysts and traders focus on the contract’s role as a benchmark for the global oil market. In that case, global inventories rather than U.S. or Cushing stocks is the relevant measure of market tightness.
On others, comment and trading seems to focus on the contract’s role in pricing a very special type of crude (light sweet) at a specific location (Cushing), acknowledging it can become “delinked” from the rest of the global oil market. In that case, U.S. or even Cushing inventories would clearly be more relevant than the global totals.
(4) In any event the inventory data have other shortcomings. DOE does not revise past totals, except in exceptional circumstances, even when reporting errors are known or suspected to have occurred. Instead fresh (and hopefully more accurate) numbers are simply published as a new inventory level the following week.
This has led to a number of instances where large stock draws (builds) one week are reversed with a large build (draw) in the opposite direction the following week as errors drop out of the stats.
The original error should not affect the correlation analysis much (markets move on perceptions of reality, even wrong ones, rather than the underlying reality itself). But the “correction” a week later might affect the correlation if it is widely anticipated and so loses “news” value compared to a similar-sized move that has come out of the blue.
Despite these analytical limitations, the lack of a strong and stable correlation between reported inventories and front-end prices (which should be most affected by near term supply-demand imbalances and the inventory cushion) is nonetheless striking.
Physical oil inventories have had less predictive power over oil prices than movements in U.S. equities since the start of 2008.
Correlations appear to have weakened in recent years, at least for total inventories, the only data for which the DOE publishes a long series. Correlations prior to 2007 were generally in the -0.6 to -0.4 range. Since then correlations have rarely been stronger than -0.2 and occasionally flipped positive.
Coupled with the increasingly strong correlations with equities and other commodities, especially since the start of 2008, it is one more piece of evidence crude oil prices have become “financialised” and are trading more like financial assets than raw materials.
The time series reveal two major structural breaks. The first occurred in the final three months of 2006 and the first eight months of 2007, when both inventories and prices were largely flat, displaying no trend. Through this period weekly inventory changes had no discernable impact on prices whatever.
The second occurred in the aftermath of the financial crisis, when the abrupt slowdown in demand and brimming tank farms at Cushing temporarily forged a strong negative correlation between prices and inventory changes at the delivery point. But there was no real correlation with wider inventory movements. Even the Cushing correlation vanished by the end of 2009.
In the past five years, inventories have been only intermittently relevant for oil prices. Even then other factors such as equities or the dollar exchange rate have had a more important influence on prices.
The data is consistent with studies suggesting current supply-demand-inventory balances have become less important, and the market is becoming more “forward-looking” and speculative as financial players have increased their share of trading dramatically.
Financial players seem more willing to “look through” temporary supply-demand dislocations to focus on the medium-term horizon (2-5 years out), using oil as an asset for positioning on the outlook for global growth, inflation and currency movements, as well as the prospects for supply meeting long-term demand projections.
Unless a U.S. inventory build/drain is seen as signalling a shift in the medium-term picture relevant to these participants, it is unlikely to have much consistent influence on the market.