Oil prices have fallen substantially since last summer. Crude West Texas Intermediate (WTI), for example, fell by about 60% between June and January, before starting to rebound somewhat in February. In today’s comment we use statistical techniques to explore the drivers of these changes in the oil price.

The idea behind our approach is to use the behavior of oil and equity prices to disentangle demand from supply shifts. Intuitively, we would expect that positive demand shocks should push both equity and oil prices up, while positive supply shocks should push equities up and oil prices down. We therefore call anything that pushes oil and equities in the same direction a “demand” shock and anything that pushes them in opposite directions a “supply” shock. (This approach is the same as the one used by our colleagues in the Asia economics team, see here.)

The intuition can be seen qualitatively in the chart below which shows the S&P 500 and WTI since January 1, 2005, using daily data. Equities and oil generally both rose between 2005 and 2007, suggesting that strong demand was the primary driver of both. From late 2007, however, equity prices started to soften while crude oil prices continued to rise sharply, pointing to supply "bottlenecks" during this time. Demand was clearly the key driver during the financial crisis, as both equities and oil both plummeted in late summer 2008 and then bounced back in 2009-2010. Oil prices then remained flat between 2011 and the summer of 2014 while equity prices rose significantly. This behavior is consistent with improvements in energy supply--primarily through shale gas--during this time. Oil prices then plummeted in the second half of 2014 but equity prices generally continued to climb, suggesting that supply factors were the key driver.

We then use statistical techniques (a so-called vector autoregression with sign restrictions) to quantify the underlying shocks. More specifically, the methodology provides an estimated decomposition of observed changes in equity prices and oil into underlying shocks by labeling positive co-movements as demand shocks and divergent moves as supply shocks. We focus on daily percent changes in the S&P 500 and WTI oil prices since January 1, 2005, and de-mean both to abstract from their trends during this period. It is important to stress that this approach uses only the time variation of oil and equities—and no other information such as actual economic data—to identify the shocks.

Exhibit 2 shows the estimated demand and supply shocks, cumulated since 2005. The demand pattern looks quite intuitive with strong demand 2005-2008, a collapse in late 2008, strengthening demand from early 2013 and some deceleration more lately. Broadly consistent with our discussion above, the oil supply pattern shows a sharp adverse oil supply shock starting in late 2007, then basically no change between late 2009 and 2011, gradual positive supply shocks 2012-13 and then a sharp increase in supply in the second half of 2014.

We next use our estimated model to decompose the history of oil and equity prices into contributions from demand and supply shocks. Exhibit 3 shows this decomposition since the summer of last year. Our model suggests that the vast majority of the decline in oil until November was driven by perceptions of improved supply. The continued sell-off in December and January, however, was driven by perceptions of both improving supply and slowing demand. The latest rebound in oil--which started in late January--appears to be driven by both demand and supply.

The main conclusion from our analysis is intuitive and consistent with our commodities team's views, who have argued that the decline in oil has been driven by an oversupplied global oil market. As a result, our commodities team expects that the new equilibrium price of oil will likely be much lower than over the past decade. Also, our framework appears reasonably robust to changes in the input variables. For example, we obtained similar results for Brent crude oil prices and when excluding energy stocks from the S&P 500.

*  *  *

Nonetheless, we need to keep in mind that our analysis is subject to a number of caveats. First, the estimated shocks not only reflect genuine demand and supply news in the oil market—but simply anything that pushes oil prices and equities in the desired direction. For example, the positive supply shock of recent months may represent not only genuine increases in supply in the oil market but also lower global inflation more generally. Second, our analysis can only help us understand market perceptions, as we only use market prices without any additional economic information. The model, therefore, cannot tell us what the “true” underlying drivers are. Finally, our model does not model explicitly other economic developments that affect equity and oil prices, including shifts in inflation and monetary policy.