On 14 December 2006, OPEC announced that it would cut production by 500,000 barrels a day from 1 February 2007. No oil went missing that day, yet the oil price rose by about 2%. Traders price oil on what they expect to be available in the months ahead, as well as on what comes out of the ground today.
That gap between news and supply matters well beyond the oil market. If the expectation of a shortage moves prices months before any shortage arrives, it can also move inventories, output and inflation in the meantime. We measured how far it does by rebuilding a published study of OPEC announcements with our open-source Bayesian library.
In brief
- News that oil will be scarcer later raises its price now. OPEC announcements act as a natural experiment that separates this news from the other forces moving oil prices.
- The economy responds in sequence. After news that lifts the oil price by 10%, inventories build within months and world oil production falls later. US industrial production is about 0.4% lower after a year and 0.7% lower after two, and US consumer prices are about 0.35% higher within a year.
- Some of these effects are near certain and others less so. The probability that US industrial production is lower a year after the news is above 99%; for a fall in world oil production after two years, it is about 94%.
- The effects on output and prices build over one to two years. That lag is time for policymakers to prepare a response.
Why the cause of a price rise matters
Oil prices rise for different reasons, and the economy responds differently to each. Strong global demand pushes the price up, but it arrives with rising incomes. A disruption today, such as a conflict that shuts in production, forces refiners to draw down their inventories. News that supply will tighten later, such as a group of producers announcing a cut, gives buyers a reason to build inventories while oil is still available. All three raise the price. They have different consequences and call for different responses.
The history of the oil price bears this out. The study we draw on, by the economist Diego Känzig, finds that supply news did little to drive the surge in prices between 2003 and mid-2008, which earlier work by Lutz Kilian attributes mainly to strong global demand. Supply news does explain a significant part of the collapse that began in mid-2014, after Saudi Arabia signalled that it would not cut production to offset rising output elsewhere.
The difficulty is that the cause is not written on the price. Prices, production, inventories and economic activity move together and feed back on each other, so the data alone cannot say which movement started what. Separating them takes outside information that moves with one cause and not the others.
OPEC announcements as a natural experiment
Känzig found that information in OPEC’s announcements. For each of 119 announcements between 1983 and 2017, he measured how much oil futures prices changed over the announcement day. He used futures because OPEC decisions usually take effect about a month later. The outlook for the world economy is already in the price before an announcement and is unlikely to change within a day, so the price change mainly reflects news about OPEC’s supply plans.
These surprises capture only part of the supply news that reaches the market: OPEC meets a few times a year, and news also arrives between meetings. Their job is narrower. They reveal which part of the month-to-month movement in oil prices behaves like supply news. A model of six monthly series (the oil price adjusted for inflation, world oil production and inventories, world and US industrial production, and US consumer prices) then traces how that news works through the economy over the next four years.
We rebuilt the analysis with Impulso, our open-source library for Bayesian time-series models. The economics is unchanged; what changes is how the uncertainty is reported. The original gives a best estimate with error bands from resampling. The Bayesian version gives a probability distribution for every effect at every horizon. Our central estimates trace the same paths as the published ones and peak somewhat lower, and the published peak effects all fall within our credible intervals. The model, code and checks are in the technical tutorial.
What happens after the news
Each chart below shows the response to supply news that raises the oil price by 10% in the month it arrives, measured against where each series would have been without the news. The model is linear, so news half that size has half the effect. The line is the central estimate. The darker band holds the middle 68% of plausible values and the lighter band 90%: given the data and the model, there is a 90% probability that the true effect lies inside the lighter band.
The oil market prepares for a shortage that has not arrived
Inventories rise at once, by about 0.4% within six months, and keep rising to about 1% after three years. World oil production edges up at first, then starts to fall after about six months and is about 0.4% lower after two years. That is what news of a future shortfall should do: firms store oil while it is still available. A shortfall today would look different, with production falling at once and inventories drawn down to cover the gap.
Inventories build at once; production falls later
- Central estimate
- 68% credible interval
- 90% credible interval
Real oil price
World oil production
World oil inventories
Change in each series, in per cent, after supply news that raises the real oil price by 10% in the month it arrives. Monthly data, 1974 to 2017, from Känzig (2021); estimates by Quant Climate with Impulso.
Show the numbers
| Series | First month | After 1 year | After 2 years | After 4 years |
|---|---|---|---|---|
| Real oil price | +10.0% (90% credible interval) +10.0% to +10.0% | +7.3% (90% credible interval) +5.7% to +9.3% | +4.6% (90% credible interval) +2.9% to +7.1% | +1.4% (90% credible interval) −0.6% to +4.6% |
| World oil production | +0.15% (90% credible interval) +0.06% to +0.24% | −0.17% (90% credible interval) −0.43% to +0.08% | −0.37% (90% credible interval) −0.77% to +0.01% | −0.47% (90% credible interval) −1.14% to +0.07% |
| World oil inventories | +0.20% (90% credible interval) +0.14% to +0.26% | +0.55% (90% credible interval) +0.32% to +0.78% | +0.82% (90% credible interval) +0.50% to +1.17% | +1.00% (90% credible interval) +0.57% to +1.59% |
Central estimate, with the 90% credible interval beneath it.
The model was not told to find this pattern. It imposes no direction on any response, so the timing is evidence that the OPEC surprises are picking up news about supply.
Output falls and prices rise, with a lag
The effects then spread beyond the oil market. US industrial production falls straight away and keeps falling for about three years: it is about 0.4% lower after a year and 0.7% lower after two. World industrial production edges up for the first few months, possibly because oil exporters gain at first, and then falls, by about 0.3% after two years and 0.5% after four. US consumer prices rise by about 0.35% within a year and stay higher for around three years before the effect fades.
Industrial production falls and consumer prices rise
- Central estimate
- 68% credible interval
- 90% credible interval
World industrial production
US industrial production
US consumer prices
Change in each series, in per cent, after supply news that raises the real oil price by 10% in the month it arrives. Monthly data, 1974 to 2017, from Känzig (2021); estimates by Quant Climate with Impulso.
Show the numbers
| Series | First month | After 1 year | After 2 years | After 4 years |
|---|---|---|---|---|
| World industrial production | +0.08% (90% credible interval) +0.05% to +0.11% | −0.07% (90% credible interval) −0.27% to +0.12% | −0.31% (90% credible interval) −0.61% to −0.04% | −0.45% (90% credible interval) −0.91% to −0.08% |
| US industrial production | −0.13% (90% credible interval) −0.17% to −0.10% | −0.44% (90% credible interval) −0.68% to −0.22% | −0.74% (90% credible interval) −1.09% to −0.43% | −0.81% (90% credible interval) −1.32% to −0.44% |
| US consumer prices | +0.14% (90% credible interval) +0.13% to +0.15% | +0.35% (90% credible interval) +0.26% to +0.46% | +0.32% (90% credible interval) +0.17% to +0.49% | +0.17% (90% credible interval) −0.05% to +0.44% |
Central estimate, with the 90% credible interval beneath it.
Weaker output alongside higher prices is the combination that makes oil shocks hard for policy. Measures that support activity can add to inflation, and measures that contain inflation can deepen the slowdown.
How sure we can be
The effects on US industrial production and consumer prices are well determined. The probability that industrial production is lower a year after the news is above 99%, and the same holds for higher consumer prices. The fall in world oil production is less certain: the probability that production is lower after two years is about 94%, and its 90% interval runs from a fall of 0.8% to roughly no change.
Uncertainty also grows with the horizon. After four years, the 90% interval for US industrial production runs from a fall of 0.4% to a fall of 1.3%, about twice as wide as after one year. A plan built on the longer-run effects should allow for that wider range, and the charts show how wide it is.
What policymakers can do with the results
Identify the cause before responding. A price rise driven by news of tighter supply comes with a forecast attached: inventories build now, output falls over the next year or two, and consumer prices rise and stay higher. A rise driven by strong demand implies something different. Futures prices around announcements give an early reading on how much of a move is supply news, long before production or output data arrive.
Use the lags. The effects on output and prices build over one to two years, which leaves time to prepare. Forecasters can build the expected paths for inflation and output into their projections, and central banks can weigh the coming rise in prices against the weaker activity that follows. Känzig finds that the Federal Reserve tightened only after about a year, once prices beyond energy began to rise, and then only slightly.
Turn the responses into scenarios. News that lifts the oil price by 10% implies a path for output and inflation over four years, with a range at every step, which makes a ready-made stress scenario. Because the model is linear, a 20% rise doubles every path, though very large shocks may not scale so simply.
Put a probability on it. “The probability that US industrial production is lower a year after such news is above 99%; for a fall in world oil production after two years, it is about 94%.” A committee, board or regulator can weigh statements like these. A single number with no range gives them nothing to weigh it against.
Watch the announcement as well as the barrels. Inventories build on news alone. For energy-security planning, an announcement about future supply can move the market before any physical supply changes.
Apply the method to other announcements. Nothing in the approach is specific to OPEC: any announcement that markets price can be studied the same way. Känzig has since applied it to the EU’s carbon market, using price changes around 114 regulatory decisions to estimate the effects of carbon pricing. Emissions and economic activity fall, energy and consumer prices rise, and poorer households bear more of the cost. The energy transition will bring many more announcements about future fossil-fuel supply, such as production caps, phase-out dates, sanctions and carbon-market reforms, and the same method can measure what each one does to the economy.
What the results do not tell you
- The analysis isolates one kind of shock: supply news that coincides with OPEC announcements. It does not explain every movement in oil prices.
- The data run from 1974 to 2017, and the model assumes the economy responded in the same way throughout. The oil market has changed since: the US has become the world’s largest oil producer, and economies use less oil for each unit of output than they did in the 1970s and 1980s. Today’s effects could be smaller or larger, so refit the model to recent data before its numbers inform a live decision.
- The economy-wide effects are measured for the US and for world industrial production. Other economies, oil exporters in particular, need their own analysis, though the method carries over directly.
- The method assumes that nothing other than supply news systematically moves oil futures on announcement days. Känzig tests this in several ways, including with an alternative method that relies only on prices being more volatile on announcement days, and finds similar, if less precise, results.
Want to know how news about energy supply, or a climate policy announcement, could move the numbers you are responsible for? Get in touch.