Earlier this year, disruptions in Libya and the resurgence of demand from the emerging economies sent oil prices up sharply, a development that many economists believe contributed to the slow growth for 2011:H1. The chaotic markets of the last few weeks saw oil prices drop back down to where they had been in December. Will that be enough to revive the struggling U.S. economy? There is some evidence suggesting that it may be too late.
I recently completed a survey of a large number of academic studies that found a nonlinear economic response to oil price changes. One very well-established observation is that although oil price increases were often associated with economic recessions, oil price decreases did not bring about corresponding economic booms. For example, when oil prices plunged in the mid-1980s, the oil-producing states in the U.S. experienced what looked like their own regional recession. An oil price increase that just reverses a recent price decrease does not seem to have the same economic effects as a price move that establishes new highs.
One measure that has sometimes been used to summarize these effects is to compare the price of oil today with where it has been over the last year. If oil prices are at a higher level than they’ve been over the previous year, I’ll denote by pt the percentage change over the previous high. On the other hand, if oil prices have recently been higher than they are now, pt is zero, where I’m measuring the price of oil by the producer price index for crude petroleum as of the last month of quarter t. It’s interesting to take a look at how this measure is related to the quarter t real GDP growth rate, denoted yt (measured at an annual rate). The regression below, estimated over 1949:Q2 to 2011:Q1, summarizes a forecast of this quarter’s GDP growth based on GDP growth over the previous 4 quarters and what oil prices had been doing the previous 4 quarters. [Cognoscente will recognize this as equation (3.2) in a paper I published in 2003 (working paper version here), re-estimated with new data that have arrived since then and quoting real GDP growth rates at an annual rate rather than a quarterly rate as in the original; standard errors are in parentheses.]
The equation implies that if oil prices are steady or falling (so that pt-j has been zero for some time), we’d expect to see real GDP growth averaging about 4% at an annual rate. But when oil prices are making new highs, we expect slower growth. It is interesting that the biggest effects on GDP come 3 or 4 quarters after oil prices have gone up. This is not a result unique to this specification, but is something one finds whether one uses linear or a variety of nonlinear specifications and regardless of the data set. We often see some economic responses right away, such as a drop in consumer sentiment, fall in sales of less energy-efficient vehicles, or build up of inventories. But it usually takes some time for the effects of these to be multiplied as they ripple through the rest of the economy.
To illustrate the implications of the above regression, suppose that the price of oil increases by 10% in quarter t over its high for the previous year, with no subsequent increases beyond that. The figure below plots the predicted consequences for the GDP growth rate in quarter t+j as a function of j, how many quarters it’s been since the price went up. Four quarters after the increase, we might expect to see real GDP in that quarter growing almost 1.4% slower at an annual rate than it otherwise would have. [Note this is not simply a plot of the lagged coefficients on pt-j because it also incorporates dynamic feedback arising from the lagged coefficients on yt-j].
 |
| Quarter | Contribution to GDP growth (annual rate) |
|---|---|
| 2011:Q1 | -0.4% |
| 2011:Q2 | -1.1% |
| 2011:Q3 | -1.6% |
| 2011:Q4 | -2.4% |
| 2012:Q1 | -2.4% |
| 2012:Q2 | -0.7% |
| 2012:Q3 | -0.1% |
The price of oil (as measured by the end-of-quarter value for the crude oil producer price index) was 9% higher at the end of 2010 than it had been over the previous year, and the price went up an additional 15% from there during 2011:Q1. The table to the right indicates how much these changes would be predicted to affect GDP growth based on the equation above. For example, if in the absence of the earlier oil price increases we would have seen real GDP growing at a 4% annual rate, given the 2010:Q4-2011:Q1 oil price increases, we would only expect 2.4% growth for 2011:Q3 and 1.6% growth for 2011:Q4 and 2012:Q1. Note that, according to the above relation, it would make no difference for that prediction whether oil prices decline dramatically in 2011:Q3 and 2011:Q4.
However, a number of other nonlinear relations could fit the data as well or better, and some of these have dramatically different predictions from the equation above. In my 2003 study, I found the evidence favored a specification with a longer memory, looking at where oil prices had been not just over the last year but instead over the last 3 years. My reading of developments during 2011 has been that, because of the very high gasoline prices we saw in 2008, U.S. car-buying habits never went back to the earlier patterns, and we did not see the same shock to U.S. automakers as accompanied some of the other, more disruptive oil shocks. My view has been that, in the absence of those early manifestations, we might not expect to see the later multiplier effects that account for the average historical response summarized in the figure above. If one uses the 3-year price threshold that the data seem to favor (e.g., equation (3.8) in my 2003 study), the inference would be that we’ll do just fine in 2011:H2, because oil prices in 2011 never exceeded what we saw in 2008.
Even using the 1-year threshold, the above relation still says that the 2010-2011 oil price increases would not be enough by themselves to bring about a recession, but would only mean slow growth for the end of this year and beginning of next. But of course, the concern is that this is not the only shock that may be hitting the economy. I continue to worry a great deal about possible consequences of credit disruptions and fiscal contraction from the unsettled situation in Europe.
It’s worries about these developments, and what they would mean for world petroleum demand, that have been the most important factors bringing oil prices down. And that most assuredly should not be read as good news for the U.S. economy.
| San Diego Historical Gas Price Charts Provided by GasBuddy.com |
Frustrating – oil putting significant downward pressure towards 2-3 more quarters of anemic real GDP growth following a recession when growth “should” be much more robust.
JDH Some of the historical data could be contaminated. For example, in 1973 petroleum was used to generate 17% of US electricity production. So an oil embargo in 1973 was likely to have very strong effects on GDP. But today we only generate about 1% of our electicity from oil (why isn’t it zero?). On the other hand, oil price increases result in higher GDP leakages through increased imports, which subtract from GDP. On the third hand, US refining capacity is now at the highest level since the bad old days of govt subsidized levels that were phased out under Reagan. Given that a lot of historical gasoline price shocks were due to seasonal capacity issues, that ought to stabilize some of the effects of oil price shocks. In other words, the price hikes at the pump could have been a lot worse this past spring were it not for increased refining capacity. Also, cash-for-clunkers may have had an unforeseen benefit, which may have dampened the oil shock effect. A lot people bought more fuel efficient cars.
And a question. Is the asymmetric response of price change to GDP due to supply shocks in the usual sense of a cost-push model, or is it due to the way people misremember most commodity price spikes and shift consumption choices? It’s big news when the price of bread jumps up one month, but barely worth a mention and completely forgotten when the price of bread falls the next month. People only remember one kind of volatility shock. If I’m reading your post correctly, you seem to be saying that it was more the former case with earlier historical data, but probably the latter case with our most recent experiences.
A watched pot doesn’t boil.
This is our third try, since the crash, of matching GDP to sustainable oil imports. We are obviously getting better.
2slugbaits: I attribute this to demand effects. When oil prices make new highs, it may cause you to postpone buying a new car. But when oil prices decline, that doesn’t cause you to go buy two new cars. More generally, I view the business cycle as a falling below potential output, in which unemployment can climb quickly. I don’t believe that similar dynamics operate to move above potential output. The observed experience is that unemployment comes down more slowly than it goes up.
Unfortunately, iirc, consumer debt increased during this time. That means we need low prices for a much longer time to both repay and bolster our inadequate personal saving (the price spike also reinforces the point that our personal saving are very far from adequate, *).
Instead of focusing on repairing balance sheets, the broader population sees what they considered discretionary income fall and savings deteriorate. We know we cannot continue to spends like we have. That suggest that companies stock prices are based on unrealistic expectations of consumer demand. The broader population on the other hand has inadequate savings to invest even if prices weren’t inflated.
I think price drops are rarely adequate for most people reach their desired saving level, let alone put away the missed savings during the spike. Don’t forget paying the debt that supplemented their consumption during the spike.
When we see a sudden price peak it means simply that truckers and car drivers recently got stuck in gas lines and their schedules got fouled. So we notice the gas line queue first and then start paying attention to price after.
Professor, now look at the effects of the rate of change of US oil consumption to private GDP and you’ve got it.
Moreover, when US reported annualized real GDP drops below ~1.9-2%, it means that the only incremental real GDP growth occurring is that resulting from total incremental gov’t spending, including personal transfers. We’re at 1.3-1.6% annualized reported real GDP, which implies that the rate of private real GDP growth is now at 0% or negative.
Similarly, given the trend deflator, nominal GDP growth below 4.5% means that the only growth is from gov’t spending. Annualized nominal growth is now decelerating below 4% from 4.5-4.9% at the business cycle peak in ’10.
This cyclical contraction will coincide with the peak Boomer demographic drag effects taking hold, as we will experience the fastest rate of change of increase in the number of Boomers (in absolute terms and per population and labor force) reaching age 62-65 (when 50-80%+ of Americans leave the labor force and begin drawing down on Social Security, Medicare, and retirement savings, if they have any) beginning this year into ’13-’16, persisting at a slower increasing rate into the early to mid-’20s, which will coincide with a marked increase in risk aversion and liquidity preference, as well as the likelihood of all-time lows for consumer confidence.
Moreover, the historical self-similar secular bear market precedent implies that the next bear market will be the largest to date in real and currency-adjusted terms, i.e., a new secular low in adjusted terms (as in the 1830s-40s, 1890s, late 1930s to early 1940s, and Japan in ’02-’03), coinciding with the worst 5- and 10-yr. real change and returns to the S&P 500 of the secular bear market to date.
Thus, the next recession and bear market will coincide with worsening Boomer demographic drag effects and extremes in liquidity preference, risk aversion, and low consumer confidence.
Were the historical secular pattern to repeat, the 10-yr. Treasury yield will fall well below 2%, and the T-bond yield will fall below 3%, as price deflation (at least core deflation) emerges.
Finally, the secular bear market will not end in nominal and adjusted terms until all of the inflation- and currency-adjusted gains from the preceding secular bull market are wiped out (leaving investors only the avg. dividend yield before fees and taxes over the course of the secular bull and bear markets), and the S&P 500 dividend exceeds the Aaa-rated yield, perhaps by a large margin.
Note that the WTI price is pretty much irrelevant to oil consumers.
WTI is really only relevant to Mid-Continent producers, who are getting much less than the global oil price, and to Mid-Continent refiners, who are making a killing.
The Brent spot price is a far better indication of global oil prices, and at about $107, it is currently only about $5 below the average price for 2011, through July, $112, which incidentally, so far considerably exceeds the previous annual high for Brent, which was $97 in 2008.
And as previously noted, our work suggests that the US is being gradually priced out of the global net oil export market, as we are forced to take a declining share of a falling volume of global net oil exports.
My approach is volume, not price, based.
So what’s going on here? What’s happening is that oil supply growth is not sufficient to keep up with oil demand, and that demand growth is coming from the non-OECD countries. Therefore, the OECD countries must re-allocate some of their consumption to the emerging economies.
So the issue is the mechanics of how that consumption is re-allocated. If you have elastic demand, then simple price rises will lead to falling consumption. Except that’s not what the historical record says for oil. In times of economic growth, countries do not give up consumption. For example, US consumption is unchanged since June 2009, even though the price of oil (Brent) has increased from $68 to $112 in June 2011. Further, the price of oil has been above our estimates for the carrying capacity of the US economy since December, but consumption has declined only marginally. In the recent price spike, US oil demand has exhibited the demand inelasticity typical for periods of economic growth. This is entirely similar to the 2007 to mid-2008 pattern.
Thus, price results from dynamics of demand re-allocation. Oil volumes must move from the OECD to the non-OECD; price is only the value needed to induce that movement.
As a result, for the US economy, one of three scenarios must occur:
1. The oil supply must grow fast enough for all consuming nations to be able to grow; or
2. US GDP growth must be able to divorce itself from the need for oil (“oil-less growth”), that is, maintain growth while reducing oil consumption annually by 1.5%, or
3. The US (and other OECD) countries must have a traumatic recession, because that’s the only way they’ll give up oil consumption.
Item 1. doesn’t seem to be in the cards; Item 2 implies that oil doesn’t matter (and many believe it doesn’t) or that the US and OECD economies can walk find a middle ground between economic growth and available oil. Item 3 is the standard, plain vanilla outome, and right now seems to be a good fit for recent events in the global economy. Hence all sorts of terrible things in the OECD economies, and pretty decent growth in the non-OECD.
So, to sum up, the approach presented here is mechanical, not statistical. To argue that the US will come out whole, you must demonstrate that either items 1 or 2 above will, or at least could, occur.
Jeffrey, excellent points, as always. That puts the US in an inexorable, Peak Oil-induced decline in per capita net energy terms, and thus in terms of real per capita private GDP.
Yet, the vast majority of economists examining the US economic conditions are 30-40 years behind the US domestic Hubbert Curve, if you will; therefore, by the time they “get it”, the US will be (is already) well past the point of no returning to the material standard of consumption of the peak Oil Age epoch.
Steven Kopits I understand your argument to be basically a supply side shock. Constrained resources lead to decreased output with limited ability to shift to substitute inputs. A world of highly curved marginal rates of technical substitution. My understanding of JDH’s argument is that he is making a quite different point. JDH is claiming that even though own-price demand for gasoline and diesel may be inelastic, consumers can and do make different consumption choices in terms of consumer durables. Of course, I could be misunderstanding JDH’s position, but when he said: “When oil prices make new highs, it may cause you to postpone buying a new car” I did not take that statement to mean that people would defer buying a fuel efficient car, but that they would defer buying a gas guzzler. So what he is talking about is a demand side effect in terms of shifting consumer preferences. This is also what I was getting at when I asked about the way price shocks are only remembered asymmetrically by consumers; that’s a demand side response. If you’ll recall our “small times small equals smaller” discussion, that too gets at the demand side. On the pure supply curve side of things the recent rounds of oil shocks are just not a large enough component of our economy to account for a recession. If oil shocks account for a recession, then the transmission belt has to be on the demand curve side.
A long time ago there used to be a big debate as to whether energy inputs should be treated as substitutes to capital or as complements. The old school view used to be that energy and capital were complements. And back in 1973 this was probably true. But more and more studies today are finding that energy and capital are substitutes, so when the price of energy inputs increases there is a capital response. This substitution effect is weak using cross-elasticities, but very strong using Morishima elasticities of substitution. Of course, all of this works itself out on the supply side, which is why I don’t think oil shocks have the same supply side effect today that they had 40 years ago. There might be economic drag, but it is coming from the demand side. Unlike the 1970s, maybe right now we could use a little Arthur Burns/Bill Miller style demand pumping in response to oil shocks. Maybe that composite Fed chair “Arthur Miller” was ahead of his time. 😉
Slugs –
I am arguing a demand-side shock. The oil supply has not fallen, it has expanded, but not by much. We can see the differences in the statistics between the two types of oil shocks. For example, in a supply shock, the rig building boom comes after the oil price shock, because it’s usually unexpected. In a demand shock, the rig building boom comes before the shock, because the price rise is more gradual and predictable. The shocks of 2008 and 2011 both fit the latter category, I would argue.
I understand Jim is using a different approach, and I will have to take time to digest it. However, I am suggesting that Jim’s approach and my approach must be reconciled. Whatever the conclusion, they must both hold true.
My more simple model, based on the necessity to transfer oil consumption and characterized by inelasticity-driven recessions, looks to be holding up better this week. Notwithstanding, betting against Jim’s judgment is a low margin game, so I’m prepared to be wrong. But whatever the model, it has to account for the need and method of transfering oil consumption to the non-OECD countries.
Last sentence should read: “But whatever the model, it has to account for the need, method and timing, of transfering oil consumption to the non-OECD countries.”
Professor,
It is easy to stop producing oil, but to make a well productioe again is not simply turning a valve. Also higher prices influence wildcatters to search for oil, but lower prices stop their search. Even if their search is a good project it is difficult to start it up again in both time and money.
But that said oil prices still do not drive an economy but are a reaction to economic events. But what this does illustrate is why when the economy surges into recovery oil prices increase so dramatically. Production cannot keep up with demand.
This can be seen dramaticlly illustrated in the deflation at the end of the last century driving oil prices down to production cost and causing a serious decline in production followed by a huge price increase in the early 2000s as demand outstripped supply. The high prices of the 2000s was a monetary errors of Alan Greenspan.
Prices need to be lower than before the shock and for longer for people just to get back to where they were consumption:asset:debt wise.
Steve,
Regarding an expanding oil supply, this is of course true if we include biofuels, but the EIA data show that annual global crude + condensate (C+C) production has been between 73 and 74 mbpd for 2005 to 2010 inclusive, except for 2009. Note that at the 2002 to 2005 rate of increase in global C+C production we would have been at about 86 mbpd in 2010.
BP data show that annual global total petroleum liquids production has been between 81 and 82 mbpd for 2005 to 2010 inclusive, again except for 2009.
But the BP data base (with minor EIA input) show that Global Net Exports* (GNE) fell from 45.5 mbpd in 2005 to 42.6 mbpd in 2010. Available Net Exports** (ANE) fell from 40.4 mbpd in 2005 to 35.1 mbpd in 2010.
The key point here is that ANE fell at average rate of one mbpd per year for the past five years.
*GNE = Net Exporters with 100,000 bpd or more of total petroleum liquids net exports in 2005 (99%+ of total global net exports)
**ANE = GNE less Chindia’s combined net oil imports
2slugbaits said:
“But today we only generate about 1% of our electicity from oil (why isn’t it zero?).”
Answer:Isolated electrical systems where oil (until recently) was the economic choice for electricity provision. Examples are islands and Alaskan villages.
Agreed, Jeffrey.
A couple of points on a very interesting post:
1. There can be little doubt that the subsequent decline in oil prices helped kill inflation in the early 80’s, which in turn allowed Volkler to take his foot off the breaks. This suggests a missing variable: to the extent that oil prices lead to generally increased inflation, you would expect to see a constrained money supply. This was clearly the case in response to the 79 oil spike, and explains the delayed effects. It was not the case in ’08.
2. The timing of the oil spike is going to matter, both in terms of labor negotiations (which obviously reflect to some extent the cpi number) and in terms of the timing of social security payment increases.
3. I looked at gas prices and unemployment, and was surprised that the R squared is actually pretty low. What is the R squared for this equation?
4. Obviously reference pricing matters here. It would be interesting to know if there have been academic studies of reference pricing.
Great post.
JJB, that’d be a start on SK’s expected shift of consumption toward developing nations.
Is it little surprise that the surge in oil prices started with the Iraq war lead military increases/Bush Tax Cuts coupled with the Chinida boom?
I have noted we need a “fiscal contraction” in some areas of the economy while we need a “fiscal expansion” in others. The US needs to inflate while the world deflates.
Steve Kopits But whatever the model, it has to account for the need and method of transfering oil consumption to the non-OECD countries.
Not following you here. I can understand how this would be relevant if the central question is one of predicting future oil prices and oil supplies; but for purposes of JDH’s post I don’t see why that’s necessary. As I read it, JDH is just treating the US economy as a price taker, so for purposes of this analysis all that counts is that prices go up for pretty much the same amount of US oil consumption. There could be a dozen different reasons for the price hike, but the bottom line is that for the US the supply curve shifts up and to the northwest (a first order supply shock) that is then followed by a change in consumer demand throughout the economy (an aggregate demand shock). If I’m following JDH’s argument, his claim is that the even if the supply shock is quickly reversed, the demand effects will persist. This is kind of an interesting angle because normally we think of supply shocks as being persistent and demand shocks as being transient (e.g., Blanchard-Quah VAR models).
2slugbaits said:
“But today we only generate about 1% of our electicity from oil (why isn’t it zero?).”
Someone already mentioned isolated areas. In addition, a half-million tons or so of petroleum coke for which there are no other buyers gets burned in coal-fired plants each year.
Fladem –
Tight monetary policy killed inflation. Volcker practised ‘tough love’ and forced the real economy to adjust. US oil consumption fell by 4 mbpd (twice the recent fall) from 1979 to 1983.
Jeffrey and I have been suggesting that we probably need to make a real adjustment now as well. If so, it stands to reason that a fiscal stimulus is unlikely to solve the problem, and why it hasn’t so far. If your real income has declined, then borrowing money to maintain an earlier standard of living will just lead to bankruptcy. So we either have to get used to a smaller transportation fuel budget, increase fuels production, or both. That’s the way forward.
fladem: The R squared is 0.23
Jeffrey Brown: About that 5 million bpd decline in ANE: Any idea on which countries are getting hit hardest in terms of per capita decline in oil consumption in the last 5 years or so?
The oil consumption decline has got to be steeper in southern Europe. A recent NY Times article reported on a bankrupt town in Spain that couldn’t afford to buy gasoline for police cars any more.
Steven Kopits: I would love to find some way to convince Americans to make the needed adjustment without severe recessions. But they seem mighty resistant to giving up their large cars and SUVs.
In the next recession fiscal policy will likely work in a cyclical fashion because governments won’t be able to afford to compensate for the downturn with more spending. The next recession will be far worse than the last one because we’ll go into it with a much higher unemployment rate.
What I’d like to see you oil-aware economists do: Model what a recession from high oil prices would do to unemployment starting from close to where we are now. If we are already at 8% or 9% unemployment what range of unemployment we likely to go up to? 11%? 14%?
JDH – that is pretty low. I would think you should mention that in your post.
I tried a bunch of different measurements, including lags and didn’t get that high.
Great work – but the correlation here isn’t huge.
Randall,
As Steve noted, it’s basically developed OECD countries, especially the net oil importers, showing lower oil consumption, relative to 2005, while the developing non-OECD countries are tending to show increased oil consumption. Broadly speaking, the developing countries are outbidding the developed countries.
As Steve noted, if we accept the premise that a finite world has finite fossil fuel resources (note that most people seem to equate this premise to a belief in space aliens), borrowing money in an attempt to maintain BAU consumption is, to use my metaphor, equivalent to stepping on the gas as we drive toward the edge of the cliff.
Jeffrey,
Is there a good source of oil consumption trends by country?
I’m guessing there must be some losers outside of the OECD who are also being outbid for oil. Perhaps some central and south American countries. Maybe some non-oil producing Middle Eastern countries. Tunisia, Jordan, and Syria come to mind. Ditto those African countries which are not making a lot of money from natural resource exports.
I’d like to also see a list of countries sorted by their percentage of total energy consumption that comes from oil. My guess is those with the lowest use of oil as part of their total energy mix are experiencing higher economic growth. That certainly describes China with its heavy reliance on coal.
How has the most recent activities (fall of Libya) changed this? I hope the price is going back down.