Today, we are fortunate to have Luciana Juvenal and Ivan Petrella, as guest contributors. In this post, they respond to Wednesday’s guest contribution by Lutz Kilian, entitled Oil Price Spike Exacerbated by Wall Street Speculation?.
Disentangling the main drivers of oil prices is a critical first step for allocating resources and designing good policy.
In our paper, “Speculation in the Oil Market,” we assess the roles of speculation and supply and demand forces as sources of oil price fluctuations.
Our main finding is that global demand has been the primary driver of oil prices.
An expanding global economy increases the demand for raw inputs, including oil, which pushes up prices. Borrowing from our abstract and conclusion our main message is:
“Our results support the view that the recent oil price increase is mainly driven by the strength of global demand but that the financialization process of commodity markets also played a role.”
“Our results highlight a major challenge faced by policymakers in the medium to long-run: Although speculation played a significant role, the high oil prices witnessed in the past decade are mainly due to demand pressures, which are likely to resurge with the recovery of the world economy.”
Our conclusions are thus in line with a large strand of the literature, including Kilian’s own contributions.
Kilian’s blog post presents a misleading interpretation of our study. His comments are partly triggered by how some media have interpreted our paper, which we are naturally unable to control. A non-technical summary of our paper can be found in our Economic Synopses. It turns out that we do not disagree with most of the substantive conclusions that Kilian puts forward and a full reading of our paper should make that clear.
Kilian refers mainly to a report in the Huffington Post, March 20, 2012. He ignores other media reports that have provided an accurate interpretation of our paper. As an example, see Brad Plumer’s post on WonkBlog of the Washington Post. Quoting from that blog: “Indeed, the idea that speculators play a secondary role is essentially what the St. Louis Fed paper, written by Luciana Juvenal and Ivan Petrella concluded. They found that about 44 percent of the price increase between 2004 and 2008 — when oil zoomed up to $140 per barrel — was driven by shifts in global demand. Countries like China, India, and Brazil kept using more oil, while production from countries like Saudi Arabia couldn’t keep up. But an additional 15 percent of the price increase was caused by speculators.”
Aside from highlighting only one, inaccurate media report about our paper, Kilian criticizes our work on various technical issues. We would like to respond to his main criticisms.
- Identification. Kilian notes that we identify speculation as “an exogenous upward shift in oil producers’ expectations of the real price of oil (possibly driven by the financialization of oil futures markets).” Under this condition supply and demand move in opposite directions and we are clear about this mechanism in our paper (see Appendix E). Kilian says that this hypothesis is not an implication from economic theory. However, our paper provides a framework that reconciles our analysis with standard theory of exhaustible resources (Appendix E).
- The 1990 oil shock. Kilian writes that according to our analysis, oil production should have declined after the invasion of Kuwait. In our model, however, this episode is captured by the oil inventory demand shock. As explained in our paper, this shock arises in the presence of uncertainty about future oil supplies, driven, for example, by political instability in key oil-producing countries (as is the case of the invasion of Kuwait). A positive oil inventory demand shock raises demand for inventories, causing the level of inventories and real oil prices to increase. Inventories of crude oil increase so that supply can meet demand in the event of supply shortfalls or unexpected shifts in demand (see Alquist and Kilian, 2010). The increase in the real price of oil provides an incentive for oil producers to increase production. Therefore, our analysis suggests that oil production should increase after the invasion of Kuwait.
- Is the oil inventory demand shock redundant? Kilian writes that adding our speculative shock renders the oil inventory demand shock proposed by Kilian and Murphy (2011) redundant. If, however, this were true we would not be able to separately identify the oil inventory demand shock and the speculative shock. In fact, we use a simple model in Appendix E to inform sign restrictions on the two shocks that distinguish between their effects on supply. We identify speculation as having a negative impact on supply. A shift in the oil inventory demand need be not matched by a contemporaneous shift in supply. For example, any shift of the convenience yield, such as the one modelled in Alquist and Kilian (2010), also implies an increase in total demand (specifically the precautionary demand for oil inventories) but does not imply a contemporaneous shift of the supply curve. Kilian then argues that we claim that the Kilian-Murphy model is wrong. We never write or imply this anywhere in our paper. Quite the opposite, we build on their contribution.
- Differences in approach. Kilian affirms that “One obvious difference is that the original Kilian-Murphy analysis was based on uncontroversial implications of standard economic models rather than ad hoc assumptions without basis in economic theory.” To the best of our knowledge, Kilian and Murphy (2011) do not present any theoretical model to guide their identification strategy. Their reference to theory is Alquist and Kilian (2010) which is nested in the simplified model we propose in our Appendix E.
- Speculation and real economic activity. Kilian describes that our findings indicate that a speculative shock is associated with an increase in real economic activity and suggests that this “would imply that what is needed in today’s economy are more speculators in oil markets.” Our paper does not suggest this. In fact, we do not impose a sign restriction on the response of real economic activity to a speculation shock as there are two forces that operate in opposite directions. In principle, the oil price increase would have a contractionary effect on demand and economic activity. We are not comfortable imposing such a restriction because we do not want to rule out the possibility that an increase of financial speculation is triggered by low real interest rates, as suggested by Frankel (1986 and 2008). As Frankel explains, low interest rates may have a number of effects on commodity markets. On the financial side, lower real rates reduce the cost of “carry trade” in the commodity markets, amplifying the effect of a mismatch between expected future spot prices and futures prices. In the physical side of the market, real rates represent the opportunity cost of holding inventories both above and below ground. This channel would imply a positive effect on real activity (see Frankel and Rose, 2010), therefore we chose to remain agnostic and not impose any restriction. This is also consistent with the framework we provide in Appendix E of our paper.
- Speculation in the last decade. Kilian argues that the fact that the historical decomposition is flat when many pundits suspected particularly large speculative pressures, is puzzling. We disagree with his conclusion. We discuss a simple explanation for this pattern in our paper: “Another feature of interest is that the contribution of speculative shocks to oil price increases becomes flatter from 2007 until 2008. This highlights that the gains from speculation decrease as the oil price goes up.” We give a simple example of why speculation is less likely to be present when prices are higher. Quoted from our paper (footnote 29): “Let us illustrate this claim with a simple example that applies to contango periods like the one observed in 2004-2007. Suppose that the spot price is 30 USD, the 1 year forward price is 60 USD, the interest rate is 10%, and there are no storage costs. An investor would borrow 30 USD, buy oil, wait for delivery and sell it for 60 USD. The total cost for the investor is 33, and the revenue is 27. Now assume that the forward curve shifts upwards, so that the spot price is 100 USD and the forward price is 130 USD. In this case the total cost for the investor is 110 USD, and the revenue is 20 USD.” Moreover, let us add that, according to the dating in Singleton (2011), the oil market was in contango between 2004 and 2007 and backwardation during the spike of 2007-2008.
Overall, we believe that we have more points of agreement with Kilian than disagreement. We encourage the reader to have a look at our paper.
This post written by by Luciana Juvenal and Ivan Petrella
Lutz Kilian responds in a comment at 7/28, 11:40AM, with link to comprehensive remarks. — MDC
I’m of course totally confused by all this. That said, my concern regarding speculation is that it increases short term price volatility and thus increases the costs born by the market. The place simply becomes more expensive.
If the amount of liquidity in these markets has grown dramatically, and I believe it has, that is a clear red flag warning of increasing speculative activity.
Well, let’s start with some of these items. You describe a $30 spot, $60 forward price. This has not occurred, but a similar situation did in early 2009, when spot prices were low. Oil trader Phibro leased a bunch of tankers to hold oil inventory and made a pile of money. So much, in fact, that Vikram Pandit over at Citi was forced to divest the unit. However, what Phibro knew was that the marginal cost of oil production is around $70 and that all countries can materially increase consumption at $70–which was considerably above the forward price in the mid-$50s at the time. Was Phibro speculating (taking a long position using physical inventory build) at the time, well, duh, yes. But no one cared, because the price of oil was low. Indeed, Phibro was doing exactly what speculators are supposed to do: support the price during periods of crisis.
Now, there was exactly one day, April 1, 2008, when the spot WTI was $100 and the forward (4) at $130. No one was putting oil on tankers at the time, although it was a period of inventory rebuild after the fantastic drawdowns of H2 2007. I personally had predicted a price spike into 2008, because it was clear that supply was not keeping up with demand and the inventory draws of late 2008–about 2 mbpd, more than 2% of global consumption–were patently unsustainable. Therefore, when the inventory holders sought to rebuild their stocks (which began in mid Q1 2008), the supply-demand re-balancing would swing prices sharply. And they did. That’s when the big gap opened up.
Remember also at the time Arjun Murti (know who he is?) was calling for $200 oil and Matt Simmons (know who he was?) was calling for $500 oil. No one had a clue what was going on. I was at an investment bank at the time, and I chatted regularly with our traders. No one understood the dynamics, although we were all pretty sure it would end badly.
Thus, your analysis seems short on historical specifics, and you are referencing hypotheticals without regard for either the carrying capacity of the global economy for oil prices or the marginal costs of oil production. Oil traders are very well informed on these topics, apparently better than you are.
“Speculators” are folks like little ol’ W.C. Varones, who has the Pimco Commodity Total Return Fund (PCRRX) in his 401(k).
“Speculators” only “speculate” on commodities because we don’t have a sound dollar, and reckless fiscal policy, abetted by easy monetary policy, promises more of the same for decades.
God forbid old folks should be allowed to protect themselves from inflation.
Menzie
Thank you for this “point – counterpoint” presentation of current research. It’s refreshing to see 2 sides of an important issue. The comments by experts in the field, like Steven Kopits are appreciated as well.
tj Thank you for this “point – counterpoint” presentation
You mean as opposed to old line from the SNL skit: “Jane, you ignorant slut.” ;->
But I agree.
When crude-oil prices surged in July 2008 to a record $145, that surge was preceded by a very significant drop in trading volumes at the NYSE and etc. It would seem then, especially considering the difference in margin requirements, that a vast wave of liquidity drenched the oil markets. This could be explained, I suppose, as a price discovery shift, but… it seems more likely that a combination of animal spirits and excess liquidity caused the 2008 super-spike, as opposed to the pent up demand explanation. How could there be that much of a difference in demand in the midst of a global meltdown?
This spike of spikes though seems to lack a clear connection to the liquidity tsunami in either of these papers. Or, do my reading skills need improvement?
Thanks for an interesting presentation here.
Luciana and Ivan –
You both look pretty young, so I’ll give you some Managing Director insights today. I write and edit reports all the time, and I oversee the production from junior analysts. So I have some professional background in the matter. Some thoughts:
When you write in mumbo-jumbo technical sytle, you’re making it hard for the reader to understand what you’re trying to say.
For example, this sentence: “This will happen if the oil price is expected to increase relative to production costs and current production is reduced as producers withhold some energy resources to sell at a greater profi t at a
future date.” might be followed by this sentence:
“Thus, by this theory, we might expect Saudi Arabia to shut in production when oil prices go up.” With that, you have a simple example to help the reader understand the previous sentence–as well as an easily testable thesis (which I believe is dead wrong in the last ten years). But OK, I can understand what you’re getting at.
If you want to know how to write a nice paper, do it the way Jim Hamilton does. He writes theoretical stuff, but I can always easily understand what he is trying to say. You can write to prove you’re smart, or you can write to communicate. If it’s the latter, write clearly so that an educated and interested, but busy, layman can understand. Simple examples are helpful.
Second, if you’re going to be committed to energy economics, then be committed. I encourage it. But don’t dabble. (Both of you are dabblers.) My chief gripe with economists is that they jump into the topic, throw a bunch of apples-to-oranges statistics in a stats blender, and say they know something about the business.
The way to learn the business is to do historical work. Again, the professional model for me is Jim. He very deliberately works through topics, often going back a long time. But as a result, he really knows his historical narrative and has a really good feel for the data. He knows what an apple looks like and what an orange looks like and how they differ. If you want to be energy economists, that’s the foundational work I would do. (And I have.) More in next post.
Juvenal and Petrella’s rejoinder raises more questions than it answers. For example, if their results are so much in line with the existing literature, why do they obtain the opposite conclusion regarding financial speculation? Have they really been misunderstood by the press? Does the additional evidence they present help make the case for financial speculation? Is their identificaton valid after all? Here is a discussion of why the answer to all four questions is no:
http://www-personal.umich.edu/~lkilian/tempkilian.html
Pick up the damn phone and make a call!
Did you sense check your thesis with Phibro, or an i-bank, or other oil analysts? Are there some pithy quotes in there?
Whenever I receive a forecast from our analysts which I find outside the mainstream consensus, my first act is to call someone. I might call Shell, or UBS, or FMC, or Phibro. Some of these folks I know, some I don’t. I have made about 1000 colds calls in my job–4 people have elected not to speak to me.
The hardest part of cold calling is physically reaching people. But if you can, here are some tips to help you:
Recognize their expertise
Tell them that you understand they are an expert–indeed, a leading expert–in the matter and you would like to ask for their insight. This has the added benefit of being true! That’s why you called them! But there’s probably not a middle-aged man alive you does not like to be considered a leading expert in his field. And he’ll want to demonstrate that to you.
Ask two good questions
Remember, a lot of the people you are calling are smart, well-intentioned people. They went to places like Harvard and Princeton. They know what a term paper is; they know what research is; they know what a cold call is; they know what the Fed is. They are nice people and they want to help society. (Really. And that includes investment bankers and almost the entire gamut of equity analysts.)
To capture their interest, try to ask a couple of good questions. I like conceptual questions. For example, you might say, “I’m a researcher at the Fed, and I’m trying to write a piece on how financial markets affect oil prices. But in truth, I’m not even sure how to define the word speculation. How do you think about it?” Or, “Do you think speculation is a real issue, or is the whole thing overblown?” Let them state a thesis and then pose a consideration: “But if that were true, wouldn’t prices have gone down instead?”
Provide Information in Return
The first few times you do this, it will be hard, but once you’ve got a few calls under your belt, it will be easier: “I was talking to Goldman Sachs, and they felt that this model was likely closer to reality. Do you agree with that?” Well, now the interviewee has learned something. Most the conversations will be private, but not confidential. Don’t abuse confidentiality–or privacy. But you’ll be speaking in generalities anyway most the time, so it shouldn’t be an issue.
Be fun to talk to
I am increasingly informal as I get older. And I make jokes–partly because that’s the kind of guy I am. But I strive to make my calls converations rather than interviews. I want the person I’m talking to to have a good time. That’s another thing you can provide. Also, remember that it’s important to be liked (especially in Texas).
The Interviewee Runs the Call
The guy talking to you is doing you a favor. Never forget it. If they want to talk about something a little different, let them. If you feel they’ve had enough, thank them graciously and let them go–even before they do so. I am always grateful to the people who have helped me–and there are hundreds of them.
Be Ethical
The interviewee has the right to know to whom and why he is giving information. Many times, we are under confidentiality–so I can’t give my client’s name. But I will generally provide the nature of the inquiry and the type of client. There are some exceptions (generally I can’t disclose if I am working on a transaction), but in general, I try to be as forthcoming with the interviewees as I can.
So, when preparing a paper, pick up the phone. Ask people. Get opinions and perspectives. Trial balloon your thesis. Interact. You’ll get a better final product and avoid otherwise obvious pitfalls and traps.
Your paper suffers the weakness that you really don’t understand the industry, and there’s not even the remotest evidence that you talked to anyone about how the real world works. Next time out, pick up the phone and ask. People will help you.
Importantly, the results of both Jim’s and Lutz’s work have been negative–they have debunked the notion that financial speculators cause high oil prices. But this is not entirely satisfactory, because we then lack a convincing reason that oil prices can exceed long term sustainable prices–just as they did earlier this year. If the world can’t sustain $125 oil, well, how did the price get there?
In “High Oil Prices are Caused by Consumers, not Speculators” (http://www.europeanenergyreview.eu/site/pagina.php?id=3761), I offer a potential explanation. Notably, price inelastic demand means that consumers reduce consumption too slowly in the face of upward demand (or downward supply) shocks. Thus, price inelasticity causes oil price shocks–what you mean by “speculation”–endogenously.
This, to my mind, is the cutting edge thesis on the topic. If you want to do a quantitative analysis, this is the hypothesis I would test.
@Kilian, 11:40AM : your link is not working, guess its with “.html” at the end :
http://www-personal.umich.edu/~lkilian/kilianblogreply.html
@Steven K.: now you are a helpful guy !
Very interesting indeed, and thanks for some great comments Steven.
Prof. Hamilton has also discussed this topic earlier. Read:
https://econbrowser.com/archives/2012/04/a_ban_on_oil_sp.html, and
https://econbrowser.com/archives/2012/04/more_on_specula.html
None of the papers are differentiating the cost of carry between free money, derivatives and cash.
On the concept of “backing up hunches with money trough contracts in futures” and “low rates driving stocks increase”.
Derivatives are for free for the large financial operators as a BBB bankers blanket bond do suffice the brokers. Since the Glass Steagall act has been repelled, most of the large brokers in futures became Banks and Banks holdings subsidiaries. The cost may be a slight impact on balance sheet contra accounts,
None of the papers include the strategic reserves in the oil price buffer, when even the draw down in stocks are far from leading to a coincident fall in prices (Please see IEA report P 6 P7 and P24) The time lag in price response is long in context.
http://www.eia.gov/analysis/requests/ndaa/pdf/ndaa.pdf
The same paper P23
Even when cumulating the most adverse prices scenari,demand chock, speculative chock, adding oil inventory and an oil supply chock the content is falling short of 20 % as an explanation to oil price.
The paper is showing a substantial work content but difficult to synthesize.
@ppcm Since you referred to both papers in your comment, let me throw in three clarifications:
1. The crude oil inventory data used in Kilian and Murphy (2011) include strategic oil reserves. My presumption is that Juvenal and Petrella include those reserves as well.
2. In the Kilian-Murphy model the four structural shocks jointly explain the observed real price of oil by construction. The Juvenal-Petrella paper is different in that regard because they embed their VAR within a factor-augmented model with additional unspecified shocks.
Their estimates on the importance of the identified shocks are expressed relative to the total for those identified shocks only, which differs from Kilian-Murphy. Qualitatively that makes little difference, however, for thinking about how the results differ.
3. The cost of carry is allowed to change under our identifying assumptions, but it is not necessary to disentangle why traders do what they do for the Kilian-Murphy model to work. More importantly, it is not necessary to know this to be able to rule out financial speculation. This conclusion relies on the fact that we do not find any evidence of speculation driving up the real price of oil, which rules out financial speculation for whatever reason in passing.
Note that this argument works for Kilian-Murphy, but it does not work for Juvenal-Petrella who report some evidence of speculation. In that case, one has to ask where that speculation is coming from and the latter question their model does not answer.
I still don’t get it. I understand that its zero sum. But I also understand that the field of economics is rife with bs and that models are often misleading.
So, in simple terms, from 2003 to 2008 the global economy expanded by about 25%. But oil prices rose from around $25 to $145, or nearly 600%. But speculation played no role because demand had outpaced supply throughout the period in a subtle way.
But there were no oil shortages noticeable enough for many thousands of analysts to notice. These analysts though are indispensable in regards to price discovery.
So during this period when supply lagged behind demand, in stealth, was global growth held back? Were there refineries in lack of crude?
I was old enough to drive back in the ’70s, and so, I know what happens when the supply of oil runs short. So is there some commensurable effort that includes a consideration of the volume of futures trading during these 2 periods of ‘supply’ shortages?
I really am trying to understand.
Ray Lapan-Love: You seem to be saying that the price should only be going up if there are actual shortages. But in a properly functioning competitive market, there are never any shortages. The price always adjusts to make sure you don’t have to wait in gas lines. The reason you remember waiting in gas lines in the 1970s was because the oil companies were prohibited by law from raising the price to where it should have been.
As for the form of evidence used, people have approached this question using a variety of different methodologies. The point-counterpoint of Kilian and Juvenal-Petrella is just one approach. Perhaps you would also be interested in some other ways of analyzing these questions, such as this one.
@ Ray Lapan-Love:
Your write the global economy expanded by 25%. I presume that’s a reference to the cumulative growth in real world GDP? Keep in mind that real GDP is a poor proxy for the use of industrial raw materials including crude oil. The measure of global real activity I developed in an effort to capture fluctuations in the demand for oil shows much larger cumulative changes (first up, then down in 2008 and back up in 2009) than real GDP. The data for this index are available on my homepage. In fact, it lines up remarkably well with observable fluctuations in the real price of oil during 2003-2009. There is nothing subtle or stealthy about this demand story. It is basic economics.
The shortages you speak of were reflected in price increases and you saw them reflected in your bill at the gas station. I very much agree with JDH’s analysis of this point. Refineries bought at market prices what they thought they could sell in the gasoline market. Indeed, in the U.S. gasoline consumption slowed down considerably, but the slack was taken up at the global level by higher demand from countries such as China.
You question of how the real price of oil held back global growth ignores the two-way causality between the economy and the real price of oil. It was global growth that drove up the real price of oil in the first place.
As to the role of oil futures during the 1970s, there is no work on this question because oil futures markets were only created in 1983 and really matured only in about 1989. The evidence we have regarding the interaction of oil futures markets and the real price of oil paid by refiners relates to the post-1989 period. Of course, we don’t need oil futures for traders to take speculative positions in oil. That happened in 1979 already, as is well documented by oil historians.
The key point is that the results I was referring to are fully consistent with the evidence on the 1970s. In fact, my research on this topic had started many years ago with the 1970s and highlights that there are striking parallels.
Mr Kilian
Thanks for your prompt explanation and invitation to dissociate the conclusion as drawn from
Kilian- Murphy model and those as drawn from Juvenal- Petrella paper.
It may be inappropriate to allocate all the movements on derivatives to the proprietary trading of Banks, but surely the increasing revenues are a deviation to trend under the items commodities.
The OCC report (P 5) on derivatives as of 4th quarter and for the full year 2011.
Drawing on both the definition and conclusion of Kilian- Murphy model, a speculator being “ anyone buying crude oil not for current consumption and not for future use” one may assume a large undetected pick- up in economic activities.
The conclusions heavily borrowing from semantic may be opposed to GAO report and findings.
“Regulators will need more information to fully monitor compliance with regulations when implemented”
@ppcm:
The Kilian-Murphy paper is first and foremost about the question of why refiners have to pay more for oil (and hence consumers have to pay more for gasoline).
Proponents of regulating oil derivatives markets have argued that speculation by index fund traders has been responsible for driving up the real price of oil paid for by refiners.
This argument relies on arbitrage between the oil futures market and the physical market for oil.
Since the late 1980s, there were two ways to speculate: One is to buy physical inventories of oil in anticipation of rising oil prices; the other is to buy oil futures contracts. Arbitrage ensures that if someone speculates for whatever reason in the oil futures market, someone else has to be taking speculative positions in inventories. As long as the price elasticity of oil demand is not literally zero, this is the only way for speculative trading of the type the GAO is concerned with to drive up what refiners pay for oil.
One can spend a lot of time thinking about what form improper speculation in oil derivatives market may take, but in the end it comes down to the fact that there is no sign of speculation having caused an accumulation of inventories in the physical market. In this sense, the GAO discussion is missing the central point in this debate.
Moreover, measuring and detecting excessive speculation in the oil derivatives market is difficult, if not impossible. The mere fact that certain traders are taking speculative positions in the oil futures market does not mean these traders “caused” the oil futures price. They in turn may have been responding to ordinary business motives. Nor is the fact that some entities enjoy high revenues or rather high profits proof of wrong doing. Speculation is risky business and that has its reward. On the issue of what speculation is and how it can be measured see the discussion in:
http://www-personal.umich.edu/~lkilian/milan030612.pdf. An even briefer synopsis is here:
http://www.voxeu.org/article/speculation-oil-markets-what-have-we-learned.
If Kilian-Murphy had found evidence of speculation during 2003-08 this may have been a sign of expectations of rising demand, but we do not. The pick up in demand that we do find, is anything but undetectable. It is in plain sight and corroborated by many different methodologies.
Lutz:
“One can spend a lot of time thinking about what form improper speculation in oil derivatives market may take, but in the end it comes down to the fact that there is no sign of speculation having caused an accumulation of inventories in the physical market. In this sense, the GAO discussion is missing the central point in this debate.”
This is probably nothing new to anyone here, but there was a WSJ article (2009?) that claimed that if all of the tankers being used for storage, were put end to end, those tankers would collectively be 27 miles long.
JDH,
Since we seem to be putting a fine point on things, I have a little dispute with this:
“The reason you remember waiting in gas lines in the 1970s was because the oil companies were prohibited by law from raising the price to where it should have been.”
Would it not be better to say that oil companies chose to restrict supply at the applicable price level. Were those companies not speculating that oil in the ground was worth more than oil in the barrel? I happen to work in the oilfields and I can tell you with no doubt that prices are less a matter of supply and demand than they are subject to the opening and closing of valves and etc.
@ Ray Lapan-Love:
Journalists have the advantage of being able to write what comes to their mind, whereas economists have to back up their views by data and evidence. An obvious question is what the data source underlying this claim is. It is government sources. The EIA, for example, does not keep track of tanker storage.
This is a difficult subject. Here are a few things I know:
(1) This is a very recent phenomenon. Reports of storage on tankers surfaced only in the last stage of the run-up of the real price of oil which makes it difficult to attribute the bulk of the surge to this form of inventory holdings.
(2) Anecdotal evidence suggests that much of the increase in tanker storage occurred only after the collapse of the real price of oil, when oil had nowhere to go, given falling demand. This type of tanker storage has nothing to do with the notion of speculators driving up the real price of oil during 2003-mid 2008. It would be interesting to see what exact time period the WSJ was referring to.
(3) One way of controlling for this phenomenon would be to reestimate the Kilian-Murphy model on data ending earlier (say in December 2007). I have estimated such models and the answer does not change materially. The surge remains about flow demand, consistent with other independent evidence.
(4) When thinking of storage of oil on tankers, we have to abstract from oil in transit and focus on stationary tankers filled with oil. Did the WSJ account for that difference?
(5) Something I do not know is how stationary tanker storage is treated by statisticians constructing oil inventory data. These tankers are moored close to shore, where the demand for oil is, so one would think that they would be part of the national statistics, but we cannot be sure.
I am in the process of procuring alternative oil inventory data from private sector sources, which will allow me to check on this question.
(6) For now my sense is that the inventory data currently being used in models are likely to be adequate at least until mid-2008. One indication that I am correct about this, is that the oil futures spread after 1989 should be able to predict the remaining model variables, provided the inventory data were mismeasured and hence informationally deficient. In our paper we show that there is no extra information in the oil futures spread.
I apologize to readers who were unable to access my earlier detailed replies on Juvenal and Petrella’s rejoinder. Apparently, the *.mth link worked for some readers (including myself), but not other readers. Here is the new link:
http://www-personal.umich.edu/~lkilian/tempkilian.html
Lutz,
I found the following and it is not the WSJ article… but informative:
Posted by Gregg Gethard on November 19, 2009 at 12:37 pm
(image: msnbc.msn.com)
An oil tanker. (image: msnbc.msn.com)
“Tankers move oil around the globe, but some of these ships aren’t currently sailing.”
“According to a Financial Times article published on Tuesday, one in 12 of the world’s biggest oil tankers are being used to store oil and oil products—in many cases, heating oil—as opposed to ship it. Two factors have caused this: oil consumption and demand are at incredibly low levels while futures prices have been climbing. This has created an incentive for investors to purchase oil now in order to resell it later at a big profit. While investors are waiting for oil demand to rise and bring prices along with it, their actions nonetheless affect today’s crude and heating oil prices, which many observers consider higher than existing demand can justify.”
“As of October, 129 ships were being used as oil shortage, according to a London-based shipbroker. Rates for crude oil carriers currently go for $28,000 per day; if this level of usage remains the same, those rates could go up to $35,000 per day.”
Ray,
Thanks, that’s helpful:
(1) The article makes explicit that some of these tankers are used to store heating oil (a refined product) rather than crude oil. Inventories of heating oil are not part of the quantity measure that matters for models of the real price of oil, so these inventories are not relevant to our discussion. We do not know what the actual number of tankers storing crude oil is (or whether these tankers are already included in the EIA data we use), but the relevant number of tankers is clearly smaller than that suggested in the WSJ.
(2) What matters for our econometric model is the change in inventories, not the level of inventories. The press articles focus on the level.
(3) The reference to low levels of oil consumption reinforces my point that this phenomemon might be an issue after the oil price collapsed, but is not an issue for the period of the run-up in real oil prices, which we were discussing initially and which Juvenal and Petrella tried to explain.
(4) In terms of the Kilian-Murphy model, an unexpected decline in flow demand all else equal would cause the accumulation of oil inventories over time. That’s what the model predicts, and that’s what happened after the 2008 recession hit. Thus, this evidence seems fully consistent with the propagation mechanisms of our model and is not an indication of exogenous positive speculative shocks.
Ray –
Arbitrage is not speculation. If I can buy oil today for $70 and sell it today for $95 for delivery four months hence, I am not speculating. I have no open financial position at all. I have locked in the price. I am merely being paid for having available physical storage–and handsomely paid for it if there is a sharp contango. But I am not a financial speculator–at least not in oil markets.
By arbitraging, I will be increasing the current oil price and lowering the future price by presumably a similar amount. I am flattening the curve (reducing the contango), but I am not driving up the price of oil overall.
When we speak of “speculation” here, and leave out the possibility of inventory accumulation, we have only two possibilities, I think.
One of these is an investing bubble, in which market participants believe the price of something will move in a way unsupported by fundamentals. The US housing market in 2006 is an example of this, although we would not call the average homebuyer in this period a “speculator”. In a bubble, there are a set of market beliefs, often predicated on the greater fool theory, as in a Ponzi scheme. But there is no actor, as such, manipulating the market. People weren’t buying houses just to drive up the market. They were buying houses because “housing prices always go up”, or because they thought they would be able to sell the house for even more to someone hoping to sell it for more in turn (the greater fool theory).
The other option is price manipulation, in which a seller creates hype to induce the market to overvalue an asset or uses collusive behavior to run up a stock. The former is legal, and of course, is the raison d’etre of a good investment banker. Facebook comes to mind.
The latter is not legal, but no one to the best of my knowledge has argued that this sort of behavior was widespread in oil markets. On the other hand, I wouldn’t be surprized if it is fairly commonly in thinly traded markets. But of course, who the heck wants to write about iridium or vanadium–or parafin. So why don’t Luciana and Ivan write about the parafin market (which really is a fantastic story of market manipulation).
Because oil matters. Voters call up their Congressmen and complain. The high price of oil causes pain. When the oil price exceeds the carrying capacity of an economy, people suffer. L&I, at the end of the day, are uninterested price manipulation or “speculation”. If they were, they would have chosen a different market. Oil is perhaps the single worst commodity market for price manipulation–it is large and deep. It is absolutely the last commodity market where you would go looking for price manipulation.
Rather, they are interested in oil because of its social impact. They believe there’s something special about it. They are responding to the emotion–and rightly so, because emotion signals relevance. But what is the emotion they are responding to? That emotion says, “The price of oil is too high, and it is reducing my standard of living in a painful way. I can’t afford it.”
So the issue here is not speculation, it’s unaffordable oil. Market forces should not be able to lift prices above generally affordable levels in the absence of a negative supply shock. A bidder can’t price himself out of the market. (Or can he?)
So how can the oil price exceed the tolerance of society? One explanation is “speculation”, by which L&I must mean either an investing bubble or price manipulation by parties certain. How one manipulates price without inventory accumulation except over very short periods is beyond me. And Lutz has spent a good bit of time burying the whole notion.
So then you need another actor to explain unaffordable oil. I would argue this is intrinsic to the notion of price inelastic demand.
It’s literally tautological. If there’s a demand shock and the consumer fails to adjust in the absence of supply accommodation, then the price must—by definition–rise above the long-run sustainable price to change behavior. You have to break the consumer’s will. The pain is not a bug, it’s a feature. It’s what makes people let go.
That’s the topic worthy of analysis.
I fully agree with Steven. The short-run price elasticity of oil demand is key. In fact, this is a topic that has been studied in detail already. Kilian and Murphy (2011) have a whole section in their paper about estimating the price elasticity of oil demand. We provide evidence that this elasticity is much higher than many economists used to think. The difference can be traced to the fact that traditional estimates of low oil demand elasticities were constructed from reduced-form regressions that we know to be biased toward zero. For example, on our data set traditional regressions yield a demand elasticity of -0.02, whereas the properly constructed structural estimate is -0.26. Even allowing for the fact that the latter elasticity may have declined recently, the notion of a zero price elasticity of gasoline demand (which can be shown to be about the same as the corresponding price elasticity of oil demand) seems empirically implausible.
Steve:
“By arbitraging, I will be increasing the current oil price and lowering the future price by presumably a similar amount. I am flattening the curve (reducing the contango), but I am not driving up the price of oil overall.”
I have trouble accepting that the assumption that any increase in current price is a benefit. If supply is beyond demand in the current market then why would the future price not benefit too?
Ray –
An increase in the current price might not be a benefit. All the futures curve tells us the is the time preferences of the market; it doesn’t make more judgments. Traders respond to that. But in the case of late 2008/early 2009, putting oil on tankers made good economic sense, because the replacement cost of that oil was greater than the market price. Speculators did exactly the right thing, by not letting a valuable resource be squandered at a time of market dislocation.
Lutz –
I am inclined to believe that elasticity is a function of the price level. So it’s like the power band for a car accelerating. We can talk about 0-60 mph time which gives us an average acceleration. Or we can talk about 50-60 mph acceleration, which will be materially different than the pro rata share of the 0-60 mph time.
So it is with price elasticity. It’s not the same at $70, I think, as at $120 Brent.
Moreover, the elasticity will depend on who’s doing the stretching. If the demand shock (upward) is affecting China, then China’s increased demand will increase the oil price–thus the price elasticity is actually positive. We saw this from mid-2009 all the way until the last couple of months. China’s demand consistently increased with the oil price (or vice versa, actually).
If the US is neutral at the time (ie, holding consumption steady), then China’s upward shock will lead to a flat to negative elasticity, depending on how elastic US consumption is at that point.
Further, it seems to me that society can toggle between elastic and inelastic modes. As I believe Jim pointed out, inelasticity makes sense if the consumer believes i)the shock is transient or ii) believes that income or wealth will rise to offset the effect. (It also helps to be able to iii. increase your debt or reduce your savings rate.)
I could argue (from a logic point of view), that once conditions i and ii (and/or maybe iii) are no longer met, elasticity will be negative beyond the carrying capacity point. Hence, US oil consumption started declining two years ago when Brent hit $85.
Now, it should be possible to arrange a kind of test of this hypothesis. We can take two persons (countries if you like) and assign them some price inelastic value, say, -0.2. Thus consumption declines modestly with increased price.
Now, allow the supply constant, what happens if country A sees a positive demand shock, and country B doesn’t. What happens? It seems to me, the oil price must increase beyond the carrying capacity of Country B.
In any event, if you have an analysis like that, I hope Jim and Menzie will let you write a post on it.
Steven,
The analysis in question is in a pdf file on my homepage, so anyone can have a look. Some clarifications:
(1) I agree that intuitively the price elasticity of oil demand depends on what happens in different parts of the world, say China and the U.S. Of course, when I am referring to the price elasticity of oil demand in my global model, I am doing so at the global level already (as opposed to referring to the price elasticity of oil demand in China or in the U.S. separately).
(2) The price elasticity of oil demand (roughly speaking) refers to the ratio of the quantity response over the price response triggered by an exogenous shift in the oil supply curve along the oil demand curve. Unless one knows for sure that such a shift has taken place at a given point in time (and that no other exogenous shifts in demand have taken place at the same time), one cannot infer the price elasticity of oil demand directly from observable data. One needs a full-fledged structural model for this.
In particular, if the observed change in price actually arises from an exogenous shift in oil demand (while keeping the supply curve fixed), the same observable movements in the data will capture the price elasticity of oil supply.
(3) There are many reasons to believe that most of the fluctuations in the real price of oil in recent years were associated with shifts in the demand curve (abstracting from the Libyan supply shock, but that event was short-lived and comparably small). Your example of China’s increased demand is a case in point.
(4) I agree that the price elasticities may evolve over time, as indicated in my previous entry, but pinning down that evolution is very difficult in practice. Some preliminary work on this point is in a forthcoming paper by Baumeister and Peersman (2012). My point was merely that a perfectly inelastic price elasticity seems empirically implausible.
(5) Actually, when you refer to “inelastic demand”, you mean a price elasticity smaller than -1 in absolute terms (i.e., between -1 and 0). Most academics working on this would consider an estimate of -0.2, as in your example, as very large in magnitude (i.e, very price elastic) because they actually are thinking of numbers closer to -0.05 or -0.02, which is what traditional regression estimates have suggested. My point was that the correct number is likely to be closer to -0.25 on average over our sample and hence further from 0. For many people this represents a move in the direction of more elastic demand, compared to what they expected. I mention this merely for clarification.
(6) I am not a big fan of models of transitory versus permanent effects of oil price shocks. For one thing that distinction is helpful only if the real price of oil is exogenous and we know it is not. Moreover, the actual oil price dynamics are somewhere in the grey zone between these extremes. Finally, there is strong evidence from consumer survey data on gasoline price expectations that people in practice virtually always form expectations of the real price according to a random walk. There is some nice recent work by my colleague at Michigan, Ryan Kellogg (and his coauthors) on this point. This suggests that the stories we economists like to tell about how people respond to oil price fluctuations don’t quite fit. People tend to view any price increase a pretty much the same and permanent, which is not unreasonable given that the random walk is a pretty tough benchmark to beat in out-of-sample forecasts.
@Ray,
I don’t want to get deeply into the other issues here, but is “27 miles” of tankers end to end really that much?
Doing a rough, back of envalope calculation – 27 miles is approximately 43,452 meters. The average VLCC tanker is about 330 meters, so 27 miles = 132 VLCCs at approx 250,000 tons = approx. 33 million tons in tanker storage. 33 million tons is about 1.25% of the 2.7 billion tons shipped by tanker each year (avg 2007-2009). Even if we took really big ULCCs, it would be about 105 ULCCs at 450,000 tons (though I don’t think there are 105 ULCC that big), it would only give us 47.3 million tons, or about 2%.
These are really, really rough numbers, to be sure, but I don’t think 27 miles of tankers is really that much.
Lutz–
Thanks for putting your rebuttal up independently on your page. Makes it very easy to explore the points and then your responses directly to those.
I think the authors probably did want to echo some popular media nonsense in order to gain that reputation or attention. It seems to me they added ad-hoc assumptions to try and tease out some kind of speculative effects from data like yours which previously suggested little to zero effects.
Thank you, in general for taking the time to address the hot-button issue of oil speculation, and calling out the critical errors that get made up and repeated in the lame-stream media.