Energy Markets: What to do when forecasting is useless

Date May 17, 2017
Speaker Robert S. PINDYCK (Bank of Tokyo-Mitsubishi Professor of Economics and Finance, Sloan School of Management, Massachusetts Institute of Technology)
Moderator KOBAYASHI Keiichiro (Program Director & Faculty Fellow, RIETI / Professor, Faculty of Economics, Keio University / Research Director, Canon Institute for Global Studies)

Professor Pindyck will explain the evolution of energy markets and energy prices, and what we can expect in the future. He will begin with the price of oil and address the following questions: (1) What drives sharp price changes? (2) Can we forecast the price of oil? The role of "option value" in oil markets. Next he will turn to natural gas, coal, and nuclear energy, and explain the linkages to oil. The main conclusion is that forecasting prices is futile, and that the focus should be on assessing uncertainty and its implications for option value.



Robert S. PINDYCK's Photo


This talk is about energy in general, but mainly about oil prices. I will discuss oil and how other energy sources are connected to oil and its pricing. I will also talk about the past and what might happen in the future with energy in general and the price of oil. Then I will discuss the linkages of other forms of energy. I want to focus on sharp price changes and why they occur. This comes down to supply and demand elasticities: what happens when supply and demand shocks occur. Why don't we know what will happen to the price of oil? The key is what might happen and the extent of the uncertainty. Finally, I want to talk about what we can do in a world like this, in which no one can predict. The key is assessing the uncertainty and thinking about option value, because oil and coal reserves can be viewed as real options.

Recent oil price volatility

Over the last five years, the price per barrel of oil went from around $100 to close to $30 and has rebounded to about $48 today. These fluctuations surprised many people, but in 2005 and 2006, the price of oil increased over a period of a year or two from $70 to over $140, dropped to about $45, and rebounded to $80. Sharp changes have begun to appear normal. All the way back to about 1985, very sharp fluctuations have occurred. Why do we see these very sharp fluctuations? Does demand for oil change that quickly? Does the supply? Why is the price so volatile? Is it possible to forecast prices?

I am the co-author of a textbook on microeconomics which contains an example about the world oil market. The purpose of the example is to compare short-run and long-run supply and demand elasticities. The competitive oil supply, added to the amount the Organization of the Petroleum Exporting Countries (OPEC) produces (less than one-third of all world oil production today), gives the total supply. In the short-run—meaning about one year to 1.5 or two years—the demand curve is very steep, very inelastic, and the supply curve is also very steep and very inelastic. In the textbook example, suppose something happens in Saudi Arabia and there's a drop in its production of three billion barrels per year. This would cause the supply curve to shift, creating a new intersection at a very high price. Any shift in the supply curve or the demand curve in the short-run can cause a very big change in the price.

The long run (from 5-10 years) looks totally different. Supply and demand are much more elastic. In the long run, people drive smaller cars, put new insulation in their homes, etc. More sources of supply will be coming online over the long term; new facilities are developed and new sources of oil are discovered. The same shift in the curve over the long term causes a much smaller change in the price. The problem is that things change in the short run.

Recent oil production and consumption trends

What has happened in the oil market over the last three or four years? In 2016, world oil consumption was about 97 million barrels per day. From 2013 to 2016, production increased by about three million barrels per day, mostly from the United States. Oil production in the United States has increased dramatically, mostly due to shale oil. Russian production increased by about a half million barrels per day and Brazilian production increased as well. Iraq increased by about one million barrels per day. Smaller changes have taken place in other countries as well. Libyan production decreased by about two million barrels per day because of the ongoing civil war. If the fighting ends, Libya can produce about 2.5 million barrels per day. It has large oil and productive capacity. However, no one can predict what will happen in Libya, so the price of oil can't be predicted. If Libyan oil comes on the market, the price will drop dramatically.

Consumption, by contrast, has increased by about one million barrels per day. It hasn't increased more partly because world gasoline consumption has not risen very much due to improved automobile and airplane fuel efficiency. The result is that either inventories increase greatly or the price drops. Inventories around the world are at their highest levels ever, but it's not enough to prevent the price from dropping. The price dropped during this period simply because production increased greatly amid a modest increase in consumption. This is simple but hard to predict.

U.S. production increased by quite a bit over the past few years. Oil prices will continue to fluctuate over the next few years. Attempting to forecast the price of oil is fruitless. The right question to ask is what might happen to the price of oil. What's the extent and nature of the uncertainty? Why this is the right question comes down to option value. In the end, we want to know the value of our options.

Problems with forecasts

The International Energy Agency produced an approximately 20-year oil price forecast in 2012 or 2013 assuming various policy differences. For example, strong carbon dioxide emission policies would cause a price drop. It claimed these forecasts were useful, but it was wrong. Large oil companies predicted three cases, but they, too, were wrong. Natural gas price forecasts out to 2030 by an energy agency were all wrong. Yen-U.S. dollar exchange rate forecasts are wrong. Forecasts are produced because people are paid to produce them. This is called "technical analysis." This is done for the stock market as well. We look at tendencies and draw lines, but the result is essentially trying to forecast a random walk with gibberish explanations. They sell this and people buy it.

The futures market uses information from the Energy Information Administration (EIA) of the U.S. government. It produces a short-term outlook every month. It looks at the futures price and makes a forecast. The forecasted price is above the futures price because people mistakenly believe the futures price to be equal to the expected future oil spot price. In other words, the futures price a year from now would be the same as the expected spot price a year from now. This is incorrect; it's only correct if the commodity has no systemic risk. Under the capital asset pricing model, the beta for oil, which measures the systemic risk, is about one. This is because oil, like copper, is an industrial commodity, and when the economy is booming, demand goes up and the price goes up accordingly. That means that these commodities have systematic risks, i.e., risks correlated with the overall market or economy. Because of these systematic risks, the futures price is biased downwards as an estimator of the expected future spot price. A positive beta causes the expected future spot price to be above the futures price. At the 95% confidence interval—which means we can be 95% sure we are within a specific range—we can find useful information because this is a range of possibilities. The EIA looked at options prices. Futures contracts involve the trading of call options and put options. From the options prices, the implicit volatility of the spot price can be determined using the Black-Scholes formula, and from that, we can determine the confidence interval. This is what the EIA did. I focus on this because the confidence interval is very large. According to the options prices in the options market, an enormous amount of uncertainty exists. Even going out only one year, we are looking at a range of about $30 to $90. That's only the 90% confidence interval. We know that from the options market. This is useful. What you might ask is not so much what the price will be, but what is the confidence interval for the price of oil next year. This can be learned from the options market and the options on futures. The answer is that the interval is very large.

Effect of uncertainty on undeveloped oil reserve pricing

How is oil produced? The first thing to do is to find oil. We take a guess and drill and if we're lucky, we discover oil. That is called an undeveloped oil reserve. We can't produce the oil, but we know it's there. Next, we have to develop the reserve to produce the oil. That's very expensive. Maybe $1 billion, at least. Developing offshore oil reserves can cost from $3 billion-$5 billion. After the reserve is developed, oil can be produced for the next 10 years, maybe longer. The revenue earned every year depends on the price of oil, but the price fluctuates. What is that undeveloped oil reserve worth? And when should it be developed?

Does greater uncertainty raise or lower the value of the undeveloped oil reserve? There is world A, with very little uncertainty, in which the oil price doesn't change much, and world B, which has a lot of volatility. In which world is that undeveloped oil reserve more valuable? It's worth more in world B, even though it doesn't appear so, because of the addition of a risk premium. Uncertainty raises the value because the reserve is exactly like a call option. The more uncertainty the greater the option value. If the price goes down, the option is worthless. If it goes down by a lot, it's still worthless. If it goes up, it becomes worth something. There is a limit on the downside: it can only be worth zero, but on the upside, there is no limit, it can be worth anything. That's why it's important to know what the level of uncertainty is. Because companies that deal with oil need to determine the value of their undeveloped reserves, the key is understanding the nature of the uncertainty.

Coal and natural gas

Natural gas and coal prices are linked—but only loosely—to the price of oil. Oil is easy to transport, so there is a world market for it. Natural gas is not as easy to transport. That's why today, natural gas in the central United States is only $3 per 1,000 cubic feet (Mcf). In Boston, it's $12. We don't have pipelines because people don't want them in their backyards. There's a capacity problem. In Europe, it's about $10. In Japan, it was about $15, but it has fallen to about $7. The issue here is transportation. Liquefied natural gas is becoming more prevalent. That might be why the price in Japan has gone down. Coal is easier to transport, but it's dirty and regulated. Its use is declining because natural gas is cheaper and cleaner. It doesn't have a strong connection to the price of oil. Nuclear is simply not economical. To build a nuclear power plant is very expensive. The capital costs are huge. The regulatory environment is difficult and uncertain. That's true in Japan, United States, and Europe. Nobody predicted that Germany would turn around completely on nuclear. The result is that there has been very little or no construction of nuclear plants.


Q1. Two things are important in forecasting the future. One is trends. The other is game-changing events, like Libya. I agree with you on uncertainty, but if you look at trends, it's very important for forecasting the future. However, we cannot forecast the game-changing events. What do you think about that?

I agree that you can't forecast what will happen with game changers. There are many other places besides Libya. Brazil potentially has enormous offshore oil reserves, but Petrobras has a lot of problems, as does the Brazilian government. Maybe it will produce much more in the future. We don't know. Nigeria is another case. There is theft and disruption there. However, even the trends are unpredictable. Five or six years ago, no one predicted that the United States would produce so much shale oil and natural gas. Nobody predicted fracking. Over three years, production went from six to nine million barrels per day. That's the difficulty. The trends are often not there.

Q2. In a related question, from a policy perspective, given the great uncertainty, what should the government do? In research and development (R&D), for example, in the field of energy conservation technology or alternative energy development, it takes time to produce outcomes. Under great uncertainty, what should the government policy be?

Many people would say that the government should subsidize R&D. I disagree with that. I don't think the government has been very good at picking winners and losers. If there is belief that an externality exists, which there is in the case of oil (because of climate, security, impact on the economy of sharp fluctuations, etc.), then a tax is imposed. The market then should be allowed to operate and companies decide if they should develop new technology, invest in R&D, etc. The pharmaceutical industry is a good example. There's a lot of uncertainty over disease. Drug companies develop new drugs to fight disease. The uncertainty does not mean the government should subsidize drug companies to develop new drugs. It's the same thing.

Q3. Oil prices were controlled by OPEC for a long time, but it now provides only one-third of the supply. Should an international organization be established to stabilize oil prices?

There are two parts to your question. The first is the role of OPEC: did it affect oil prices? The second is whether the government should stabilize prices by stockpiling, etc. OPEC had very little influence on oil prices. Since its inception in 1962, OPEC has many members, but they really never agree. They have meetings and then all go and do what they want. OPEC has not had that great an impact since it started. Given this volatility, should the governments do anything? No, I don't think the governments should do anything. These are world markets. Similarly, the copper cartel has no effect on the price of copper. First of all, there needs to be some sort of world stabilization fund. Stockpiling is necessary, and knowing how much and when to release it is also necessary. It's impossible. If there is worry about fluctuations, then hedging should be done on the futures market. We have futures markets for this purpose. Governments should not come in and try to stabilize prices.

Q4. Aside from using option prices, how can we measure the degree of uncertainty?

Options prices are one good way of doing this, by looking at actual market transactions. The other way is to conduct simulations—"what-if" experiments. What if Libya does this or that, what if Brazil does this or that, etc. There are so many such experiments that can be done. Which are the important ones? Which are more likely? That's very hard to decide. Options are the best choice. If there are no options to look at to get price data, then scenarios have to be considered. Using oil as an example, by making a list of things that could happen to increase supply, a price of $20 or $140 can be imagined with all of the possible factors for the scenarios. The key is understanding how many different things can happen and how much uncertainty exists.

Q5. You mentioned copper. Currently, an enormous steel and aluminum oversupply exists globally. The government is trying to control this. U.S. President Donald Trump is introducing an investigation. What would your recommendation be to the U.S. administration regarding the oversupply issues?

With regard to the word "oversupply," a buyer of steel does not think an oversupply exists; it would think there's exactly the right amount. Whether there is a feeling that an oversupply or undersupply exists depends on whether the party is a buyer or a seller. I think there's an oversupply of economists. I am a big believer in free trade. We have the World Trade Organization (WTO) so we can have a playing field where we can trade freely and have rules. Dumping and other things can be a problem. If a problem exists, it might be appropriate to go to the International Trade Commission. That has to do with the rules for free trade and how to enforce them to continue trading freely.

Q6. If the options price is unavailable, is the data on past volatility important for predicting the confidence interval?

Good point. Yes, definitely. That's a very common way of looking at volatility and uncertainty. Over the last 30 years, the price has fluctuated enormously. Looking at the past and treating it as some type of stochastic process, volatility can be estimated.

Q7. If we consider a rare disaster-type shock that occurs only once every hundred years, does this estimation change?

Rare disasters are my favorite. This is a world in which people produce, buy, and trade oil. Life goes on. What would happen in the event of some huge catastrophe? Who knows?

Q8. Price is influenced by market expectations. How can we understand the role of market participants' expectations?

Are you asking whether expectation influences price? The answer is no. Many people have argued—and they only argue this when the price goes up—that, for example, heating oil price increases occur because speculators speculate on the futures market and push the price up. The futures market does not push the price up. The futures market is like a bet on a football game. I bet on one team and you bet on the other. If my team wins, you give me money, and if your team wins, I give you money. It has no effect on the game. Expectations don't affect price. Chinese demand increased because they were locking up supply at a much more rapid rate than what they needed. Some other countries were doing the same thing. Then, China said it didn't need so much oil and its consumption, etc. dropped. How could this be predicted?

Q9. My question is about the EIA forecast. This forecast is based on a variety of scenarios which take many variables and conditions into account. I assume it has simplified the variables. Does it have specific principles for the variables? Also, in scenario making, the key is how to use them. Could you discuss the extreme left and right of the scenarios, the range of possibilities? We could then decide what measures to take.

The forecast was useless. If a line is drawn, what does it imply? Nothing. I don't know what the EIA did and what it assumed, but it made a projection. The projection is useless, based on information it doesn't have. What it comes to scenarios, the key is that there are possibilities, and the EIA can't deal with this. It would be useful to say there are possibilities that could affect the price. If a party is thinking about getting into the energy industry and producing oil, and spend money, sure, it should look at scenarios and possibilities. It should know that the price could drop to $20 or go up to $140. The same is true for an oil consumer. It would want to know the possibilities so that it can take actions. It might hedge its risk on the futures market to reduce exposure.

*This summary was compiled by RIETI Editorial staff.