Is the Age of Scarcity upon us?
If we could forecast the scarcity of such natural resources as oil, then we could remove much of the uncertainty in Central Asian economies that frightens investors. But analysts disagree over the extent of scarcity, largely because they differ over how to measure it.
Perhaps the most famous study of scarcity, by Harold Barnett and Chandler Morse, examined the cost of extracting resources in the United States from 1870 (the end of the Civil War era) to 1957. They treated rising extraction costs as signals of growing scarcity. In most extractive industries, unit costs fell, particularly after 1890. In fact, they fell even faster than those for non-extractive industries. The exception was the forestry.
Follow-up work found that extraction costs fell even more steeply from 1957 to 1970 – and that they kept falling in the 1970s for ferro alloys and nonferrous metals, noted a resource economist, Jeffrey Krautkraemer. However, extraction costs for coal and oil rose in the United States throughout the Seventies. Whether this was due to scarcity or to OPEC is a matter of conjecture.
Barnett and Morse explained that when extractors exhausted high-grade resources in most industries, they would discover low-grade resources in even greater abundance. Also, the rise in the price of a resource that has become scarce for a while would induce searches for new deposits and cheaper substitutes. Innovation has sharply cut extraction cost “even as the quality of exploited deposits has declined,” observed Krautkraemer.
Keep these data under your hat
Writing in 1963, Barnett and Morse didn’t consider the cost of energy in extraction. In 1991, Cutler Cleveland found that the costs of labor and manmade inputs (capital) in extraction had fallen because extractors were replacing these with fossil fuels. Were fuels becoming more scarce, then extraction costs in general should rise (which is exactly what some statistical studies of the Seventies didn’t find). Data on extraction cost are hard to collect, because producers regard them as confidential. And as a measure of scarcity, extraction cost is flawed because it looks backward; it does not directly reflect expectations of scarcity.
A measure that does reflect expectations is the cost of exploring for another unit of resources. An expected increase in profits down the road will fuel exploration and thus push up its cost at the margin.
Unless we find cheap new ways to extract resources, or to substitute for expensive ones, resource prices will eventually rise because of scarcity. The prices may follow a U-shaped curve over time, falling and then rising. In 1982, Margaret Slade reported evidence of this curve in the prices of 11 of 12 metals and fuels studied over the period from 1870 to 1978. Other studies confirm that the prices of exhaustible resources don’t always rise or fall. However, not all of such prices kept rising after the 1970s, which is what a U-curve price path might suggest.
It’s about time
Most statistical analyses of resource scarcity use data that change over time, called “time series.” An example is the annual price of oil from, say, 1970 to 2010. The methods used to examine time series have changed dramatically over the past 30 years. Old methods assumed that the basic traits of a time series did not change over time. For example, the price of oil would have the same basic average in any year, although random events – such as Mideast wars -- may cause the observed price to differ from this fundamental average in a particular year. Today, we know that most economic time series do change in basic ways over time. For example, the value of production in the United States – gross domestic product – has usually risen for two centuries. It would be difficult to argue that GDP has the same fundamental average now as it had in 1812. GDP is “nonstationary.” For such a time series, the usual statistical model is not accurate, because it assumes that given factors, such as the number of workers, will affect GDP in the same way over time when in fact the response of output to labor is changing. Unfortunately, this was the approach taken by many statistical studies of resource prices in the early Eighties.
One way to handle nonstationarity is to put the time series into a form with essential characteristics that do not change over time. Such a form is “stationary.” Although GDP may be nonstationary, the annual change in it may well be stationary. Another approach is to estimate a GDP model that explicitly controls for a time trend. In either case, once the time series is rendered stationary, we can apply the usual statistical techniques to it, since we no longer need to worry that the estimated parameters, which are assumed constant over time, may mislead us. We can then reverse-engineer the model in order to get forecasts of the original variable – say, in the level of GDP rather than its annual change. (The Notes offer an example.)
In 1996, Peter Berck and Mike Roberts took this approach in estimating the price paths of natural resources like oil. From an extension of Slade’s dataset, they used the annual price changes for their stationary time series. They found “only a weak supposition that natural resource prices will rise….We would predict rising prices but be much less surprised about being wrong than were the previous authors.” Price increases were most likely for zinc and copper. The Age of Scarcity remains a strong possibility, but it is not entirely clear that it already engulfs us. – Leon Taylor, tayloralmaty@gmail.com
Notes
Suppose that we estimate the following model for the price of oil in year t: P(t) = 2 P(t-1). According to this model, whenever the price of oil rises by one dollar in the previous year (t-1), it will rise by two dollars in the current year (t). This model may work fine for the time period over which it was estimated – say, from 1970 to 1990. But if the time series for oil prices is nonstationary, then the model may not fit other time periods. Using it to forecast the 2013 price of oil would be futile.
Suppose, then, that we estimate a model for the annual change in oil prices. Denote this change as D(t) = P(t) – P(t-1). Suppose that our new model is D(t) = .5 D(t-1). Also suppose that D(t) is stationary. Then the new model would fit time periods in general, and we could use it to forecast the 2013 price of oil. To do this, note that we can write the new model as P(t) – P(t-1) = .5 D(t-1). Rearrange this: P(t) = P(t-1) + .5 D(t-1). Specifying the values of these variables will give us the forecast: P(2013) = P(2012) + .5 D(2012). For example, if P(2012) is $100 and D(2012) is $10, then the forecast for 2013 is $105.
Good reading
J. A. Krautkraemer. Nonrenewable resource scarcity. Journal of Economic Literature 36(4). 1998. Pages 2065-2017.
References
Harold Barnett and Chandler Morse. Scarcity and growth: The economics of natural resource availability. Baltimore: Resources for the Future. 1963.
Peter Berck and Michael Roberts. Natural resource prices: Will they ever turn up? Journal of Environmental Economics and Management 31. 1996. Pages 65-78. Online as Working Paper 699, California Agricultural Experiment Station, Giannini Foundation of Agricultural Economics.
Cutler J. Cleveland. Natural resource scarcity and economic growth revisited: Economic and biophysical perspectives. In Robert Costanza, editor, Ecological economics: The science and management of sustainability. New York: Columbia University Press. 1991.
Margaret E. Slade. Trends in natural resource commodity prices: An analysis of the time domain. Journal of Environmental Economics and Management 9. 1982. Pages 122-137.
