How Easy Is It to Forecast Commodity Prices?
Over the last decade, unprecedented spikes and drops in commodity prices have been a recurrent source of concern to both policymakers and the general public. Given all the recent attention, have economists and analysts made any progress in their ability to predict movements in commodity prices? In this post, we find there is no easy answer. We consider different strategies to forecast near-term commodity price inflation, but find that no particular approach is systematically more accurate and robust. Additionally, the results warn against interpreting current forecasts of commodity prices upswings as reliable and dependable signals of future inflationary pressure.
In our NBER conference volume paper, “Commodity Prices, Commodity Currencies, and Global Economic Developments,” we do not attempt to answer questions such as “why are commodity prices so persistently high or low?” or “when will they start affecting inflation expectations?” Our purpose is more modest in that we assess how the information from a large dataset of indicators of global conditions may help predict future movements in commodity prices. Our forecast variables are cross-commodity price indexes, that is, we consider ten indexes taken from four distinct sources going back as far as 1973. Before we discuss our results, it may be useful to summarize briefly the approaches usually adopted to forecast commodity prices.
Alternative forecasting approaches
The first approach argues that economic modeling is not particularly helpful. The best information one can use to predict future prices is what is already embedded in current and past prices. Thus, to obtain the best out-of-sample prediction one could just estimate a simple statistical autoregressive process (what goes up must come down) or a random walk specification (the best forecast of tomorrow’s prices are today’s prices). A similar emphasis on simple statistical processes has proven remarkably powerful in the analysis of several asset classes, from stock prices to exchange rates. Let’s call this the “atheist” approach, as only historical information embedded in a commodity price is used for forecasting.
At the opposite extreme is the claim that commodity prices would be easily predictable if only one used the right tools, the right theory, the right model. Let’s call this the “true believer” approach. Examples of this approach are:
- Predictions based on observed or expected movements in fundamentals, such as strong demand growth confronting stagnating world production and little spare capacity due to past sluggish investment. Because commodity markets are characterized by relatively inelastic demand, even small revisions in the expected path of future supply expansion can have large and highly volatile effects on prices.
- Forecasts focused on speculative behavior in the futures markets. The claim here is that speculative strategies that drive commodity futures prices up must be reflected in higher spot prices today regardless of long-term fundamentals; otherwise, agents would have an incentive to accumulate inventories that could be sold later at higher prices. Such forward-looking behaviors are likely to be stronger in an environment of rapid declines in short-term interest rates, when the opportunity cost of physical commodity holding is relatively low, and investors in money-market instruments seek higher yields in alternative asset classes such as commodity derivatives.
- Emphasis by some market observers that exchange rate fluctuations of relatively small and predominantly commodity-exporting economies such as Australia, Chile, or South Africa are privileged predictors of future global commodity prices. Primary commodity products represent significant components of output in the above-mentioned countries, affecting a large fraction of their export earnings. At the same time, these countries are too small to have much of an impact on world markets. This observation implies that global commodity price changes represent external shocks for these countries, and their exchange rates move today in anticipation of future terms of trade adjustment.
Finally, some are unwilling or unable to choose among the glut of “true beliefs.” Let’s call this approach “agnostic” (whatever works, we'll take it). Pragmatically, this approach translates into throwing into the cauldron of possible predictors disparate things such as macroeconomic time series across major developed and developing countries (for example, industrial production, business and consumer confidence data, retail sales volumes, money aggregates, and interest rates), data on inventories and production of industrial metals and energy commodities, and data on ocean shipping costs across many different routes. To distill the relevant information from such a brew of raw data, we follow the approach in our FRBNY working paper “Revisiting Useful Approaches to Data-Rich Macroeconomic Forecasting” and use partial least squares (PLS) regressions which, in essence, take those linear combinations of predictors in our large dataset that have maximum explanatory power for future commodity price changes. The main advantage of this approach is that it tends to outperform empirically the more traditional principal components-based methodologies—where predictor combinations would be constructed in isolation (of future commodity price changes or any other variable).
To summarize, we obtain different sets of forecasts based on different “atheist,” “true-believer,” or “agnostic” approaches.
And the winner is . . .
Well, there is no obvious winner. Information from large panels of global economic variables can help, but their forecasting properties are by no means overwhelming. It all depends on the choice of the specific index and the forecasting horizon. For example, for one specific commodity price index, PLS regressions provide significantly better predictions than both autoregressive and random walk benchmarks when used to forecast one-month and one-quarter-ahead commodity prices. But when the forecasting horizon is six months or longer, the forecast performance of PLS regressions is no better than the statistical benchmarks. PLS does perform relatively better with aggregate commodity price indexes than with commodity subindexes such as metals or energy.
If we focus on specific subsets of explanatory variables—as emphasized by the “true believers”—we do find some, but not overwhelming, evidence for the notion that commodity currencies are useful predictors. We find even less empirical support for the notion that commodity futures have strong predictive power.
Ultimately, the basic message is one of inconclusiveness. No easy generalization or pattern emerges, and the results look almost random. In fact, we are unable to generate forecasts that are, on average, more accurate and robust than those based on autoregressive or random walk specifications. If a policy lesson can be drawn from our results, it is that one should be very cautious when interpreting the forecast of a forthcoming commodity price surge as an early signal of recrudescence in global headline inflation. As forecasts of commodity prices provide only highly noisy hints about their actual future trajectories and persistence, excessive confidence in such forecasts may bias policymakers' views and beliefs about future inflation risks in the direction of a premature—and unwarranted—tightening of the global policy mix.
The views expressed in this blog are those of the author(s) and do not necessarily reflect the position of the Federal Reserve Bank of New York or the Federal Reserve System. Any errors or omissions are the responsibility of the author(s).