By Stephen Snudden, JDI Student Fellow, Queen’s University

Accurate forecasts of the price of crude oil are of central interest for investors and public policy institutions. For example, an oil producer deciding to invest in drilling or an airline company deciding to purchase a new fleet of aircraft cares about the payoff over the lifetime of the investment. Hence, the expectations and realizations of crude oil price movements influence the economy and respective policy stance. However, generating accurate and robust crude oil price forecasts is notoriously difficult. The International Monetary Fund produces crude oil price forecasts for up to five years, but historically these forecasts have not been very accurate. Even crude oil future prices contain very little predictive power beyond a few months (Alquist et al., 2013). That is why some policy institutions, including the Bank of Canada, rely on the current crude oil price, the no-change forecast, as the official forecast. This leaves no room for policy to react to future crude oil price movements.

The existing oil price forecasting literature has provided evidence that individual model forecasts of the real price of oil can outperform the no-change forecast at short horizons (Alquist et al., 2013; Baumeister and Kilian, 2011; Baumeister and Kilian, 2014; Baumeister and Kilian, 2015). In particular, these studies show that univariate models can outperform the no-change forecast for up to six months, and multivariate models can forecast for a little beyond a year. None of these studies intended to forecast at horizons beyond two years. Moreover, these methods are not always robust over time. Individual model forecasts perform well during certain times, but they underperform the no-change forecast at other times. Individual monthly models of the price of crude oil are not able to out-perform the no-change forecast the real price of crude oil at horizons between one and five years.

My paper, “Targeted Growth Rates for Long-Horizon Forecasts with an Application to the Real Price of Crude Oil,” seeks to fill this gap. It proposes a statistical method to forecast at longer horizons. The analysis begins by considering simple benchmark models for forecasting the real price of crude oil for horizons of up to five years. The average real oil price over the last year is found to provide a simple alternative forecast of the real price of oil at horizons beyond one year and works particularly well at the two- and three-year horizons.

This paper then proposes the method of targeted growth rate filtering, a modification to the standard model construction method. The method uses insights from spectral analysis to show that the lag in growth rate transformations can be used to target lower frequencies of the data. The method removes high frequencies and emphasizes select low frequencies which correspond to respective forecast horizons. This contrasts with traditional data transformation methods which use period-over-period growth rates, a form of a high-pass filter that overemphasizes high frequencies in the data.

Targeted growth rate transformations are then applied to various univariate and multivariate models to forecast the real price of crude oil. The method significantly improves mean-squared forecast errors relative to the no-change forecast and compared to models that rely on period-over-period growth rates. The method also improves directional accuracy at horizons of up to five years. The method can achieve the same degree of accuracy at horizons up to five years that has only previously been achieved at shorter horizons. For specific models, the method exhibits robust crude oil price forecasts over time.

There is little evidence on the success of alternative data series for use in out-of-sample forecasts for multivariate forecast models of the real price of crude oil, especially at longer horizons. Hence, the paper systematically explores the forecast performance of alternative measures of global real activity and crude oil inventories. Targeted growth rates applied to world industrial production and Kilian’s global real activity index (Kilian, 2009) are found to produce comparable forecasts at horizons up to five years. Moreover, U.S. petroleum inventories are found to produce superior forecasts at these horizons, compared to the often used measures in the literature.

The results suggest that the method of targeted growth rate transformations can improve the accuracy and robustness of crude oil price forecast compared to traditional methods. Model generated forecasts of crude oil prices using targeted growth rates can out-perform the no-change forecast at horizons between one and five years. Further extensions of the method should be explored such as extending the analysis to satisfy real-time data constraints (Baumeister and Kilian, 2011). Moreover, potential gains from forecast pooling techniques should be considered (see, for example, Baumeister et al., 2014). This paper suggests that institutions can exploit reliable crude oil price forecasts at horizons of up to five years to inform policy.

Alquist, R., Kilian, L. (2010), “What do we Learn from the Price of Crude Oil Futures?” Journal of Applied Econometrics, 25(4), 539–573.

Kilian, L. (2009), “Not All Oil Price Shocks Are Alike: Disentangling Demand and Supply Shocks in the Crude Oil Market,” American Economic Review, 99(3), 1053-69.

Baumeister, C., Kilian, L. (2011), “Real-Time Forecasts of the Real Price of Oil,” Journal of Business and Economic Statistics, 30(2), 326–336.

Baumeister, C., and  Kilian, L. (2014), “What Central Bankers Need to Know About Forecasting Oil Prices,” International Economic Review, 55(3), 869–889.

Baumeister, C., Kilian, L. (2015), “Forecasting the Real Price of Oil in a Changing World: A Forecast Combination Approach,” Journal of Business and Economic Statistics, 33(3), 338–351.

Baumeister, C., Kilian, L., Lee, T. (2014), “Are There Gains from Pooling Real-Time Oil Price Forecasts?” Energy Economics, 46(S1), S33–S43.

Snudden, S. (2016) “Targeted Growth Rates for Long-Horizon Forecasts with an Application to the Real Price of Crude Oil”, Mimeo.