Using various real-time forecasting strategies, estimation windows, transaction costs, and forecaster loss functions, the authors analyze the link between international business fluctuations (measured by G–7 output gaps) and gold price movements from 1975 to 2012. The evidence suggests that the international business cycle has limited predictive power for gold price fluctuations. These findings differ from those of other researchers who show that the international business cycle has some predictive ability for the stock market.
What’s Inside?
The authors use the real-time forecasting approach developed by Pesaran and Timmermann (Journal of Finance 1995, Economic Journal 2000) to see whether the international business cycle, as measured by estimated output gaps and output growth rates of the G–7 countries, can forecast gold price fluctuations. This forecasting approach accounts for model uncertainty and model instability by using real-time available information and a search-and-updating technique to help dynamically identify the best forecasting model. The authors also examine two different techniques to update the forecast model (recursive and rolling estimation windows), and three different periods (5 years, 10 years, and 15 years) of back history are used as the initial model training periods.
How Is This Research Useful to Practitioners?
The international output gaps weakly forecast gold prices, according to the authors. In light of these results, they recommend that although the output gap model has been reported to have predictive power for the stock market, it should not be applied to the gold market without further study. The gold market appears to be reasonably efficient with respect to international business cycle fluctuations, but in terms of (risk-adjusted) excess returns, this finding does not preclude the possibility that other financial and macroeconomic variables have systematic predictive power.
How Did the Authors Conduct This Research?
The authors use end-of-month US dollar gold price data from the Federal Reserve Bank of St. Louis for the January 1975–December 2012 period. They compute continuously compounded excess returns by subtracting the one-month short-term interest rate from the change in the natural logarithm of the gold price. Short-term interest rates for the period are obtained from the Federal Reserve Bank of St. Louis.
Using industrial production data for the G–7 countries (Canada, France, Germany, Italy, Japan, the United Kingdom, and the United States) from the Organisation for Economic Co-Operation and Development, the authors estimate the time series of each country’s output gap with Cooper and Priestley’s (Review of Financial Studies 2009) simple filtering technique, which does not require specification of a smoothing parameter and is not subject to endpoint bias.
The authors use two alternative modeling strategies to determine an optimal forecast each month. First, they consider four forecast-selection strategies: the adjusted coefficient of determination criterion, the Akaike information criterion, the Schwarz information criterion, and the direction of change criterion. Second, they consider four forecast-aggregation strategies: a simple averaging criterion, a mean-based averaging criterion, a weighted-mean averaging criterion, and a truncated-mean combination criterion.
They also examine a simple trading rule that builds on real-time out-of-sample forecasts as well as a buy-and-hold strategy. They find that after accounting for modest transaction costs, the simple trading rule based on real-time out-of-sample forecasts does not lead to superior performance relative to the performance of the buy-and-hold trading strategy.
In terms of the accuracy of forecasts, the results can be summarized into three conclusions. First, the rolling-window real-time approach performs better than the recursive real-time forecasting approach. Second, an optimal forecast using a forecast-aggregation modeling strategy does not perform better than a forecast using a forecast-selection strategy. Third, rolling-window estimation updating often performs better than recursive updating when the goal is to outperform the buy-and-hold strategy. The rolling-window approach better predicts the upswing in the price of gold after 2000 and its continued increase to December 2012.
The focus of the authors’ research is the US-based investor; they recommend that other investor perspectives be considered. In addition, they acknowledge that other technical and tactical gold market-timing strategies may have more predictive power than their simple real-time forecasting and two trading strategies.
Abstractor’s Viewpoint
The authors leverage prior research to focus on the relationship between the international business cycle and gold price fluctuations—an approach few researchers have taken. Gold prices have increased since 2000 because of geopolitical turmoil, the global financial market crisis, and the subsequent European sovereign debt and banking crisis. This article is well organized and logical because the authors clearly present definitions and their data sources. It should be of interest to many investment practitioners.