An empirical paper suggests that the risk premium and excess return on gold have been time-varying and predictable, also out-of-sample. The key predictors have been the variance risk premium and the jump risk premium of gold. Gold has historically also served as a hedge and “safe haven” for equity and bond investments, but this could not have been expected based on forecasting models. Common sense suggests that the hedge value of gold depends on the dominant market shock. For example, gold hedges against inflationary policies but not against rising real interest rates.
Nguyen, Duc Binh Benno, Marcel Prokopczuk and Chardin Wese Simen (2017), “The Risk Premium of Gold”, November 20, 2017
The post ties in with SRSV’s lecture on strategies based on implicit subsidies (excessive risk premia), particularly the part about using information of volatility markets.
The below are excerpts from the paper. Emphasis and cursive text have been added.
Predicting the return on gold
“We first show that the excess return of gold is time-varying and predictable both in-sample and out-of-sample using a parsimonious forecasting model.”
“Our primary data set consists of end-of-the-day futures for gold traded on the New York Mercantile Exchange/New York Commodities Exchange…Our analysis covers the period from January 1996 until February 2015.”
“Our empirical analysis shows that… the best model for the gold risk premium includes the left [downside] jump risk premium of gold and the gold variance risk premium as explanatory variables….The jump tail premium and the variance risk premium are strong predictors for the gold risk premium, with high explanatory power both in-sample and out-of-sample and for all horizons investigated, varying from one month to two years…The monthly variance risk premium is defined as the difference between the implied volatility and the [expected] realized variance. [The jump risk premium is a compensation paid for protection related to extreme rare or tail events]… The left and right tail measures are estimated in a two-step procedure where the tails are extrapolated from the short maturity and deep out-of-the-money options.”
“There is a growing literature investigating the predictive power of the risk premium using…the variance risk premium, which proxies the aggregate degree of risk aversion in the market…Recent studies address the ability of rare disaster events to explain the risk premium…The jump risk premium…has been shown to amount for a large fraction (two–thirds) of the equity premium.”
See also related post on downside variance risk premium here.
“Our prediction model shows good out-of-sample forecasting performance across all horizons, where it is able to outperform the historical mean model. The higher performance relative to the historical mean is statistically significant for five of the six horizons.”
Gold as a hedge or “safe haven” asset
“The media often claim that gold is a hedge and safe haven asset and the recent literature has empirically tested this claim… these studies use realized returns and compute covariances or other dependence measures… and only answer the question whether gold and other assets co-moved ex-post. However…from an asset-pricing perspective, it is much more important to understand, whether gold is also expected to be a hedge or safe haven asset.”
“We differentiate between hedges and safe havens…Gold is a hedge for another asset if it is uncorrelated or negatively correlated in general, while it serves as a safe haven asset for another asset when it is uncorrelated or negatively correlated in times of market stress.”
“We examine the co-movements between the risk premia of gold and other important markets. We investigate the hedging and safe haven properties of gold by examining their expected and unexpected relationship. We find for the stock and bond markets that gold is generally not expected to be a hedge, but it is realized as such ex-post. Also, gold is not expected to act as a safe haven asset, but it does [over the sample period].”
“We…investigate the expected relationship of gold and equity relying on linear regression models. Our equity risk premium model relies on the two most predominant predictors: the dividend yield and the variance risk premium… One can observe that gold is not expected to serve as a hedge or safe haven throughout all horizons…We find that the expectation of gold as a hedge and safe haven ex-ante differs from its actual role ex-post not only by magnitude but also by sign depending on the horizons.”
“The results are similar for bonds…We find that gold is not expected to serve as a hedge and safe haven, but it does serve as both ex-post.”