Bank Runs and Gold-Price Returns and Volatility: Evidence from Centuries of Data

Gold is commonly viewed as a safe haven asset to which investors rush in times of economic and financial crises. It is attractive during a bank run because it is not a li- ability of any party and does not involve counterparty or bank credit risk. Thus, bank runs as a type of financial crisis should significantly affect the conditional return and volatility of gold prices. Surprisingly, the implications of bank runs for gold price dynamics have not been studied. Our empirical study addresses this gap. To this end, and using annual data going back to 1694, we estimate variants of GARCH models that account for time-varying conditional volatility, as well as time-varying higher-order moments (conditional skewness and conditional kurtosis). We find that accounting for bank runs improves the in-sample performance of the estimated GARCH models and also significantly leverages their out-of-sample performance in terms of standard statistical and economic criteria. We, thereby, establish that taking bank runs into account, as well as the dynamics of conditional higher-order moments, matters for the empirical modeling of the dynamics of the gold price and, thus, for the the safe-haven property of gold. In addition, we show that the higher-order GARCH models extended to include bank runs help investors to improve the statistical and the economic value of out-of-sample forecasts of gold-price returns and volatility.