Download Article

During periods of financial stress, investors tend to shift toward safer assets. An asset is considered a safe haven if it is uncorrelated or negatively correlated with riskier assets during times of stress (Baur and Lucey 2010). Although U.S. government securities are often considered the quintessential safe haven against the stock market, previous studies have suggested that gold, too, has acted as a safe haven in the past.

In recent years, investors have questioned whether Bitcoin might function as a safe haven as well. In theory, Bitcoin’s fixed supply and algorithm-based issuance allow it to be independent from traditional markets, which could make it a desirable asset in times of economic stress. Indeed, in a survey, many holders of Bitcoin stated they believed it to be similar to gold as opposed to stocks, U.S. dollars, or a “new technology” (Hundtofte, Lee, Martin, and Orchinik 2019). Consequently, the returns of other safe-haven assets may have changed after Bitcoin’s introduction.

To assess how the 10-year Treasury note and gold behaved in periods of stress before and after the introduction of Bitcoin, we examine the correlations of their daily returns with the daily returns of the S&P 500, a diversified portfolio of large U.S. publicly traded companies that is commonly used as a proxy for risk asset performance. We examine these correlations from January 1995 through February 2020 to allow for a near-equivalent number of periods before and after the introduction of Bitcoin. In light of recent events, we also separately analyze correlations in March 2020, when the bull market ended amid concerns about the coronavirus pandemic. We separate the sample into periods in which financial stress was either above or below its long-term average using the Kansas City Financial Stress Index (KCFSI). Although assets are only considered safe havens in times of stress, we include periods without stress to capture any changes in asset behavior between the periods. Chart 1 shows that the KCFSI and S&P 500 index generally have an inverse relationship: over most of our sample, the KCFSI rises as the S&P 500 falls. The gray bars identify our periods of financial stress, which begin when the KCFSI (green line) rises above zero.

Chart 1: Periods of Financial Stress

Chart 1 shows how periods of financial stress from 1995 to 2020 are determined using the S&P 500 index and the Kansas City Financial Stress Index (KCFSI). Periods of stress are delineated when the KCFSI rises above zero. The KCFSI and S&P 500 index generally have an inverse relationship: the KCFSI rises as the S&P 500 falls.

Note: Gray bars indicate periods of stress.

Sources: Federal Reserve Bank of Kansas City and Bloomberg.

Table 1 shows the historic correlations of the 10-year Treasury, gold, and Bitcoin across the entire sample. Both the 10-year Treasury and gold have negative, statistically significant correlations with the S&P 500, suggesting both assets have the properties of safe havens. In contrast, Bitcoin has a weak positive correlation with the S&P 500 during periods of financial stress, suggesting Bitcoin behaves more like a risk asset than a safe haven.

Table 1: Historic Correlations with the S&P 500, 1995–2020

Table 1 shows historic correlations for the 10-year Treasury, gold, and Bitcoin with the S&P 500 from 1995 to 2020. In periods without financial stress, all three assets show negative correlations with the S&P 500, though only the 10-year Treasury and gold are statistically signifcant. During times of financial stress, both the 10-year Treasury and gold still have negative correlations with the S&P 500, while Bitcoin has a weak positive correlation with the S&P 500, and all three are statistically significant.

Source: Bloomberg.

Table 2 breaks the full sample down into individual periods with or without financial stress and shows that the safe-haven properties of assets have varied over time. Panel A shows the correlations for gold and the 10-year Treasury before the introduction of Bitcoin, while Panel B shows the correlations after Bitcoin was introduced. As in Table 1, the 10-year Treasury consistently behaves like a safe haven: the correlations for the 10-year Treasury are negative and statistically significant across all stress periods. In contrast, gold has only behaved like a safe haven in certain periods of financial stress. Finally, Panel B shows that Bitcoin has failed to exhibit the behaviors of a safe-haven asset—its correlations with the S&P 500 are not statistically significant in any period and positive in all but one.

Table 2: Historic Correlations with the S&P 500 in Varying Periods of Stress over Time

Table 2 breaks the 1995 to 2020 period into periods with and without financial stress to show how assets perform over time. Panel A shows the correlation for the 10-year Treasury and gold before the introduction of Bitcoin. Panel B shows the correlations after Bitcoin was introduced. The 10-year Treasury exhibits consistent statistically significant negative correlations across all stress periods, thus behaving like a safe haven. Gold shows statistically significant negative correlations during certain periods of financial stress, thus sometimes behaving like a safe haven. Bitcoin exhibits positive correlations with the S&P 500 during all but one period of financial stress, and its correlations are not statistically significant in any period, thus showing no safe-haven behavior.

Sources: Federal Reserve Bank of Kansas City and Bloomberg.

A natural question is whether the recent period of financial stress related to the coronavirus pandemic has altered these correlations. Table 3 shows that during March 2020, none of the assets exhibited statistically significant safe-haven behavior. The result for the 10-year Treasury is somewhat surprising, as it was a safe haven for every stress period from 1995 through February 2020, regardless of duration. Although its correlation with the S&P 500 in March 2020 is still negative, it is not statistically significant. Gold has a positive but insignificant correlation, which is unsurprising given that it has not consistently demonstrated safe-haven properties across stress periods. As in the full sample, Bitcoin has a positive, statistically significant correlation, suggesting it performed like a risk asset in March rather than a safe haven.

Table 3: Correlations with the S&P 500 in March 2020

Table 3 shows that in March 2020, the 10-year Treasury, gold, and Bitcoin all failed to exhibit statistically significant safe-haven behavior as measured by their correlation with the S&P 500 index. Although the 10-year Treasury still has a negative correlation with the S&P 500, it is not statistically significant. Gold has a weak positive correlation though this is not statistically significant. Bitcoin shows a moderate positive correlation, suggesting it performed like a risk asset rather than safe haven.

Source: Bloomberg.

Overall, our results suggest that the 10-year Treasury has generally exhibited safe-haven behavior, gold has occasionally exhibited safe-haven behavior, and Bitcoin has never exhibited safe-haven behavior since its introduction. Moreover, the introduction of Bitcoin does not appear to have materially changed the safe-haven properties of government bonds or gold. Instead, Bitcoin at times appears to have behaved more like a risk asset than a safe haven.

Endnotes

  1. 1

    He, Krishnamurthy, and Milbradt (2016) build a model to explain why U.S. government bonds exhibit this property. Coudert and Raymond (2011) find that gold qualifies as a safe haven.

  2. 2

    Academics such as Baur, Dimpfl, and Kuck (2018) and Smales (2019) have asked this question, too.

  3. 3

    To include Bitcoin in the sample, we do not use periods of extreme stress. As a robustness check, we also run the analysis using periods where stock markets fell more than 15 percent, 17.5 percent, and 20 percent and find nearly identical results to the main analysis using the KCFSI stress periods.

  4. 4

    Using a t-test, we find the difference between the higher and lower stress periods to be statistically significant.

  5. 5

    Although uncorrelated assets may be classified as safe-haven assets, none of the near-zero correlations are statistically significant.

  6. 6

    The correlation is still negative even though it is not statistically significant. The lack of significance could be due to the volatility within this short sample period.

  7. 7

    This result could also be due to the volatility within this short sample period.

References

Jesse Leigh Maniff is a payments specialist at the Federal Reserve Bank of Kansas City. Sabrina Minhas is a former assistant economist at the bank. David Rodziewicz is a senior commodity specialist at the bank. Becca Ruiz is a research associate at the bank. The views expressed are those of the authors and do not necessarily reflect the positions of the Federal Reserve Bank of Kansas City or the Federal Reserve System.

Author

David Rodziewicz

Regional Research Senior Economics Specialist

David Rodziewicz is a senior economics specialist at the Denver Branch of the Federal Reserve Bank of Kansas City. His research focuses on energy economics, natural resource econ…