Portfoliothese (Issue 1): Has Bitcoin failed as a safe haven?

Even among professional researchers, there is no agreement on what protection Bitcoin can actually provide.

In April 2025, JPMorgan analysts led by Nikolaos Panigirtzoglou questioned Bitcoin’s role as digital gold. Almost a year later, the same team reported that Bitcoin had shown signs of safe-haven demand during the Iran war, while gold and silver suffered outflows.

Other studies reach a markedly more sceptical assessment.

The question therefore needs to be framed more narrowly at first: did Bitcoin offer protection when equity markets came under stress? That can be tested. So we test it first, before discussing whether this is the right yardstick.

What is being tested

In 2010, Baur and Lucey formulated a distinction that is still widely used today. They separate three properties that are regularly confused: A diversifier is positively, but not perfectly, correlated with equities on average. A hedge is uncorrelated or negatively correlated on average. A safe haven is uncorrelated or moves in the opposite direction precisely when equities lose heavily.

An asset can therefore diversify over the long term and still fail as a safe haven.

What follows is not a formal test of conditional correlation but a practical one: did Bitcoin preserve capital when the equity market destroyed it? This is exactly the question the following test is meant to answer.

The data comprise ten years of daily closing prices from August 2016 to July 2026 for the S&P 500, gold and Bitcoin. Weekly returns are derived from them for the correlation and the weakest equity weeks. The stress periods are identified on a rule basis from the S&P 500’s daily closing levels: every decline of at least 10% from the previous high, measured from equity peak to equity trough and over the same window for all three assets. A new event only begins once the earlier high has been regained. All calculations were carried out in US dollars and in euros.

The findings

First, a frequently cited metric. Over the full period, the correlation of weekly returns with the S&P 500 is 0.19 for Bitcoin and 0.18 for gold in US dollars, and 0.16 and 0.08 in euros. The values are not identical, but all of them point only to a weak positive correlation.

That is already a finding: by this metric, both assets are only weakly positively correlated with equities on average. It says nothing yet about the safe-haven question. Anyone who cites it as evidence of diversification properties is describing average behaviour, not behaviour under stress.

It gets interesting where protection is needed. In the weakest 10% of equity weeks, i.e. 53 weeks with an average loss of about 4.1% in the S&P 500:

Gold was unchanged on average and finished in positive territory in the majority of cases. Bitcoin lost 3.4% on average in US dollars and 2.4% in euros and was positive in only about one-third to two-fifths of the weeks. A control calculation using the weakest 20% does not change the qualitative result.

The period saw five completed equity market declines of 10% or more:

Bitcoin fell in all five windows: in four of five cases by more than the equity index, and in three of them considerably more. Gold never lost more than 7.1% and gained twice.

In the practical test against equity market stress, Bitcoin did not pass. Gold fulfilled this function considerably better. In euros the same holds: Bitcoin likewise fell in all five windows and in four of five by more than the index; gold never lost more than 3.3% there.

The finding does not stand alone. DIW Berlin, using a different data frequency and methodology, arrives at a similar qualitative conclusion: over the period examined, Bitcoin was not suited, as gold is, to diversifying an equity-heavy portfolio. In “Gold and Bitcoin”, Campbell Harvey considers equating Bitcoin with digital gold too simplistic and sees Bitcoin more as a complement to gold than as a substitute for it.

This analysis does not confirm those studies; methods and questions differ too much for that. It does show, however, that the qualitative conclusion does not stand in isolation.

Two qualifications belong here. In March 2020, Bitcoin fell 32.3%, marginally less than the S&P 500 at 33.9%. That does not qualify as protection, but “always falls by more” would be wrong. And gold was not unfailing protection: it lost in three of the five episodes, though far less often and substantially less than Bitcoin.

The correlation is also time-dependent: between 2016 and 2019 it was close to zero for Bitcoin, and since 2020 it has been markedly higher. That does not change the stress finding: already in the two declines of 2018, Bitcoin lost 25.5% and 37.8%. A low average correlation is no proof of a safe-haven property.

What the findings do not say

Bitcoin did not pass the test, and that does not change. It is worth looking at the yardstick itself, however.

The concept of a safe haven is older than this financial-statistical operationalisation. Historically, gold’s special role rested not only on its behaviour relative to equities but also on monetary and institutional properties: scarcity, the possibility of physical possession, and the absence of a claim against an issuer.

The practical test used here asks about one dimension of safety: capital preservation in equity market corrections. It does not capture what protects against monetary dilution, counterparty default or seizure of assets. Those are different risks and, for some mandates, possibly the more relevant ones.

This is not a retrospective redefinition but the observation that the test is narrow. What has been refuted is protection against equity market stress, not every conceivable portfolio function.

This can be shown with the starting point. The JPMorgan assessment of March 2026 rests on fund inflows, institutional positioning and a markedly increased use of crypto in Iran, including transfers into self-custody. It therefore describes a different risk from the one tested here. Both statements can be true side by side.

There are structural differences from gold in individual properties. With direct control of the private keys, Bitcoin can be transferred without physical transport. Possession and use of gold have at times been legally restricted historically, for example in the US from 1933. It does not follow that Bitcoin is in principle beyond state access; actual resilience depends, among other things, on key control, identifiability and access to the technical infrastructure.

Anyone investing through a regulated investment vehicle generally does not control the private keys at the investor level. The extent to which the relevant properties of the underlying asset nonetheless remain effective for the investor depends on the legal structure, the legal ownership of the assets, insolvency protection, the custody chain and any delivery or redemption rights. That is a separate question and will be the subject of a later issue.

My conclusion

Anyone who allocates to Bitcoin as protection against equity market declines relies on a rationale that these data do not support. Clients should be told this plainly, even if it shortens the conversation.

Anyone who explicitly defines the defensive allocation as protection against equity market declines cannot empirically justify such a protective effect for Bitcoin with these data. For mandates that hold Bitcoin for other reasons, the history examined argues for modelling simultaneous losses in equities and Bitcoin in stress calculations. No compensating effect can be derived from it.

What it depends on

The empirical finding applies to the data basis chosen, the S&P 500 as reference and the period examined. Whether it holds with the same strength against another equity market or for other types of crisis was not tested. Within the chosen scope, its direction is robust: in both currencies, in both the weakest 10% and the weakest 20% of equity weeks, and in all five event windows. The magnitude does vary, however: in euros, Bitcoin falls by roughly one percentage point less in weak weeks.

Protection against discretionary expansion of the money supply or institutional dependence is a different thesis. This analysis neither confirms nor refutes it, and it depends on the relevant properties actually remaining effective at the investor level. For a regulated vehicle, this must be assessed on the basis of the legal structure, the legal ownership of the assets, insolvency protection, the custody chain and any delivery or redemption rights.

Definition according to Baur/Lucey (2010), “Is Gold a Hedge or a Safe Haven? An Analysis of Stocks, Bonds and Gold”, Financial Review.

On the historical restriction of gold ownership in the US from 1933: Federal Reserve History, “Roosevelt’s Gold Program”.

Sources for the introduction: JPMorgan Research, analyst team led by Nikolaos Panigirtzoglou, April 2025 (assessment of the digital-gold narrative), and report of 26 March 2026, as reported by The Block on 26 March 2026 (safe-haven demand during the Iran war).

Further studies with a similar qualitative conclusion: DIW Berlin, Wochenbericht 9/2025, “Bitcoin ist nicht das neue Gold” (“Bitcoin is not the new gold”; monthly returns); Campbell R. Harvey, “Gold and Bitcoin”, SSRN Working Paper No. 5530719, October 2025.

Data basis: daily closing prices for the S&P 500, Bitcoin and gold; own calculations. The S&P 500 is the price index excluding dividends. Raw data from 1 August 2016 to 4 August 2026; the last complete weekly observation ends on 31 July 2026. Weekly returns are based on the last jointly available trading day of a calendar week, regularly Friday (Thursday in 17 of 522 weeks, mostly because of public holidays); simple returns, Pearson correlation; 521 weekly returns. The event windows are based on daily closing levels.

The euro series were derived from the same data; the conversion used EUR/USD. The euro calculation uses the same trading days; US stock-market holidays on which the index level is carried forward and only the exchange rate moves were excluded. The event boundaries remain unchanged from the USD definition so that the same window applies to all three assets.

The rule was applied without retrospective adjustment: the 2023 decline remained part of the 2022 drawdown that had not yet been recovered; the largest decline up to the data cut-off of 4 August 2026 was 9.1%, below the threshold.

The three series follow different trading hours; the daily closing prices are therefore not synchronised to the same point in time. This may shift weekly returns and the values at the edges of the event windows; no separate quantification of this effect was carried out.

Boris Jurczyk is the founder of 21Nexus Capital and is therefore economically connected to the topic of digital assets. Previously he was responsible for equity and quant selection at Berenberg and for liquid investments at AURETAS family trust.

This article is for information purposes only and constitutes neither investment advice nor a recommendation to buy or sell financial instruments or crypto assets. Crypto assets are subject to substantial price fluctuations; a total loss of the capital invested is possible. Past performance is not a reliable indicator of future results.