|
Matthias Knab, Opalesque for New Managers: Investors from 34 countries joined the recent Opalesque webinar "Profiting from Panic: Portfolio Strategy for Volatile and Failing Markets" with Daniel Gelernter, Director of AI and Machine Learning at R.G. Niederhoffer Capital Management and Chief Investment Officer of the Niederhoffer Family Office.
The session - delivered, as it happened, on Gelernter's birthday - laid out a provocative thesis: the diversification frameworks most family offices, allocators and CIOs still rely on were built for a market that no longer exists. Gelernter calls the result "Diversification In Name Only" (DINO) - portfolios that look diversified on paper but behave like a single concentrated bet when stress hits.
The math has stopped working the way investors think it does
Modern portfolio theory has not changed much since Markowitz published it in 1952. The market, however, has changed enormously. Gelernter walked the audience through updated correlation research showing that the average pairwise daily-return correlation across US equities has nearly tripled - from 5.7% in the 1962-1997 period (Fang, Jiang, Sun, Yin and Zheng) to 13.4% from 1998 through 2020, and 15.9% in the 2021 to April 2026 window.
More importantly, that correlation is not static. It is roughly 90% correlated with market volatility itself - meaning correlations spike precisely when investors most need diversification to work. The highest single quarter on record was Q2 2022 at 27.7%. The highest single month came in April 2025 at 36.8% - more than six times the pre-1998 average, and concentrated in exactly the moment investors most depended on diversification.
The flip side of that picture is the share of risk that diversification can actually remove. Using 1 minus average pairwise correlation as a proxy, the diversifiable share of equity risk has collapsed from 94.3% in the 1962-1997 regime to 84.1% in the current period - and to just 63.2% in the April 2025 stress month. The market in which Markowitz's framework was designed to operate has, for practical purposes, disappeared.
A NASCAR detour that explains everything
To illustrate, Gelernter shared an unexpected analogy from a weekend AI project of his own: a quant model for betting on NASCAR, connected through Robinhood and sized using a Kelly criterion. Across 101 NASCAR Cup Series races, the model captured the eventual winner inside its top-5 ranked drivers 67.7% of the time on short tracks - where individual driver skill and idiosyncratic risk dominate. On superspeedways, where cars run nose-to-tail at 250 mph and a single mistake can take out 15 cars at once, that figure collapsed to 33.3%. The top-5 miss rate roughly doubled, from 32.3% on short tracks to 66.7% on superspeedways. And this happens because there's much more systemic risk (with heavy accidents) on that type of race.
His point: today's markets are no longer a short track. They are a superspeedway. Systematic risk dominates, idiosyncratic risk shrinks, and adding more names to a portfolio mostly sacrifices edge without buying meaningful protection.
Bull markets are mostly just recoveries from the last crash
Looking at major US drawdowns over the last 150 years in real (inflation-adjusted) terms, Gelernter showed that "bull markets" are in many cases just the long recovery from the previous crash. Inflation does much of the damage. The Great Depression (Sep 1929) took roughly 29 years to recover in real terms. The 1968 inflation malaise added almost 20 years to nominal recovery. The 1907 Panic added another 19. The oil shock of 1973 added seven, the dotcom bubble seven and a half. Excluding 1958-1968, there is essentially no full decade in the 20th century that was not spent either in or recovering from a crisis.
The uncomfortable conclusion: zero correlation is not enough
The core takeaway is one most investors find counterintuitive. An uncorrelated strategy is not protection. Only a negatively correlated strategy is. Uncorrelated strategies neither help nor hurt during a crash - and crashes, Gelernter argued, are precisely the events that determine long-term terminal wealth.
The math on protective allocations is also more aggressive than most CIOs assume. Using a two-asset mean-variance framework and a protective sleeve with only half the Sharpe of the risk book:
- A zero-correlated strategy warrants an optimal allocation of 33.3%
- A -0.25 correlated strategy warrants 40%
- A -0.50 correlated strategy warrants 44.4%
The impact on recovery time is dramatic. Applying a half-Sharpe, -0.50 correlated sleeve at its optimal weight, the Great Depression's 29.2-year real recovery compresses to 5.9 years. The inflation malaise drops from 23.1 years to 3.2 years. The Panic of 1907, from 22.0 years to 2.7 years. The dotcom bubble, from 14.2 years to 6.1 years. The 2008 financial crisis, from 6.1 years to 3.3 years.
And at the moment the passive index merely claws back to zero, the protected portfolio is already substantially ahead in real terms - 97% ahead in the Great Depression case, 46% ahead in the inflation malaise, 43% ahead in the 1907 Panic. Those surpluses then compound for the rest of an investor's life.
Allocation size is where most portfolios fail
The most striking finding is that token allocations barely move the needle. With the same -0.50 correlated, half-Sharpe sleeve applied to the Great Depression:
- 0% allocation (passive): 29.2 years to recover
- 5% allocation: 26.0 years
- 15% allocation: 25.2 years
- 44.4% optimal allocation: 5.9 years
The same pattern holds across every major crisis. Recovery time only collapses when the hedge is large enough - and the gap between "what most people hold" and "what the math says" is the single biggest unaddressed risk Gelernter sees in private wealth today.
Properties of a real protective strategy
Asked what investors should look for in such a strategy, Gelernter pointed to three properties: liquidity (so it is actually available to rebalance with in a crisis), correlation (the more negative, the more effective - though strategies more negative than -0.5 are very hard to find, partly because many were dropped from portfolios during the CTA winter of 2010-2019 and did not survive), and volatility, which is counterintuitively a feature rather than a bug for a negatively correlated sleeve. The more volatile a negatively correlated strategy is, the more it reduces overall portfolio risk.
Watch the replay
This is a summary of a much richer one-hour conversation that included live audience polls, the underlying mean-variance derivation, detailed walk-throughs of recovery times and terminal wealth across ten major crises, and a discussion of trend-following pitfalls in violent reversals such as the February-March 2023 Silicon Valley Bank episode.
Watch the full replay here: https://www.opalesque.com/webinar/index.php?id=94#pw94
A note for fund managers
Dan's audience at this live session was made up of Individual High Net-Worth Investors, Single Family Offices, Wealth Managers/RIAs, Banks, Multi Family Offices, Fund of Funds, Private Pension Funds, Endowments, Public Pension Funds, Foundations and Sovereign Wealth Funds.
And with that, I congratulate Dan, because he is doing something really smart here. Using one hour of his time, he is talking to hundreds of investors.
If you are a fund manager or portfolio manager and could benefit from presenting what you do to a similar audience at one of our future webinars, please email me at knab@opalesque.com.
|
RSS
