Skip to content

Drawdowns (Van Hemert et al.): summary and key takeaways

The paper simulates how often a strategy hits a given maximum drawdown, then tests rules that replace or de-risk managers after drawdowns. Total-return rules work best when manager quality is fixed. Drawdown rules catch managers who lose skill, but cutting risk alone lowers expected return.

Hemert, O.V.; Ganz, M.; Harvey, C.R. et al. · 2020 · The Journal of Portfolio ManagementSummary updated 20 November 2026
Published2020Journal article
Asset focusAsset-agnosticDrawdowns
TypePortfolio MgmtPortfolio Construction & Risk
Reading time4 minthis summary

Drawdowns (Van Hemert et al.) at a glance

The question
How likely is a given drawdown for a strategy of known quality, and how useful are drawdown rules for replacing or de-risking managers?
Dataset & period
Simulated normal, independent monthly returns; baseline of a 10-year window, 10% annualized volatility and 0.5 annualized Sharpe ratio, with 100,000 simulations per setting. A bootstrap uses actual US equity returns since 1926. Appendix lists worst monthly moves for equity indices over 25 to 50 years.
Method
The authors simulate monthly returns and measure how often the maximum drawdown reaches 1 to 4 annual standard deviations, varying volatility, horizon, Sharpe ratio, autocorrelation and gap risk. They then model a pool of Good managers (Sharpe 0.5) and Bad managers (Sharpe zero). They compare replacement rules based on total return, drawdown and maximum drawdown, using one test or monthly checks. A final test cuts risk by half after a drawdown and restores it after half the loss is recovered.

Key findings of Drawdowns (Van Hemert et al.)

  • In the baseline case, the chance of reaching a drawdown of 1, 2, 3 and 4 annual standard deviations is 97.1%, 43.0%, 9.9% and 1.5%.
  • Holding the Sharpe ratio fixed, changing volatility barely moves the probability of a drawdown measured in sigma units. Longer windows raise it. A higher Sharpe ratio lowers it by a large amount.
  • An autocorrelation of 0.1 in monthly returns changes the expected maximum drawdown about as much as cutting the Sharpe ratio from 0.5 to 0.4.
  • Bootstrapping actual US equity returns lifts the probability of a 2-sigma drawdown from 43.0% to 55%, because of non-normal returns and volatility clustering.
  • When manager quality is fixed, total return separates Good from Bad managers better than drawdown rules. When Good managers can turn Bad, at 0.5% a month, drawdown rules become about as effective, and with monthly checks they do better.
  • A rule that cuts risk by 50% after a drawdown can raise the Sharpe ratio slightly but lowers annualized return.

Robustness, as reported in the paper

  • The bootstrap with actual US equity returns gives similar sensitivity to window length and Sharpe ratio as the normal simulation.
  • The paper reports that rerunning the bootstrap with longer blocks gave similar results.
  • A time-varying drawdown cutoff that grows with the square root of time gives a higher average Sharpe ratio than a constant cutoff for the same number of replacements; the gain fades when replacements are frequent.

Caveats and limits

  • Results rely on a stylized two-type manager model (Good and Bad) and simulated returns. Real manager skill is less clean (inferred).
  • Analysis uses monthly data only; daily drawdown decisions are left to future work.
  • The paper does not study how a drawdown rule changes manager behavior.
  • Replacement costs are discussed in words, but the main figures plot Sharpe ratio against the number of replacements rather than modeling a cost (inferred).

Items marked (inferred) are our reading of the design, not statements by the authors.

Why it matters

Allocators and risk managers can see how often a good strategy will still breach a drawdown limit by chance. The results show why a fixed limit gets hit more often as time passes. They also show that a drawdown rule and a total-return test answer different questions.

Questions about this paper

Does a good strategy still hit large drawdowns?

Yes. With a Sharpe ratio of 0.5 over a 10-year window, the paper finds a drawdown of two annual standard deviations is reached in almost half of simulations.

Is a drawdown rule better than a total-return rule for replacing managers?

Only when managers can lose skill over time. If quality is fixed, total return is the better statistic.

Does cutting risk after a drawdown raise returns?

In the paper's test it lowered annualized return unless risk is added elsewhere. Gains need negative expected returns after losses, such as strong autocorrelation.

Where can I read the original paper?

The original is at its publisher: https://doi.org/10.3905/jpm.2020.1.170. We link to the paper and do not host the PDF.

Follow new quant research

Download MacroFinance market data as CSV or Excel, or pull it by API, and test the ideas in these papers yourself. New summaries are added as we review them.

Create a free account
Summaries are written by MacroFinance from the published papers, in our own words, for information only. They are not investment advice and may omit detail. Papers are not hosted here: each page links to the original at its publisher.