Conditional Drawdown at Risk (CDaR)

The Conditional Drawdown at Risk (CDaR) is a risk measure that quantifies the average loss in the worst drawdown episodes above a chosen confidence level. Unlike maximum drawdown, which captures only the single deepest decline from peak to trough, CDaR spreads the risk across multiple bad scenarios, averaging the worst episodes. It is a “coherent” risk measure, meaning it respects diversification and scales appropriately with portfolio size—making it a sounder tool than maximum drawdown for portfolio construction and risk management.

What Is Drawdown and Why Measure It?

A drawdown is the decline in value from a previous peak. If a portfolio reaches $1 million on day 100, then falls to $900,000 by day 200, the drawdown is $100,000 or 10%. Investors care deeply about drawdown because it represents real losses from the high-water mark—the worst moment in a holding period to have bought the portfolio.

Traditional risk measures like volatility (standard deviation of returns) treat gains and losses symmetrically. A volatility-based model does not care whether a 20% swing is a gain or a loss. But investors hate losses far more than they enjoy equivalent gains—a behavioral bias called loss aversion. Drawdown metrics respect this asymmetry by measuring downside directly.

The oldest drawdown measure is maximum drawdown—the largest peak-to-trough decline over the period. A portfolio with a 50% maximum drawdown experienced at least one moment when it was 50% underwater. This metric is intuitive but has a critical flaw: it responds to a single outlier event. A portfolio might have dozens of small 5% drawdowns and one catastrophic 50% drop; the maximum drawdown ignores the pattern of pain and fixates on the worst day.

From Maximum Drawdown to CDaR

Conditional Drawdown at Risk fixes this by averaging over the worst drawdowns, not just the single worst. This move mirrors the evolution from a single risk measure to a more sophisticated one.

The maximum drawdown is equivalent to asking: “What is the deepest loss my portfolio ever hit?” The answer is a number, and it is often driven by one volatile day. In a leveraged or hedge fund context, that one day might be an outlier caused by a tail risk event, which investors correctly fear but which may be statistically rare.

CDaR answers: “On average, how deep are the worst drawdowns?” By averaging over multiple bad episodes, it captures the severity of repeated stress, not just the one unluckiest day.

Mathematically, CDaR at a confidence level (e.g., 95%) is the average drawdown among all historical drawdowns that exceed the 95th percentile. If historical data show 1,000 days of returns, CDaR-95 averages the drawdowns on the 50 worst days (top 5%).

Coherent Risk Measures

CDaR is a coherent risk measure, a term with a precise definition in quantitative finance. A coherent measure must satisfy four properties:

1. Monotonicity: If portfolio A always returns more than portfolio B, the risk of A is less than or equal to the risk of B.
2. Homogeneity: Doubling the portfolio doubles the risk. This prevents distortions when scaling strategies.
3. Translation invariance: Adding cash reduces risk proportionally.
4. Subadditivity: The risk of combined portfolios is at most the sum of individual risks. This ensures diversification never increases risk.

Maximum drawdown fails subadditivity. Two portfolios with 20% maximum drawdowns can combine into a portfolio with a 40% maximum drawdown if their drawdowns align perfectly. A coherent measure like CDaR would cap the combined risk at 40%, reflecting genuine diversification benefits.

Value at Risk (the maximum loss at a given confidence level over a set holding period) is also coherent. CDaR is sometimes called “conditional maximum drawdown” and is the drawdown analogue of VaR. Both are expected shortfall concepts—averages of the tail of a loss distribution.

CDaR in Practice

Portfolio optimization. Investors can optimize portfolio weights to minimize CDaR instead of volatility or Sharpe ratio. A CDaR-minimizing portfolio will hold assets that together avoid the worst drawdown episodes. This often leads to higher diversification and a different asset mix than volatility-based optimization.

Hedge fund and active strategy evaluation. A hedge fund manager claims a “15% volatility” but suffered a 50% maximum drawdown in a single year. The average investor cares more about the drawdown. CDaR-95 might reveal that the fund’s typical bad month (when it is in the worst 5% of months) sees an 8% drawdown. That is a more honest picture than the maximum.

Tail risk and black swan hedging. CDaR surfaces how expensive the worst episodes are, which guides decisions on tail risk hedges (e.g., put options). A portfolio with a CDaR-95 of 15% should expect serious drawdowns; this information can justify the cost of protective puts.

Comparative strategy assessment. Strategy A has a 30% maximum drawdown and a 12% CDaR-95. Strategy B has a 40% maximum drawdown and a 10% CDaR-95. The larger single drop in B is offset by its lower average of the worst episodes; CDaR reveals that B’s bad episodes are less severe overall, even if one bad month was worse.

Calculation and Interpretation

To compute CDaR-90 for a portfolio over 252 trading days:

1. Calculate the drawdown on each day (maximum loss from that day’s starting peak).
2. Rank drawdowns from largest to smallest.
3. Identify the top 10% worst days (roughly 25 days).
4. Average the drawdown values on those 25 days.

The result is CDaR-90 in dollars or percentage terms. A CDaR-90 of 12% means that, on the average of the 10% worst-case drawdown days, the portfolio loses 12% from peak.

A critical interpretation: CDaR-90 is not a worst-case bound (unlike maximum drawdown). It is a conditional expectation. The portfolio could still experience worse drawdowns in new data; CDaR quantifies the average severity of the historically worst episodes, not an absolute floor.

Strengths and Limitations

Strengths:

• Coherent, so it respects diversification and avoids scaling oddities.
• Sensitive to multiple bad events, not just one outlier.
• Aligns with investor loss aversion; focus is on downside, not upside.
• Yields meaningful comparisons across different time periods and strategies.

Limitations:

• Requires substantial historical data; CDaR-99 needs at least ~100 observations to be reliable.
• Backward-looking; past drawdowns may not predict future ones.
• Can be dominated by one or two extreme events in short data histories (same pitfall as maximum drawdown, though less acute).
• Less widely understood than volatility or Value at Risk, so institutional adoption lags.

Relationship to Other Risk Measures

Maximum drawdown is the simplest; it captures one worst day, making it easy to explain but prone to outlier distortion.

CDaR averages over multiple worst days, mitigating outlier effects and satisfying coherence axioms.

Value at Risk measures the tail of return distribution (e.g., 99% confidence that losses won’t exceed X). CDaR measures the tail of the drawdown distribution. Both are coherent and related; CVaR (Conditional VaR) is the return-space analogue of CDaR.

Volatility treats gains and losses equally. Drawdown-based metrics (including CDaR) are one-sided, capturing downside only. For investors with loss-averse preferences, drawdown metrics are more behaviorally salient.

See also