How does the KMV model improve on Merton, and what is the Expected Default Frequency?

The KMV model (now owned by Moody's Analytics) takes the Merton framework and makes it practically useful by addressing two of Merton's biggest weaknesses.

Improvement 1: Better Default Point

Merton uses total debt at maturity. KMV empirically found that firms typically default when assets reach:

Default Point = Short-Term Debt + 0.5 x Long-Term Debt

This makes intuitive sense: short-term debt must be repaid immediately, but long-term debt gives the firm breathing room. The 0.5 multiplier was derived from observing thousands of actual defaults.

Improvement 2: Empirical EDF Mapping

Merton converts DD to default probability using the normal distribution: P(default) = N(-DD). But empirical evidence shows this severely underestimates actual defaults. For example:

The KMV approach: Instead of using the theoretical normal distribution, KMV built a massive empirical database of firms, their DDs, and whether they actually defaulted within 1 year. The mapping from DD to EDF is derived from this historical data.

The KMV Process:

Worked Example — Ironbridge Corp:

Ironbridge Corp has:

• Market cap (E): $400M
• Equity volatility (sigma_E): 40%
• Short-term debt: $300M
• Long-term debt: $200M
• Default point: $300M + 0.5 x $200M = $400M

After solving the simultaneous equations:

• Asset value (V): $780M
• Asset volatility (sigma_V): 22%
• Expected asset return (mu): 10%

DD = [ln(780/400) + (0.10 - 0.5 x 0.0484) x 1] / (0.22 x 1)

= [0.6678 + 0.0758] / 0.22

= 0.7436 / 0.22

= 3.38

Merton: N(-3.38) = 0.04% default probability

KMV EDF: approximately 0.28% (from empirical mapping)

The KMV EDF is 7x higher than Merton's theoretical estimate — and historically much more accurate.

Strengths of KMV:

• Updates daily with equity prices (forward-looking)
• Empirically validated across thousands of defaults
• Captures firm-specific risk through equity volatility

Weaknesses:

• Still assumes lognormal asset dynamics
• Proprietary database not publicly available
• Less reliable for private firms (no equity price)

Practice KMV model calculations in our FRM Part II question bank.