The problem with IRR

The Internal Rate of Return (IRR) is beloved by private equity (PE) and venture capital (VC) firms, often regarded as the ultimate measure of an investment’s attractiveness. Its simplicity and comparability make it a favored metric to evaluate projects across industries. However, in the world of biotech—dominated by uncertainty, fat-tailed risks, and non-linear outcomes—IRR is not just insufficient; it can be dangerously misleading.

Why IRR Is a Mirage in Biotech

Biotech investments operate in an environment where rare, extreme events—both positive and negative—define outcomes. Drugs with transformative impact (e.g., Humira or Keytruda) generate outsized returns, while most other projects fail entirely. This results in fat-tailed cash flow distributions that IRR cannot capture. Let’s dissect why this matters.

The Formula and Its Simplistic Assumptions

Mathematically, IRR is defined by solving the equation:

$$
\sum_{t=0}^n \frac{CF_t}{(1 + r)^t} = 0
$$

Where:

• CFt​: Cash flow at time ttt,
• n: Total number of periods,
• r: IRR.

This equation assumes deterministic cash flows, static time horizons, and uniform discounting. These assumptions fail catastrophically in biotech, where cash flows are stochastic, development cycles are uncertain, and outcomes are highly variable.

1. Biotech’s Long Timelines and IRR’s Myopia

Biotech investments often span 10–15 years, with significant uncertainty at every stage of the process. For example:

• Drug discovery and preclinical testing: 3–6 years.
• Clinical trials (Phases I, II, III): 6–8 years.
• Regulatory approval and commercialization: 2–3 years.

The time between initial investment and meaningful revenue generation can exceed a decade. IRR heavily discounts cash flows occurring far into the future, often rendering long-term tail events—such as blockbuster drug success—nearly irrelevant.

Discounting Long-Term Success

IRR applies an exponential discount factor:

$$
PV = \frac{CF_t}{(1 + r)^t}
$$

For large ttt, even substantial future cash flows are discounted to negligible values. Consider a hypothetical blockbuster drug generating $1 billion annually starting 15 years from now. With an IRR of 20%, the present value of this revenue is:

$$
PV = \frac{1,000,000,000}{(1 + 0.2)^{15}} \approx 66,460,000
$$

This severely undervalues transformative long-term opportunities, which dominate biotech portfolios. Conversely, the metric overvalues near-term milestones that contribute relatively little to the overall value of the investment.

Short-Termism and Fund Cycles

PE and VC funds operate on fixed lifecycles, typically 7–10 years. IRR’s sensitivity to time incentivizes a focus on quick exits, even at the expense of maximizing value. In biotech:

1. Rushed development: Investors may push for faster clinical trials to meet IRR targets, increasing the risk of failure.
2. Premature exits: Promising assets are sold early, forfeiting the outsized returns of long-term tail events.

This short-term bias is particularly harmful in biotech, where the greatest value often materializes after long, uncertain timelines.

2. Fat-Tailed Risks and IRR’s Blind Spots

Biotech cash flows follow a fat-tailed distribution, characterized by rare, extreme events that dominate overall outcomes. These events can be modeled using a power-law distribution:

$$
P(X > x) \sim x^{-\alpha}, \quad \alpha > 1
$$

Where:

• Right tail: Rare successes like blockbuster drugs (e.g., Humira, Keytruda) generate billions in revenue.
• Left tail: Failures, such as clinical trial setbacks or regulatory rejections, can wipe out entire investments.

IRR smooths over these extremes by focusing on average cash flows, which are not representative in fat-tailed domains. For example:

• A single blockbuster drug might contribute 90% of a portfolio’s returns, but IRR averages this success across all investments, obscuring its impact.
• Conversely, IRR fails to account for the catastrophic downside of rare but severe risks, leading to over-optimistic projections.

3. IRR’s Fragility to Non-Linear Cash Flows

Biotech cash flows are often lumpy and non-linear, driven by milestone payments, licensing deals, and royalty streams. These irregularities create two problems for IRR:

Multiple IRRs

The polynomial nature of the IRR equation means that cash flows with alternating signs (e.g., negative R&D costs followed by positive revenues) can produce multiple solutions. This ambiguity makes the metric unreliable.

Misrepresentation of Optionality

Biotech investments often contain embedded options—such as pivoting a failed drug to a new indication or expanding a platform technology. These options introduce non-linear payoffs that IRR cannot capture. For example:

• A gene-editing platform might appear unattractive based on early cash flows, but the optionality to target multiple diseases creates a convex payoff structure.

4. Discount Rate Instability in Fat-Tailed Contexts

IRR implicitly relies on a uniform discount rate, but in fat-tailed contexts, the appropriate rate varies depending on the magnitude and timing of cash flows. In biotech:

• High-risk early stages: Require higher discount rates due to uncertainty.
• Post-approval revenue streams: Should use lower discount rates reflecting reduced risk.

In fat-tailed environments where variance may be infinite (α<2), standard discounting approaches break down entirely.

5. IRR’s Mathematical and Statistical Failures

From a mathematical perspective, IRR is ill-equipped to handle biotech’s realities:

1. Infinite variance: In fat-tailed distributions, variance may not converge. IRR assumes finite variance, making its projections meaningless.
2. Central limit fallacy: IRR relies on the central limit theorem, which assumes aggregated cash flows converge to a normal distribution. In biotech, extreme outliers dominate, invalidating this assumption.

6. The Convexity of Long-Term Biotech Investments

Biotech investments are inherently convex: Small inputs of capital can yield massive, asymmetric returns. Convexity thrives in uncertainty, as optionality and scalability amplify the impact of rare successes. IRR, however, penalizes convexity:

1. Short-term penalties: Early-stage R&D costs lower IRR, even if they pave the way for outsized long-term payoffs.
2. Linear assumptions: IRR treats cash flows as additive, ignoring the exponential growth potential of platform technologies or follow-on indications.

Conclusion

IRR’s reliance on deterministic cash flows, short-term bias, and inability to handle fat-tailed risks make it fundamentally incompatible with biotech. By trivializing long time horizons and extreme outcomes, IRR misrepresents both the risks and rewards of drug development. For investors in this space, clinging to IRR is not just suboptimal—it is an act of fragility in the face of uncertainty. To succeed, investors must adopt metrics that embrace the stochastic, fat-tailed nature of biotech, valuing patience, convexity, and the transformative power of rare successes.