Irrational Escalation in Investment Decisions: The Data

The sunk cost fallacy doesn't explain why intelligent investors throw good money after bad—something far more systematic is happening in the decision architecture itself.

When a portfolio underperforms, the rational response is straightforward: reassess the thesis, compare it against alternatives, and reallocate if warranted. Yet institutional investors routinely increase positions in failing investments, and the behaviour intensifies when losses mount. This isn't weakness or stupidity. It's a predictable output of how commitment and identity become entangled with capital allocation.

The mechanism works like this: an initial investment creates what researchers call "commitment escalation." The decision to invest isn't merely financial—it becomes a statement about judgment. The investor has publicly backed a thesis, perhaps defended it in committee, built a narrative around it. When early returns disappoint, the investor faces a choice that feels binary: admit error or increase conviction. But the choice isn't really binary. It's filtered through reputation, through the need to appear consistent, through the psychological weight of having been wrong.

What the data reveals is that escalation correlates most strongly not with the quality of new information, but with the magnitude of prior loss. Investors who have lost 15% on a position are statistically more likely to add to it than investors who have lost 5%, even when the underlying fundamentals are identical. This isn't because the larger loss provides better information. It's because the larger loss creates greater pressure to recover—and recovery requires either time (which feels passive) or capital (which feels active and therefore more within one's control).

The pattern holds across asset classes. In venture capital, follow-on investments in struggling portfolio companies increase when the initial loss is largest. In real estate, developers are more likely to inject additional capital into stalled projects after significant write-downs. In equities, retail investors add to underwater positions more aggressively than to positions that are merely flat. The behaviour is consistent enough that it appears in the data as a reliable signal—not of opportunity, but of psychological distress.

What makes this particularly consequential is that escalation decisions are made under conditions of maximum uncertainty. By definition, you're deciding whether to commit more capital to something that has already disappointed you. The information environment is typically poor. Yet the decision-making confidence often increases. Investors describe "averaging down" as a disciplined strategy, which it can be—but the language itself reveals the psychological mechanism at work. The frame shifts from "this investment is failing" to "this investment is cheap," and the shift happens most dramatically when losses are largest.

The institutional context amplifies this. An investment committee reviewing a struggling position faces collective pressure. The original decision-maker has reputational capital at stake. Colleagues may have supported the thesis. Reversing course requires not just admitting error, but doing so publicly, in a room, on record. Adding capital, by contrast, can be framed as conviction, as seeing through short-term volatility, as having a longer time horizon than the market. The narrative is available and socially acceptable.

The data also shows that escalation decisions are rarely revisited with the same rigour as initial investment decisions. Once you've committed additional capital, the position becomes harder to exit—not because the fundamentals have improved, but because you've now doubled down on the original error. The psychological investment deepens. Selling becomes an admission that both decisions were wrong, rather than just the first one.

What separates sophisticated investors from others isn't immunity to escalation. It's the presence of pre-commitment devices—rules that constrain the ability to add to losing positions without explicit, separate approval; processes that treat follow-on capital as a distinct decision requiring fresh analysis; cultures that reward the ability to reverse course quickly rather than the ability to defend prior decisions.

The pattern is robust enough that it should inform how investment governance is structured. The question isn't whether escalation bias exists. The data confirms it does. The question is whether your decision-making architecture accounts for it.