Survivorship Bias in Institutional Success Stories

There's a story from World War II that I keep coming back to. The U.S. military wanted to know where to add armor to their bombers. They examined the planes that returned from missions and mapped where the bullet holes clustered — the wings, the fuselage, the tail. The obvious conclusion: reinforce those areas. But a statistician named Abraham Wald pointed out the error. The planes they were examining had survived. The holes they were seeing were the holes that didn't kill a plane. The places that showed no damage — the cockpit, the engines — were where the planes that never came back had been hit.

We've been applying the same mistake to institutions for centuries.

When we want to understand what makes a great university, we study Harvard. When we want to understand democratic longevity, we study the United States. When we want to understand corporate culture, we study Apple or Southwest Airlines. And it all makes intuitive sense — find what works, study it, replicate it.

But what if Harvard's particular combination of practices isn't what makes it work? What if dozens of universities with the same endowment structures, the same selective admissions models, the same faculty governance systems simply ceased to exist — while Harvard, through some mixture of geography, timing, and political fortune, survived long enough to become the template everyone else copies?

That question doesn't have a clean answer. And the fact that it doesn't is precisely what makes survivorship bias so difficult to dislodge from how we talk about institutional success.

What Survivorship Bias Actually Does to Institutional Thinking

Survivorship bias isn't just a statistical quirk. In the context of institutions — universities, governments, religions, corporations — it operates as a kind of active distortion, because the institutions that fail tend to disappear from the record entirely. They close their doors, get absorbed, collapse under political pressure, or simply fade. Their archives scatter. Nobody writes business school case studies about the companies that had strong cultures and failed anyway.

The result is that every lesson we extract from studying successful institutions is drawn from a sample that doesn't represent the full range of what was tried.

I've come to think this creates a specific kind of intellectual trap. We observe features of surviving institutions — strong leadership, clear mission, adaptive governance — and conclude those features caused the survival. But those same features appear in institutions that failed. We just don't notice because we aren't studying the failures. The correlation between practices and outcomes is real, but we've truncated the dataset in a way that makes it look far cleaner than it is.

The clean version of institutional wisdom is almost always the survivorship version. And the survivorship version leaves out most of the story.

How Often Do Institutions Actually Fail?

The numbers are less comfortable than most institutional case studies suggest.

Of the 500 largest U.S. companies in 1955, only about 52 remained on the Fortune 500 list by 2020 — meaning roughly 90% of the firms that defined American corporate dominance in the mid-20th century are gone, merged, or dramatically diminished. That isn't a fringe phenomenon; that's the baseline.

The picture is similarly sobering for constitutional governments. Research by political scientists Tom Ginsburg, Zachary Elkins, and James Melton found that the average national constitution lasts only about 19 years before being replaced or substantially rewritten. The United States Constitution, at over 230 years old, is a genuine anomaly — not a template for what constitutions do, but an extreme outlier in a distribution most institutional scholars rarely surface.

Universities show the same pattern. More than 80 private nonprofit colleges closed between 2016 and 2022 alone, according to data from the National Center for Education Statistics. Those institutions had missions, faculties, traditions, and endowments. Many had the structural features that appear in every higher-education "best practices" guide.

And startups — which are in some ways just the fastest-moving version of institution-building — fail at a rate of around 90% within ten years, with approximately 20% not surviving their first year, according to the U.S. Bureau of Labor Statistics.

The graveyard is enormous. We just rarely visit it.

The Features We Attribute to Success Often Appear in Failures Too

This is the part that tends to make people uncomfortable, and I think it's worth sitting with.

When Jim Collins published Built to Last in 1994, he identified eighteen "visionary companies" — Motorola, Sony, Ford, Merck among them — and catalogued what distinguished them from their peers. The book became one of the best-selling management texts of all time, and its insights were applied across industries and sectors.

By the 2010s, many of those eighteen companies had stumbled badly. Motorola, once the paradigmatic example of core ideology combined with operational excellence, lost its mobile division to Google and became a cautionary tale. The problem wasn't that Collins was wrong about what those companies were doing. The problem is that the same things were being done by companies that never made it into the study at all — companies that had strong cultures, clear ideologies, and adaptive leadership, and still failed.

Selection bias shaped what got analyzed, and analysis of the selected cases produced principles that felt universal but were drawn from a skewed sample.

I think the same dynamic runs through almost every major body of institutional wisdom. The features we identify as causal — strong culture, adaptive leadership, clear mission, stakeholder trust — are real features of surviving institutions. But they're also features of failed institutions. What actually differentiates the two is a harder question than most frameworks want to acknowledge.

A Comparison That Rarely Gets Made

The right column isn't a list of institutions that lacked these features. Many of them had precisely these features, and failed anyway. What the comparison reveals is that the features themselves may be necessary but are rarely sufficient — and our habit of studying only the left column makes them look more sufficient than they are.

The Geography and Timing Problem

Here's something I find myself returning to: how much of institutional survival is about what an institution did versus when and where it existed?

The United States Constitution survived, in part, because it was ratified by a relatively small, educated, property-owning electorate in a period before mass communications, before industrialization created the economic pressures that toppled later democratic experiments, and in a geographic context that provided extraordinary natural protection from external threat. Those conditions had nothing to do with Madison's design choices.

Harvard survived because it was founded in a colonial context where proximity to Boston and political patronage created conditions simply unavailable to later-founded institutions. The practices Harvard developed are practices that could only be developed by an institution that had already survived. Replicating the practices doesn't replicate the conditions.

I'm not suggesting that what institutions do doesn't matter — it clearly does. But I think we systematically overweight practice and underweight circumstance, because practices are visible, learnable, and teachable, while circumstance is mostly luck and retrospectively obvious. The luck becomes invisible once the institution is established; the practices become the story.

This connects to a broader pattern I've written about on PatternThink — how institutions maintain their authority by curating the narrative of their own origins. Survivors don't just survive; they write the history of what survival requires.

What Does the Graveyard Actually Tell Us?

If we could study failed institutions with the same rigor we study successful ones, what would we find?

In my view, we'd find that the features most tightly correlated with survival aren't the inspiring ones. They're features like: access to capital during early periods of vulnerability, geographic or political protection from existential threats during formative years, and the presence of patrons or networks willing to absorb early losses. Those things don't make for compelling case studies, but they appear to be more robustly predictive than mission clarity or leadership quality.

We'd also likely find that some of the practices we attribute to successful institutions were developed after the critical period of survival — that they're the result of survival, not its cause. Institutions that clear the vulnerable early years have the luxury of developing strong cultures and adaptive governance. Institutions that don't survive never get the chance to develop those things at scale. So when we observe the features of mature, successful institutions, we may be observing what institutions look like after they've already won, and mistaking that for the reason they won.

That distinction matters enormously if you're trying to build something that lasts. The question isn't "what do successful institutions do?" It's "what helps institutions survive long enough to do the things that successful institutions do?" Those are different questions, and I think we've been answering the first while assuming it answers the second.

Why This Pattern Persists

It's worth asking why survivorship bias is so durable in institutional thinking, given that the statistical concept has been understood for decades.

Part of it is practical. Failed institutions don't produce the same volume of accessible data — case studies, memoirs, documented practices — as successful ones. If you want to teach a class on organizational design, you use the material that exists, and the material that exists is overwhelmingly produced by and about survivors.

Part of it is motivational. The lesson "context and timing matter more than strategy" is far less useful than "here are the practices that work." The former gives leaders nothing actionable; the latter does, even if the latter is partly illusory. There's a market for lessons drawn from survivors, and that market shapes what gets studied and published. This connects to a deeper question about how institutional patterns spread and calcify — the selection pressure runs through what we teach, not just what we build.

And part of it, I think, is that we are drawn to stories of survival for the same reason we find comfort in most stories of success. The survivors are here. They can speak. The failures have gone quiet.

The Honest Implication

If survivorship bias shapes our understanding of what makes institutions work, then most of the wisdom literature around institutional design is built on a truncated sample. That's an uncomfortable thing to say about fields that have produced genuine insight. But I think the discomfort is worth sitting with rather than explaining away.

The honest implication isn't that we should stop studying successful institutions. It's that we should hold the lessons with more humility — and actively seek out the failure record to test whether those lessons generalize. When a practice shows up consistently in surviving institutions, the meaningful question is whether it also shows up in failed ones before we conclude it's causal. In most cases, we haven't asked that question rigorously.

What we call "best practices" in institutional design may be better described as "practices common among survivors" — a meaningful distinction, because the same practices appear frequently in institutions that failed. That's still a real thing and worth knowing. But it's a different thing than a robust theory of what causes institutional success.

And the difference, in my view, is everything.

Frequently Asked Questions

What is survivorship bias in institutional success stories?

Survivorship bias occurs when we study only the institutions that succeeded — universities, companies, governments — and draw lessons from their practices, without examining the much larger population of institutions that had similar practices but failed anyway. The result is a distorted picture of what actually causes institutional longevity, because the sample we're learning from isn't representative.

Why do we rarely study failed institutions?

Failed institutions tend to disappear from the accessible record — they close, collapse, get absorbed, and their archives scatter. Surviving institutions produce the memoirs, case studies, and documented practices that become the raw material for institutional research. There's also a supply-and-demand problem: there's an active market for lessons about success, and almost no market for lessons about failure that cannot be reframed as a path to eventual success.

Does survivorship bias mean we can't learn anything from successful institutions?

Not exactly. Successful institutions can still tell us real things — but those lessons need to be tested against the failure record. When a feature appears in surviving institutions, the meaningful question is whether it also appears in failed ones. If it does, it's probably not the causal factor we think it is. The goal isn't to stop studying survivors; it's to hold the conclusions with more humility and widen the sample.

What are some documented examples of survivorship bias in institutional thinking?

Jim Collins's Built to Last is one of the clearest cases — many of the "visionary companies" identified in 1994 struggled significantly in subsequent decades, suggesting the identified practices weren't as robustly predictive as they appeared when the sample was limited to survivors. The same pattern appears in constitutional design: the U.S. Constitution is so frequently cited as a model that most people are unaware the average national constitution lasts only about 19 years — making the U.S. document an extreme outlier, not a representative example.

What should institutional builders do differently given survivorship bias?

Two things help practically. First, actively study failed institutions with similar profiles to the successful ones you're learning from — the contrast often reveals which features are genuinely causal and which are just common to the era or sector. Second, distinguish carefully between features that preceded an institution's survival versus features that developed after it cleared its most vulnerable years. Institutions that survive develop strong cultures partly because they survived — not just the other way around.

Last updated: 2026-06-16

Jared Clark is the founder of PatternThink, where he writes about the hidden structural patterns that shape institutions, organizations, and human systems.