Keep It Simple, Stupid: Why Wall Street's Models Keep Failing Us

Eighteen years after the financial crisis nearly collapsed the global economy, the banking system is repeating the same fundamental mistake that caused it.

The technology is different. The instruments have evolved. But the underlying problem remains unchanged: complexity that creates an illusion of safety while reducing the very transparency and resilience that actual safety requires.

This pattern, as one prominent financial researcher observed in 2008, follows a predictable arc. The system becomes more complex. Models proliferate. Confidence grows. Then reality intrudes, and everyone discovers that the models were never modeling reality at all.

The question now is whether the industry will learn before the next crisis forces it to.

How Models Became More Important Than Markets

In the years leading up to 2008, complexity was celebrated as a virtue in banking.

The most sophisticated firms—those able to construct the most intricate financial instruments and identify profit opportunities in minute detail—dominated their competitors. Merrill Lynch built its empire on structured products. AIG sold credit default swaps with the confidence of executives who believed the risk could be engineered away.

Then the mathematics encountered the real world.

Banks operated under what became known as "mark to model" accounting. Rather than valuing assets based on what they could actually sell for, they valued them based on what their mathematical models said they were worth. The models, however, contained fundamental blind spots. They ignored liquidity—the possibility that you might need to sell an asset when you need to sell it. They assumed that securities which moved together in prosperous times would remain diversified during downturns. They failed to account for a housing collapse because the mortgage industry had not yet experienced one.

As Jon Danielsson, a finance researcher at the London School of Economics, observed at the time: "Mathematics often assumed far greater importance than the accurate depiction of reality."

In August 2007, the consequences became apparent.

Quantitative hedge funds—institutions run by teams of PhDs who had spent years perfecting market-neutral strategies—suffered losses that the underlying models said should occur once every few thousand years. Yet these events happened multiple times within a single week.

David Viniar, chief financial officer of Goldman Sachs, captured the bewilderment in an interview with The New York Times: "We were seeing things that were 25-standard deviation moves, several days in a row. There have been issues in some of the other quantitative spaces. But nothing like what we saw last week."

We were seeing things that were 25-standard deviation moves, several days in a row. There have been issues in some of the other quantitative spaces. But nothing like what we saw last week.

The models were not modeling reality. They were modeling a world that no longer existed—if it had ever existed at all.

The Illusion of Control

The deeper problem with relying on complex models for financial regulation is not that they exist, rather that complications tend to act as a substitute for judgment rather than tools for testing it.

A bank's risk committee, presented with a thirty-page report containing sophisticated models and statistics calculated to the third decimal place, naturally assumes the system is functioning properly. Regulators examining the same elegant mathematics reach the same conclusion. The numbers say everything is under control.

But control requires understanding.

And once a financial system reaches a certain threshold of complexity, understanding becomes impossible. The traders do not understand it. The risk managers do not understand it. The executives do not understand it. The regulators do not understand it. Everyone agrees to believe the models.

This creates what might be called a mass hallucination.

As long as no external shock tests the models' assumptions, belief persists. When liquidity evaporates—when the real world suddenly diverges from the model's assumptions—the system freezes. What the banks thought was precision turns out to be fiction.

After 2008, regulators attempted to address this through new rules and more sophisticated modeling. By 2010, the Basel III accord had formally enshrined complex mathematical models as the foundation of banking regulation worldwide.

The Basel III accord is an internationally agreed set of regulatory standards designed to strengthen the regulation, supervision, and risk management of the banking sector developed by the Bank for International Settlements (BCBS) in response to the 2007–2009 financial crisis.

Banks hired more risk managers. The industry invested billions in more elaborate algorithms. They responded to a crisis caused by excessive complexity by mandating additional complexity.

This should have been surprising.

Instead, it followed a predictable pattern. "It is the nature of financial regulations that they tend to be reactions to previous crises and slow to adapt," Mr. Danielsson said. Banks found loopholes in the new models. They discovered ways to exploit regulatory gaps. With each cycle, the financial system became less transparent, more interconnected, and harder for anyone to understand.

The Case for Complexity

There is, to be fair, a coherent defense of financial complexity.

The global economy is itself genuinely complex. A major international bank managing trillions in currency hedging, cross-border trade, and multinational capital flows cannot operate according to the same principles as a retail investor purchasing an index fund. The complexity within financial institutions reflects the underlying complexity of the world they inhabit.

Moreover, quantitative risk management provides a standardized language for discussing risk. Without it, institutions would make credit decisions based on intuition and historical anecdotes—arguably a more dangerous approach. Complex derivatives, like credit default swaps, allow investors to separate and hedge specific risks, enabling credit to be extended that would otherwise be unavailable.

Even regulation cannot escape the trap. A simple rule—say, a straightforward leverage ratio—might seem elegant until banks begin finding creative accounting methods to circumvent it, moving assets to less transparent entities and exploiting loopholes regulators did not anticipate. Complex regulations exist, at least in theory, to close these infinite gaps.

These arguments contain genuine force. They explain why complexity persists even after systemic crises. But they conflate two distinct problems. The issue is not that models exist, rather how such frameworks are put to use.

At best, a model, built to measure, and test risk, should be a diagnostic tool. It should stress-test scenarios. It should ask: What breaks if interest rates spike? What happens if liquidity evaporates? A model should reveal vulnerabilities that require management attention.

Instead, models have become instruments of certainty. They are used to declare that risks are contained, that the system is sound, that no further scrutiny is required. A model's output becomes proof of safety rather than an initial hypothesis to be tested against reality.

That inversion is what kills.

A Pattern Emerges Again

Eighteen years after the crisis, warning signs have reappeared in different forms. Private credit markets are expanding with minimal oversight. Artificial intelligence systems are trading automatically, identifying patterns in data that no human fully understands. Traditional banks remain tangled with shadow banks in configurations that nobody has fully mapped.

The technology has changed. The instruments are different. The terminology has evolved. But the fundamental dynamic—complexity that obscures risk while creating confidence in safety—remains.

Warren Buffett's investment philosophy, unchanged for decades, offers a direct challenge to this approach. Stick to what you understand, he advises. For most investors, this means a low-cost index fund: ownership of productive companies, compounded over time. No leverage. No derivatives. No strategies requiring a PhD to explain.

For those managing the financial system, the principle is identical, though the implications are more costly. It would require walking away from instruments not fully understood. It would mean accepting lower profit margins on deals that can be clearly explained. It would require admitting that much of what passes for financial innovation is obfuscation dressed in mathematical language.

The industry has resisted this approach with consistency. As recently as 2019, more than a decade after the previous crisis, regulators were still debating whether to implement a simple leverage ratio—a straightforward cap on borrowing that requires no modeling at all. Even today, banks that fail basic stress tests based on the leverage ratio continue operating because more complex models say everything is fine.

The Path Forward

True resilience in finance requires one thing: the willingness to admit what cannot be known. It does not come from more granular data, faster computers, or better algorithms.

The history of financial disasters is not a history of insufficient information. It is a history of misplaced confidence in information poorly understood. The system has repeatedly confused precise numbers with precise knowledge, believing that mathematical precision could engineer away fundamental uncertainty.

For regulators and institutional leaders, the path is clear. Move toward transparency and away from the black box. Stop asking for more complex models. Start demanding structures simple enough that they do not require a supercomputer to monitor or a PhD to understand.

Demand leverage constraints that do not depend on modeling. Demand markets where investors can understand what they are buying and selling.

Will this happen? The incentive structure suggests otherwise. Complexity remains profitable—at least until it explodes. And by then, the losses have been distributed in ways that insulate those responsible for building the system.

Warren Buffett understood this decades ago. Jon Danielsson understood it in 2008. Whether the industry will learn before the next crisis becomes unnecessarily painful remains an open question.

The pattern, however, suggests an answer.

• The author is the Head of Research and Analysis at Icarus Asia, a Hong Kong-based risk and advisory business

Sources:

Jon Danielsson, "Complexity Kills" (2008), Centre for Economic Policy Research. Updated commentary available on RiskLab.

Amir E. Khandani and Andrew W. Lo, "What Happened To The Quants In August 2007?" (MIT, September 2007).

Carmen Reinhart and Andrew Felton, eds., "The First Global Financial Crisis of the 21st Century" (VoxEU/CEPR, 2008).