An Auditable Framework for Stablecoin Depeg Detection: Validation, Bias Quantification and Evidence from a Decade of Stablecoin Data

The Great Digital Dollar Detective Story

Imagine a world where people use digital tokens that are supposed to be worth exactly one US dollar, no more, no less. These are called "stablecoins." Think of them like digital IOUs that promise to always be exchangeable for a real dollar bill. They are the lifeblood of a new kind of internet money, used for everything from sending money to friends to buying things online. But here's the tricky part: sometimes, these digital tokens lose their grip. They might drop to 98 cents or spike to $1.05. When that happens, it's called a "depeg."

For a long time, scientists and regulators have been trying to figure out how often these depegs happen and how bad they get. But there's a huge problem: nobody agrees on how to measure them. It's like trying to count how many people are "tall" in a room, but one person says anyone over 5'10" is tall, while another says you have to be over 6'2". If you change the ruler, you get a completely different count. This paper dives into that exact mess. It asks a simple but vital question: Does the number of "bad days" for these digital dollars depend on the rules we use to spot them? The answer turns out to be a resounding "yes," and the way we measure stability changes everything we think we know about these digital assets.

---

The Paper: A Detective's Guide to Digital Dollar Glitches

So, what did the researchers actually do? They built a super-strict, auditable "detective kit" to hunt down these depegging events in the data. Instead of just guessing, they created a set of rules to catch a stablecoin when it strays from its $1.00 promise. They looked at ten different US dollar-pegged stablecoins over a decade, from 2016 to 2026, using a massive database of daily prices.

Here is the big reveal: The number of depegs you find is entirely dependent on the rules you set.

If the researchers set their "sensitivity" low—meaning they flagged a depeg if the price moved just 1% away from $1.00—they found 191 events. But if they made the rules stricter and only counted moves bigger than 5%, the number of events plummeted to just 21. Even changing the "duration" rule mattered: if they required a price to stay weird for at least seven days, the count dropped from 191 to just 29. The paper shows that the "frequency" and "severity" of these crashes aren't fixed facts of the market; they are just functions of the math we choose to use. It's like saying a room is "crowded" only if you decide that a crowd starts at 10 people, but "empty" if you decide a crowd starts at 100.

The "Ghost" Events

One of the most playful and important discoveries was finding "ghost" events. The researchers found that 21.4% of the raw depeg events they initially caught were actually fake. These weren't real market crashes; they were just the digital equivalent of a ghost town. These "ghosts" happened when a stablecoin was discontinued or delisted, and the few remaining prices were just random noise from an empty market. If you didn't filter these out, you would think the market was crashing way more often than it actually was. The paper explicitly rules out counting these as real stability failures.

The Hierarchy of Trust

Once they cleaned up the data and removed the ghosts, they ranked the stablecoins by how well they stuck to their $1.00 promise. The results were clear:

• The Gold Standard: Stablecoins backed by real, central banks and fiat money (like USDC) were the most faithful, with tiny deviations of about 2.4 basis points (that's 0.024%).
• The Risky Kids: Stablecoins backed by other cryptocurrencies (like DAI and SAI) were much more jittery, with deviations of 20.3 and 125.5 basis points respectively.
• The Size Myth: The paper explicitly argues against the idea that "bigger is safer." The largest stablecoin, USDT, was actually in the middle of the pack, not the most stable. So, if you think the biggest coin is the safest, the data suggests you might be wrong.

The Perfect Rule

The researchers didn't just complain about bad rules; they proposed a better one. They used a statistical trick called "Youden's J" to find the sweet spot for detecting real problems. They found that a 2.5% daily close threshold was the optimal rule. This rule caught every single documented real-world crisis (100% sensitivity) while avoiding most false alarms. The paper suggests that regulators and investors should stop using arbitrary rules (like "anything over 1% is bad") and start using rules that are proven to work against real historical data.

How Sure Are We?

The paper is very honest about its limits. While they found that the ranking of stablecoins (centralized vs. crypto-backed) is robust, they admit that some of their deeper statistical models are "under-powered." This means they don't have enough data points to say for sure why some coins fail more than others, only that they do.

They also tested their findings against a second, independent data provider. The result? Only 61.9% of the "severe" depeg events (the big crashes) were confirmed by the second source. This means roughly two-fifths of the biggest crashes they found might just be errors in the first data provider's reporting. The paper concludes that we need to be very careful when looking at "severe" events, as they might be more about data glitches than actual market panic.

The Takeaway

This paper teaches us that in the world of digital money, how you measure stability is just as important as the stability itself. You can't just look at a chart and say "it crashed" or "it's fine" without knowing the ruler you're using. The study suggests that for the most reliable safety, we should stick to stablecoins backed by real dollars, ignore the "biggest is best" myth, and use a 2.5% threshold to spot real trouble. But most importantly, it warns us that if we don't agree on the rules of the game, we'll never agree on who is winning.