Decentralised stablecoins are impossible

The recent collapse of the stablecoin TerraUSD (UST) unveiled a hard truth — decentralised stablecoins are not stable. At the outset, stablecoins are cryptocurrencies that aim to hold a stable price against some target asset. The most common target is US$1 ($1.37), always redeemable at any point. Proponents of stablecoins assert that stablecoins will reduce transaction costs and make economic activities more efficient. In particular, the most desirable form of stablecoin is trustless and capital-efficient: it depends only on code to maintain the target price with no trusted party and requires less than full backing.

There have been quite a few attempts at decentralised, capital-efficient stablecoins. IRON famously failed a year ago. UST broke its peg by falling to below US$0.20.

Why is it that stablecoins are so, well, unstable? Many developers and investors believe that past instability stems from some coding error or design bug. We can show that to the contrary, stablecoins face fundamental limitations that affect their longterm viability.

Our results come from research at the intersection of computer science, finance, and economics. Computer science tells us that some questions about computer programmes are unsolvable. Finance tells us some trading strategies leak value over time. Economic theory tells us there are limits on the amount of value we can raise from hastily selling assets. While none of these principles sound controversial on its own, in combination they allow us to prove that decentralised stablecoins are technologically infeasible. No coder can solve this problem; at issue is a contradiction in the design of any decentralised stablecoin, even those that have not been invented.

To see why, consider different stablecoin designs. First, a stablecoin backed by 100% or more of the target asset will work with no issues. Such a stablecoin involves trust because some person or some entity has their name on the fiat account storing the target asset. This is unsurprising, but it is clearly not capital-efficient. Can we do better?

Now, let’s consider a simple strategy: the stablecoin provider holds some bag of assets as long as the backing is greater than 100%, and convert the stablecoin into the backing if the price falls to 100%. At face value, this kind of “stop-loss” approach may sound appealing—except that derivative finance tells us this leaks money over time due to non-zero external financing costs. This is not simply a transaction cost or trading efficiency issue. A classic result by financial economists Peter Carr and Robert Jarrow in 1990 shows that this is a natural feature of so-called “stop-loss” strategies.

What if, by contrast, we hold a bag of assets that is allowed to be converted on demand? Does that produce capital efficiency with stability? Auction theory tells us we cannot be sure how much value is realised from each auction, and computer science tells us we cannot work out how many auctions are needed in advance. Together, these mean that we cannot promise the asset will be redeemable at the desired value. Traders may try to do this, and they may succeed for some period of time. But that is not a stablecoin, it is a “probably-stable” coin.

What that means is that the only feasible strategy to guarantee redeemability at a specified price is to hold 100% or more of the target asset. For non-decentralised assets, that requires trust. This means we cannot have a decentralised stablecoin with a target fiat price, and we cannot have a capital-efficient stablecoin at all. Importantly, these arguments are not about “market efficiency” or some other feature of financial markets. Our results state that one cannot expect to find unlimited risk-free profits in markets at any time.

This finding has broad applications. From a product and engineering perspective, we now know that no one will ever develop a decentralised stablecoin that works. As a result, any project equivalent to, or relies upon, such a construct will not succeed. Such products are like building a faster-than-light spacecraft based on Newtonian physics: it is not going to work.

From a policy perspective, our results tell us that precisely one type of stablecoin can work, and that regulations need only admit that design. While regulators should leave room for innovation, there is no reason to hold a table for a guest who has already cancelled.

Lastly, from a market perspective, both infrastructure operators and participants should redouble their efforts to ensure systems and strategies are robust in the face of inevitable stablecoin failures and likely regulatory actions.

Ben Charoenwong is an assistant professor of finance at the National University of Singapore (NUS) Business School and head of research at Chicago Global, a quantitative global investment fund. Jonathan Reiter is co-founder of Data Finnovation. The views in this article do not represent those of NUS Business School, Chicago Global, or Data Finnovation