Doom Loop

Some problems resolve themselves over time. Others make themselves bigger. That’s exactly what the English expression “doom loop” describes, literally something like a “loop toward doom.” It refers to a chain of events in which each consequence becomes a new cause. One example: a department store loses customers, so it cuts back on staff, and because of the poor service, even fewer customers come. The reaction to the problem worsens the problem. Experts call this a self-reinforcing feedback loop.

Why spirals are harder to stop than single crises

A normal crisis has a beginning and an end. A doom loop, on the other hand, no longer needs an external trigger once it has started. It keeps propelling itself forward as long as no one intervenes. That’s why it’s considered especially dangerous.

There’s also a timing problem. At first, the damage usually looks small and harmless. Whoever intervenes early needs little effort, but has hardly any evidence of urgency. Whoever intervenes late has the evidence, but by then the spiral is already turning fast. This trap explains why governments and boards often watch for too long before acting.

It’s important to distinguish this from a mere buzzword. Not every bad development is a doom loop. Falling revenue alone is just a trend. A doom loop only exists once you can name the circle: A worsens B, and B worsens A again.

The loop between banks and government debt

The best-known example comes from the euro crisis starting in 2010. Banks had bought large amounts of government bonds, i.e., debt notes from their own country. When doubts arose about the solvency of states such as Greece or Italy, these bonds lost value. This suddenly left the banks worse off.

Now the loop closed. Wobbling banks must be bailed out by the state in an emergency. This prospect burdens public finances, which caused the bonds to lose even more value. That weakened the banks again. Economists called this connection the “bank-sovereign doom loop” — the loop between banks and the state.

Such spirals always have the same structure. Two quantities depend on each other, and in both directions. Expectations also play a role: investors don’t sell only once a state is actually bankrupt, but already when they fear it might be. The only way to break the loop is by cutting one of the connections. In Europe, it was mainly the European Central Bank that took this on from 2012, with its pledge to buy bonds itself if necessary.

From empty office towers to AI models

In business news, the term has appeared for several years in connection with the “urban doom loop,” the downward spiral of city centers. Since many people started working from home, office towers in US cities like San Francisco stand half empty. This drives down property values and, with them, the city’s tax revenue. The city cuts back on public transit and safety, which makes the area less attractive and drives away even more businesses.

The pattern is also being discussed in the AI industry. One case is model collapse: AI systems learn from text and images found online, but increasingly, that content is itself AI-generated. New models thus train on the outputs of old models. Errors and biases amplify with every round, and quality declines.

A second example is recommendation systems, such as those used by video platforms. They show what gets clicked, and what gets shown gets clicked. Anyone who reads the term in a headline should therefore ask one question: which two things are reinforcing each other here? If that can’t be answered, “doom loop” is just a dramatic word for bad news.