Why do so many people choose to lend out ETH on Aave for 2% instead of stake via Lido for 3% yield? In a new paper with Joel Hasbrouck, @cryptoeconprof, and @casparschwa we build a structural econometric model and estimate it with data from Aave and Lido. The result? There is a huge market inefficiency that cannot be explained by smart contract risk, depeg risk, or any other risk.

Even if Lido were perceived riskier than Aave, the yields should co-move: 📈 If Aave yield goes up, Lido yield should go up as well. 📉 If Aave yield goes down, Lido yield should go down. Theory predicts the yields should be positively correlated.

What happens though is that as Aave yields increase, Lido yields surprisingly decrease! Empirical yields are negatively correlated.

The persistent yield gap is not a risk premium. It is driven by other factors. Capital that may not move can explain some of the yield gap. If large capital allocators contractually may only deposit in Aave, then Aave and Lido yield would not co-move. But, this surely doesn't hold for all the $6B worth of ETH lent out on Aave.