When Equilibria Converge: The Mathematics Behind DSGE Stability
This lightning talk reveals a breakthrough in understanding how dynamic economic models reach their steady states. The authors establish precise mathematical guarantees for both numerical solutions and statistical estimation in stochastic equilibrium models, proving geometric convergence rates and discovering a super-consistency regime where parameter estimates improve 500 times faster than classical theory predicts. The work unifies three major pricing models in macroeconomics and explains why some frameworks capture real-world dynamics while others systematically fail.Script
Most economic models converge to a steady state, but no one could prove exactly how fast or under what conditions. This paper changes that by establishing the first rigorous convergence guarantees for an entire class of dynamic stochastic models used to guide monetary policy worldwide.
The authors prove that any recursive equilibrium converges geometrically to its stochastic steady state, with the decay rate pinned down by the model's largest eigenvalue and shock persistence. Higher-order terms like price dispersion vanish even faster, at super-exponential rates, erasing the memory of initial conditions with remarkable speed.
The framework unifies three canonical pricing models in macro. Calvo pricing uses random resets with constant probability, Taylor contracts fix duration at multiple periods, and menu costs allow endogenous state-dependent adjustments. The paper proves these are mathematically equivalent at the ergodic level when calibrated to match average reset frequency.
A striking result bounds the impulse response peak to the maximum lag in the model. Calvo and menu cost frameworks peak within 2 periods, far too fast to match empirical inflation and output dynamics that take years to reach maximum response. Taylor contracts with longer durations generate the realistic hump shapes observed in data, explaining why empirical studies consistently favor multi-period specifications.
When second-order terms vanish around linear equilibria, parameter estimators become super-consistent, converging at rate 1 over T instead of the classical 1 over square root of T. This 500-fold improvement in short samples transforms the power of existence tests and opens new possibilities for estimation in macro data.
These convergence guarantees bridge numerical methods, statistical theory, and equilibrium existence, giving researchers confidence that simulations and regressions will recover the true dynamics. If you want to explore how stochastic equilibrium reshapes macro modeling, visit EmergentMind.com to dive deeper and create your own explainer videos.
