Non-linear behavioural New Keynesian models: Design and estimation

This thesis studies empirical Behavioural New Keynesian (BNK) DSGE models incorporating heterogeneous expectations and reinforcement learning to endogenise the distribution of agents’ type, focusing on higher-order estimation methods to capture the non-linear features of the learning mechanism.
Chapter 1 analyses the properties of local Gaussian filters for Bayesian posterior inference on the parameters of non-linear DSGE models. Various filters are assessed by estimating parameters on simulated data: second-order Extended Kalman filter, risky linear approximations, and sigma-point filters. Results show that these filtering techniques help with empirical problems characterized by high computational burden.
Chapter 2 develops a BNK model enriched with portfolio adjustment costs to study long-term asset purchases. Adjustment costs on the composition of the households’ financial portfolio allow for bond-market segmentation by introducing a wedge on the yields paid by bonds with different duration. Reinforcement learning combined with bounded-rational agents is exploited to study state-dependent asset purchases multipliers, by linking policy measures to the endogenous economic sentiment. Simulations support the role of asset purchase programs as counter-cyclical measures and emphasize the importance of Central Bank credibility for monetary policy transmission.
Chapter 3 estimates a small BNK model with trend inflation. A formal test for parameters identification shows that core reinforcement learning parameters can only be jointly identified using higher-order approximations of the model while expanding the information set with a proxy for the share of agents’ types based on consumers’ survey data on expectations. Estimates relying on non-linear filters outperform a rational expectation counterpart in matching higher-order empirical moments.