PhD Lunch Seminars Rotterdam

Aysil Emirmahmutoglu, Didier Nibbering
Thursday, June 16, 2016

Aysil Emirmahmutoglu (Erasmus University Rotterdam)

‘Zooming in on Ambiguity Attitude’

Typical ambiguous events people are facing are rare, extreme events such as catastrophes. Yet, empirical studies of ambiguity attitudes so far have focused on events with moderate likelihood and extrapolation to rare events must be subject to caution. We conducted an Ellsberg-like experiment to study ambiguity attitudes for very unlikely events, and for both gains and losses. We measured ambiguity attitudes with additivity indices, with neither assumptions on subjects’ beliefs nor restrictions to specific ambiguity models. Overall, very unlikely events were overweighted, looming larger in isolation than when part of larger events. Using latent profile analysis, we classified the subjects based on the extent they deviated from ambiguity neutrality. One third of the subjects behaved close to ambiguity neutrality. The others exhibited various patterns of overweighting of rare events. Such behavior can lead to money-pump situations. (Joint work with Aurelien Baillon)

Didier Nibbering (Erasmus University Rotterdam)

‘A Bayesian Infinite Hidden Markov Structural Vector Autoregressive Model’

We propose a Bayesian hidden Markov model with an infinite number of states to estimate time-varying parameters in a structural vector autoregressive (SVAR) model. The Markov structure allows for heterogeneity over time while accounting for state-persistence. By modelling the transition distribution as a Dirichlet process mixture model, parameters can vary over an infinite number of regimes. The Dirichlet process favours a parsimonious model without imposing restrictions on the parameter space. In an application on a small monetary SVAR we find evidence for heterogeneity in the volatility of monetary policy shocks and impulse responses to these shocks. These features can only be captured when we allow both the coefficient and the covariance matrix in the SVAR model to be time-varying.