Seminario Académico - Models, inattention and Expectation Updates
Fecha de inicio: 09 de Diciembre, 2015, 13:00 hrs.
Fecha de término: 09 de Diciembre, 2015, 14:00 hrs.
El Seminario se realizará el miércoles 09 de diciembre de 13:00 a 14:00 hrs, en la Sala P-307 de FEN
El Departamento de Economía de la Universidad de Chile tiene el agrado de invitar a usted a un nuevo Seminario Académico:
Título - "Models, inattention and Expectation Updates".
Autores - Raffaella Giacomini (University College of London (UCL), Vasiliki Skreta (University College of London (UCL), Javier Turén.(University College of London (UCL)
Presentador - Javier Turén.
Abstract:
How do agents update expectations of an economic variable? Are their updates con- sistent with Bayes’ rule? What information do they incorporate: publicly available data relevant to the variable, or do they copy each other (“herding”)? Do they behave dif- ferently during a crisis? We analyze new micro data on dynamic forecast updates of US annual CPI inflation from Bloomberg’s ECFC survey over the years 2007-2014 and formu- late a theory that fits the facts and sheds light on these questions. Our theory postulates that agents use different models to produce an initial forecast and can be inattentive, in the sense that they don’t always update as new information arrives. Those who are at- tentive use their model to interpret information and update according to Bayes’ rule. We obtain structural estimates of agents’ “faith” in their model and find that it is remarkably high in non-crisis years, while it sharply drops during the crisis. Regarding the source of information that agents incorporate, we find that the scenario where they update us- ing monthly CPI fits the data better compared to herding. Bayesian updating fits well in normal times, but not during the crisis when, perhaps, agents do not have adequate estimates about the precision of their information nor of their model. We structurally estimate a number of variations of the baseline theory to assess the relative importance of its key ingredients (inattention, heterogeneous models, Bayesian updating) and find that the most important one is heterogeneity in agents’ models. Our theory nests and empirically outperforms most prevalent theories, and is simple enough to be incorporated in economic theories studying phenomena that depend on the dynamics of expectation formation.
El Seminario se realizará el miércoles 09 de diciembre de 13:00 a 14:00 hrs, en la Sala P-307 Tercer Piso del edificio Placa de la Facultad de Economía y Negocios de la Universidad de Chile, ubicada en Diagonal Paraguay 257.