Peking University, May 15, 2019: Professor Thomas J. Sargent, a renowned economist as well as 2011 Nobel Economics Prize Winner delivered a speech titled “Risk, Uncertainty, Value, and Public Policy” in Dongxu Academic Hall in School of Economics.
Professor Sargent introduced the definition of uncertainty and the reasons why the uncertainty of models shall be paid attention to. He explained that modeling is actually a random process and the uncertainty is about worries that models might be wrongly set. However, since all subjects share common knowledge, the possibility that a model is wrongly set is very likely to be ignored in rational expectational models. On the one hand, Savage’s axiomatic system of subjective probability was questioned by Ellsberg. On the other hand, in statistical sense, it is very difficult to distinguish different macro models based on macro data sets of different sizes. Therefore, during the process of setting models and making decisions, the uncertainty of models shall be taken into consideration.
Professor Sargent elucidated on how to understand the uncertainty of models. The essence of introducing uncertainty to models lies in making use of max-min expected utility to calculate the boundary of value function. He also further pointed out relative entropy can be used to measure uncertainty.
Professor Sargent introduced the impact that the uncertainty of models has on the balance of it. He pointed out that a balance approximate to rational expectancy is expected. In reality, the expectation will help form a new balance. Therefore, in terms of equilibrium price, the uncertainty of models enlarges the influence of ordinary random discount factor and consequently produces “market prize of uncertain models.” In addition, since portfolio holders' beliefs about worst cases scenarios could affect state-dependent prices, this will prompt Ramsey planners to adopt more disciplined and purposeful strategies.
Professor Sargent illustrated that the reasons why other methods to analyze uncertainty of models are not applied are their changeability and indistinguishability.
Professor Sargent also introduced two classic consumer preference as far as model uncertainty is concerned: constraint preferences and multiplier preferences. The prior is more straightforward while the latter is easier to deal with.
Written by: Xu Yuanrui
Edited by: Huang Yadan
Source: PKU News (Chinese)
