Embedded Files
My research on "Forecasting with Feedback" is supported by the National Bank of Austria's Jubiläumsfonds under project No. 19011. The official start date is September 2025; the project is expected to run for three years.
Project members
Robert Lieli, principal investigator (PI), Central European University, Vienna
Augusto Nieto-Barthaburu, international collaborator, Universidad Nacional de Tucuman, Argentina
Katrin Rabitsch, national collaborator, Wirtschaftsuniversität Wien
Blandine Ledoux, part time research assistant
Another research assistant position and a 2-year postdoctoral position will be advertised in Spring 2026.
Synopsis
The goal of the project is to study the properties of forecasts subject to feedback. Feedback occurs when an action taken in response to a forecast affects the realized value of the target variable. A case in point is central bank inflation forecasts, which play a dual role. First, they are meant to provide an accurate prediction of realized inflation over some horizon. Second, they also serve as an input for setting monetary policy, which of course influences future inflation. The project consists of i) developing theoretical models that describe forecast production in the presence of feedback; ii) developing methods to test the empirical implications of these models; iii) collecting data on inflation forecasts to carry out these tests.
Motivation
The left panel of the graph below shows the average error of the 4-quarter ahead Greenbook inflation forecast in a 10-year rolling window. (Inflation is measured by the growth rate of the GDP deflator, last release.) The right panel depicts the evolution of the slope coefficient from a corresponding Mincer-Zarnowitz regression, i.e., the regression of realized inflation on the forecast. Why do we see these surprising patterns? We argue that feedback has something to do with it.
Available results
The working paper "Forecasting with Feedback" presents a simple model of forecast production in the presence of feedback and derives the statistical properties of optimal forecasts in such environments. The paper currently has a "revise and resubmit" status at the Economic Journal; the revision has been submitted.
Abstract: Systematically biased forecasts are typically interpreted as evidence of forecasters’ irrationality and/or asymmetric loss. In this paper we propose an alternative explanation: when forecasts inform policy decisions, and the resulting actions affect the realization of the forecast target itself, forecasts may be optimally biased even under quadratic loss. The result arises in environments in which the forecaster is uncertain about the policy-maker’s reaction to the forecast, which is presumably the case in most applications. We motivate our theory by reviewing some stylized properties of Greenbook inflation forecasts. Our results point out that the presence of policy feedback poses a challenge to traditional tests of forecast rationality.
Google Sites
Report abuse