TIMESERIESFOREC · Advances in Empirical Methods for Time Series and Forecasting in Unstable Environments

The environment we live in is both complex and time varying. Examples of recent instabilities include the recent financial crises as well as the more recent COVID-19 pandemic and the war in Ukraine, which have substantially altered our world. Currently, however, the way researchers implement forecasts as well as the way they estimate the effects of economic policies is based on methods that either impose restrictive assumptions on the nature of instabilities or quickly become computationally demanding in the presence of instabilities.This project proposes to develop a local projection-based estimator for time-varying parameter models and their impulse response functions in unstable environments, which we refer to as the “time-varying parameter local projection” estimator (or “TVP-LP” in short). The proposed estimator is expected to provide a feasible approach to conveniently and flexibly estimate economic models in unstable environments as well as, more broadly, forecasting and assessing the effects of economic policies. The proposed estimator can be generalized to include instrumental variables as well as be applied in vector autoregressive models with external instruments while being robust to the presence of instabilities. The proposed methodology is expected to have widespread applicability given the increasing interest in using convenient local projection estimators and the substantial evidence of instabilities in macroeconomic models.In contrast to conventional time-varying parameter VAR models, the proposed time-varying parameter local projection estimator is robust to the presence of non-invertibility due to omitted variables and misspecification and its estimation is less computationally challenging.The advantages of the methodology will be illustrated in terms of forecasting ability as well as the evaluation of the effects of economic policy (in particular, fiscal policy).