A1720
Title: Using heterogeneous panels to estimate nonlinear VAR dynamics
Authors: Peter Pedroni - Williams College (United States) [presenting]
Abstract: A new technique is developed for estimating nonlinear VAR representations of transition dynamics in heterogeneous time series panels. Specifically the technique uses a two-step approach by first estimating a heterogenous sample distribution of linear approximations for the dynamics and then using the cross-sectional variation to estimate state-dependent nonlinear expansions. The approach is thereby able to exploit the greater variation in historical experiences that are typically present in multi-country panels relative to single country time series in order to estimate a state dependency function. Monte Carlo simulations show promising small sample performance. The technique is illustrated with the estimation dynamic fiscal multipliers which are represented as a vector function of the GDP growth rate and fiscal expenditure as a share of GDP these vary over the phases of the business cycle.