A1362
Title: Nonlinearities in estimation of the Phillips curve
Authors: Giulia Gitti - Brown University (United States) [presenting]
Abstract: The consequences of introducing nonlinearities in the estimation of the slope of the Phillips curve are investigated using US regional data. Panel variation in inflation and unemployment rates effectively deal with threats to the identification from aggregate endogenous policy responses aiming at stabilising inflation or economic activity. However, local unemployment rates could still be driven by local labour supply shocks, leading to an omitted variable bias. Following the literature, a shift-share instrument is used, exploiting regional sectoral variation to isolate demand-driven fluctuation in local unemployment rates. Identification based on instrumental variables assumes a linear relationship between the instrument and the instrumented variable. This assumption is relaxed by allowing for a piecewise linear relationship between the shift-share instrument and the unemployment rate. Using this novel identification strategy, a standard log-linear functional form fits the data better than nonlinear ones when estimating the Phillips curve from 1992 to 2023. The estimated slope of the standard log-linear Phillips curve over this sample period is -0.74, negative and significant. Moreover, it is shown that the estimated slope changes over time. Such results corroborate the hypothesis that the relationship between inflation and unemployment changes over the last 80 years is consistent with a time-varying slope of a log-linear Phillips curve.
