Estimating the New Keynesian Phillips Curve with Regional Data and Selected Instruments

Despite the increase in nationwide unemployment following the Great Recession, the United States only saw a moderate fall in inflation, which has led to numerous debates on whether the Phillips curve has indeed flattened. The empirical discrepancies leading to this debate could stem from estimation issues related to confounding cost-push shocks and the many weak instruments encountered by the aggregate New Keynesian Phillips curve. This paper resolves these two issues by incorporating regional variation and instrument selection. Monte-Carlo simulations demonstrate that regional data help with the identification of the Phillips curve when cost-push shocks bias the aggregate estimation, and the NKPC estimation can be further improved in finite samples with instrument selection. I apply these methods to US metropolitan data and find that the aggregate Phillips curve has not flattened; on the contrary, the trade-off between inflation and unemployment remains strong when using regional data from more recent periods.