Min(d)ing the President: A text analytic approach to measuring tax news

TL;DR

This paper introduces a novel text analytic method to quantify tax news from presidential communications, predicting future tax changes and analyzing their macroeconomic effects, especially on output over long horizons.

Contribution

The paper develops a new text analytic approach to measure tax news from presidential communications, capturing signals beyond traditional narrative measures.

Findings

Tax news predicts future tax changes accurately.

Pre-implementation tax news signals cause output contractions.

The method uncovers signals not detected by previous measures.

Abstract

Economic agents react to signals about future tax policy changes. Consequently, estimating their macroeconomic effects requires identification of such signals. We propose a novel text analytic approach for transforming textual information into an economically meaningful time series. Using this method, we create a tax news measure from all publicly available post-war communications of U.S. presidents. Our measure predicts the direction and size of future tax changes and contains signals not present in previously considered (narrative) measures of tax changes. We investigate the effects of tax news and find that, for long anticipation horizons, pre-implementation effects lead initially to contractions in output.