Econometrics Seminar - Nalan Bastürk (Maastricht University)
Title: Min(d)ing the President: A text analytic approach to measuring tax news
(with A. Jassem, R.J. Almeida, L. Lieb and S. Smeekes)
Abstract: Assessing the effect of macroeconomic policy requires a proper understanding of the role of foresight. This is particularly important in case of fiscal policy, since economic agents receive clear signals, or news, about future changes in legislation. Several authors have constructed measures to capture policy news which economic agents receive at a given point in time. Such narrative measures are often based on information contained in text-based resources, e.g. newspapers. In such a way, tax shocks have been constructed in an ex-post manner for all major changes in post-war U.S. tax legislation (cf. Romer and Romer, 2010). However, such an ex-post measure does not account for the “nature” of tax news. News about possible future policy changes often do not materialise (ex-post). In this paper we propose a text mining approach to construct a measure of real-time tax news. We assess 35,000 official statements made by US presidents using a two-step LDA-based topic model. We then identify the prevalence of topics related to particular directions of changes in a given quarter and interpret this quantitatively as the strength of the tax news signal.
Password: moses Hosted by: Tomasz Woźniak