Geeta Garg - Research

Household Beliefs and Japan's Lost Decades: The Role of Fiscal Policy Credibility (Job market paper)

The fiscal authorities in Japan were unsuccessful in credibly committing to the future path of fiscal policy. Concerns over rising deficits led to the promises of fiscal consolidation and weak economic conditions led to the reversal of those promises (chart below), creating uncertainty about their future fiscal stance. In a New-Keynesian DSGE model with rational expectations, I examine the extent to which the uncertainty due to these repeated promises explain the slowdown experienced by the Japanese economy. I assume Markov-switching tax rules such that the response of taxes to debt vary according to the fiscal stance of the government. I document that these promises generate time-variation in both the expected value and volatility of the tax rates. Even in the regime, in which taxes do not stabilize debt, the rising level of debt create expectations of higher future taxes causing economic contraction in the current period. These expectations lead to a decline in consumption and an increase in debt. The investment and labor hours decline depending on whether it is the uncertainty about future capital or labor-income tax. The model does not generate low inflation as seen in the Japanese data.

Are Tax Cuts Expansionary? Narrative Evidence based on Japanese Data

In this paper, I examine the macroeconomic impact of anticipated tax liability changes in Japan. I find that the tax cuts that have been announced but not yet implemented lead to a decline in output, consumption, investment, hours worked and real wages, generating a slowdown during the period before their implementation. These anticipation effects depend on the size of implementation lag. Lower is the implementation lag, less pronounced is the slowdown generated by the announcement of tax cuts. Once the tax cuts are implemented, tax cuts are expansionary only when the implementation lag is less than 2 quarters. After consumption taxes are excluded from the data, the implemented tax cuts are not only contractionary, the magnitude of decline is also large. Thus, implemented tax cuts are not expansionary as observed in the U.S. data. These findings imply that long implementation lags in the tax policy can be detrimental to the economic recovery.

Impact of Fed Chairmen Appointments on the Financial Markets

This paper examines the reaction of financial markets to the Federal Reserve chairmen appointments using a new dataset based on the daily counts of news articles that discuss these appointments. To the extent these appointments convey new information about future monetary policy, financial markets respond to them as a result of revision in their expectations of the future path of interest rates or inflation. I find that financial markets reacted adversely (Yen appreciated against USD, bond yields increased and stock returns slightly declined) in response to Volcker's departure or Greenspan's (first) appointment. However, there was a muted response of financial markets to the appointments/departures that occurred afterwards. I plan to extend the analysis to examine the reaction of financial markets to the appointments of Presidential candidates.

Work in Progress: Papers and Projects

Construction of Real-time measure of Interest Rate Expectations using Textual Data

The goal of this project is to construct a real-time measure of public-signals about pending interest rate changes during "the Great Inflation" period. The objective is to understand the timeline of change in interest rate expectations before and after the announcement of the appointment of Volcker as the chairman of the Federal Reserve. To construct this measure, I utilize textual data which consists of a corpus of newspaper articles, collected from two major newspapers - the New York Times and the Wall Street Journal. These newspaper articles are scanned pdf images that undergo several text pre-processing techniques to bring them into a useful form. I then apply two different supervised probabilistic topic models on this dataset - Latent Dirichlet Allocation (LDA) and the Non-Negative Matrix Factorization (NMF) - which are forms of dimension reduction techniques that collapse together semantically similar terms, to discover topics to which the documents belong. In this document, I describe each of these methods in detail, their pros and cons, their comparison, and commonly used posterior inference techniques such as Variational Inference and Gibbs sampling. I find that NMF method, despite its simplicity, is at least as good as LDA in terms of extracting topics from the textual data. This project is largely a work-in-progress. I plan to extend this analysis to a longer time-series of textual data.