Research
Working papers
Abstract
The study proposes an alternative way to decompose Federal Reserve (Fed) information shocks from monetary policy shocks by employing a textual analysis to Federal Open Market Committee (FOMC) statements. I decompose Fed statements into economic topics using Latent Dirichlet Allocation (LDA). The model was trained on the business section from major US newspapers. After decomposing surprises in Fed futures into a part that is explained by topics from the Fed statements and that is not explained, the study employs these purged series as proxies for monetary policy and Fed information shocks. The results show that, compared to surprises in 3-month federal funds futures, a policy shock identified in this study has a more negative effect on GDP and a more prolonged negative effect on inflation. In the short-run it causes S&P500 to decline and the Fed to raise its interest rate. Identified Fed information shock affects the macroeconomy as the standard news shock: it has positive long-run effects on S&P500, interest rates, and real GDP, whereas it has a negative short-run effect on inflation. Moreover, the Fed information shock reduces credit costs.
Keywords: FOMC, statements, Latent Dirichlet Allocation, monetary policy, information, shocks
JEL Classification: E52, E31, E00
Central Bank Communication and Forward Guidance Shocks
Abstract
Does it matter what the central bank had said during a monetary policy announcement? The paper proposes a new approach to identifying the effects of forward guidance taking into account what the central bank had said and how financial markets perceived it. I use computational linguistic methods to disentangle the topics the central bank was talking about during its announcements. Moreover, I found that newspapers cover monetary policy announcements and positivity/negativity of the coverage might serve as a proxy to what the markets have heard during the announcements. The study finds that the markets react to monetary policy announcements by incorporating the new information in the yield curve. Furthermore, financial markets are forward-looking and obtain information before its publication in newspapers.
Keywords: forward guidance, shock, statements, Latent Dirichlet Allocation, central bank
JEL Classification: E52, E31, E00
Abstract
I investigate the role that news sentiment plays in the macroeconomy. Using an approach that combines Doc2Vec embedding and Latent Dirichlet Allocation with lexical-based models I show that the news the media choose to report and the tone of these reports contain important information for household unemployment, interest rates, and inflation expectations. Topic time series derived from the news and the sentiments they express are employed to estimate how the news affects the macroeconomy.
Keywords: expectations, sentiment, news, Latent Dirichlet Allocation (LDA), Doc2Vec
JEL Classification: E52, E31, E00
Abstract
Whether Federal Open Market Committee (FOMC) discussions contain additional information for Taylor rule estimation? While researchers usually estimate a Taylor rule with official forecasts of the output gap and inflation, or official statistical data as instruments, this paper sheds light on the importance of additional information contained in the FOMC discussions. Using text analysis techniques I detect all economic phrases from FOMC transcripts. Employing sentiment analysis of the most frequent economic phrases and Bayesian LASSO with time-varying coefficients I show that the Fed changes the federal funds target also in response to FOMC members’ uncertainty about inflation expectations, financial markets and monetary aggregates.
Keywords: FOMC, Transcripts, Sentiment, Taylor, Uncertainty, Positiveness
JEL Classification: E52, E31, E00
Abstract
I propose a new approach to identifying exogenous monetary policy shocks that requires no priors on the underlying macroeconomic structure, nor any observation of monetary policy actions. My approach entails directly estimating the unexpected changes in the federal funds rate as those which cannot be predicted from the internal Federal Open Market Committee’s (FOMC) discussions. I employ a Neural Network and basic machine learning regressors to predict the effective federal funds rate from the FOMC’s discussions without imposing any time-series structure. The result of the standard three variable Structural Vector Autoregression (SVAR) with my new measure shows that economic activity and inflation decline in response to a monetary policy shock.
Keywords: monetary policy, identification, shock, deep learning
JEL Classification: E52, E31, E00