| The first chapter shows that the Fed reacts to financial (in)stability indicators such as credit spreads over and beyond their information content on future inflation and future output. When credit spreads are rising, the Fed lowers the Fed Funds rate more than what would be required by a standard forward-looking Taylor rule. In a non-parametric framework, I show that, when credit spreads are on the rise, the probability that the Fed (Greenbook forecasts) will make a large error in forecasting future output and inflation increases. Thus, the Fed's behaviour could be seen as an insurance policy against the downside risks to the baseline forecasts, aimed at protecting the economy against an extreme event (a deflation), with a law probability of occurrence but with devastating consequences, should it occur.; As it is the case for every insurance policy, the Fed's has a cost. In the second chapter, we measure it by running counter-factual simulations in a VAR framework. We find that, if the Fed did not react to credit spreads per se (i.e. over and beyond their information content on future inflation and future output), output gap variability would be higher and inflation and interest-rate variability would be lower. Ultimately, the cost of the insurance policy over the last twenty-five years was minor: it did not produce any significant loss of economic welfare.; In the third chapter, I investigate whether GDP revisions are mainly "noise" or "news". In particular, focusing on the first GDP estimates (the "advance" estimates), I analyse whether the Bureau of Economic Analysis could produce more accurate forecasts with a better use of the data at its disposal at the time of the releases. Exploiting the information from a large real-time macroeconomic data set (using a Dynamic Factor Methodology), I find that about 10% of GDP revisions can be forecasted at the time of the initial release. It is possible to utilize the (relative) predictability of the revisions in order to improve the accuracy of the first estimate. |
