Browsing by Subject "Econometrics"

The research question that my thesis is based on is two-fold: on one hand, my goal is to investigate using instrumental variables in identifying the shocks in a structural vector autoregressive model, while on the other hand my goal is to replicate an already existing paper by Antolin-Diaz and Rubio-Ramirez, which uses narrative sign restrictions in identifying the shocks. Both methods are new, which makes the contributions that I am making relevant and interesting. The use of instrumental variables in economics has been hindered by the fact that one must be careful about the instruments that they use. Strong instruments have all the desired properties and present no issues when using them in identification, but if the instruments are weak, most of the inference measures are wrong and cannot be used in drawing conclusions. This is the contribution that the method proposed by Olea, Stock and Watson present, and I am using. Their method is robust to weak instruments as well, which enables a freer usage of instruments in identification of models. I am using two different instruments and the data that Antolin-Diaz and Rubio-Ramirez use in their paper. The main result of the comparison that I do is that the methods provide results that are distributed in a similar way, which indicates that the two methods are similar. The one difference was the existence of a price puzzle when using one of the instruments, and this is something that is investigated in detail, discussing about whether the existence of a price puzzle can be a feature of the model or if it is always an indication of the model in question being misspecified. I believe that my thesis provides support for both methods being used as they are confirming the results that they arrive at. I also conclude that my evidence can provide some more information about the price puzzle, specifically suggesting that in some cases the price puzzle can appear due to the monetary policy decisions of the decision-making authorities rather than due to the model being misspecified.