A reduced form ARDL approach to counterfactual policy analysis : Assessing the effect on euro area GDP growth of non-standard measures introduced by the ECB

Abstract

The aim of this thesis is to investigate the effect on GDP growth in the euro area from the non-standard measures introduced by the ECB in response to the financial crisis in late 2008, using a method for counterfactual policy analysis developed by Pesaran and Smith (2012). The method they developed is interesting because it does not require the specification of a structural model. For a complicated subject like monetary policy transmission such structural models can become quite large. Instead of a fully specified structural model, Pesaran and Smith show that under certain assumptions, a reduced form model is enough to estimate the ex-ante counterfactual policy effect. This set up has the potential to allow quicker and easier analysis of policy, by bypassing the complicated structural models. In the dynamic case the structural form takes the shape of an ARDL model containing a GDP growth variable, the policy variable and a set of conditioning variables. As long as there are no structural breaks a forecast of the dependent variable a number of periods into the future can be undertaken. This forecast, conditional on realised policy variable values, is then compared to a forecast conditional on the counterfactual values of the policy variable. The difference between these two conditional forecasts will represent the effect of the policy compared to the counterfactual “no policy” scenario. The main picture from my estimations of the euro area is that the results support the view that the measures have had an effect on GDP in the euro area. The first specification of the model, which is a direct application of Pesaran and Smith’s model gives quite low, and rapidly decreasing, effects, but expanding the model by including more lags produces larger effects. The results of a second set of models which are better adapted to the euro area data set, indicates a stronger effect of the measures. Depending on which policy variable is used in the estimation, the annualised effect on GDP growth in the third quarter after introduction ranges from -0,4 to 2,7 percentage points.