The Government Spending Multiplier at the Zero Lower Bound: Evidence from the Euro Area

We use an Interacted Panel Vector Autoregressive (IPVAR) model, to investigate the effects of a government spending shock when the interest rate is at zero lower bound (ZLB). We also compare the responses of variables of interest at the ZLB with what we get when a government spending shock occurs in normal times (i.e. when the interest rate is larger than 0.25). We identify the government spending shock by sign restrictions and use the European Commission forecasts of government expenditure to account for fiscal foresight. For the baseline specification we find lower multipliers in times in which the ZLB is binding. However, fiscal foresight is not the only problem in fiscal VARs related to limited information problems. Usually, VAR models can only consider a limited number of variables due to degree of freedom problems. Several authors have shown (see Stock and Watson (2005) for a survey) how principal components extracted from a larger number of variables, can approximate unobserved factors driving most (if not all) of the macroeconomic variables. Therefore, we develop a Factor-Augmented IPVAR model (FAIPVAR) and find that the multipliers are very similar among states, ranging between 1.08 and 1.41 at the ZLB and between 1.26 and 1.39 away from it. We also divide our sample, considering two groups of countries in terms of high and low debt-to-GDP ratios. We find that countries with high levels of debt-to-GDP ratio show relatively lower multipliers. Considering the FAIPVAR model, the government spending multiplier ranges between 2.69 and 3.54 for core countries and between 0.82 and 1.37 for peripheral countries. Therefore, our findings support some recent studies, which suggest that the government spending multiplier is even larger if the debt-to-GDP ratio is low.