Upon realizing that I could easily get the Cyclically Adjusted Primary Balance data from the IMF as well as Real GDP data from the OECD without expending an extreme amount of energy, I decided to add to the empirical findings that already exist about austerity. I also sought to answer some of the concerns that Scott Sumner likes to express about every attempt at drawing a correlation not provided by Mark Sadowski.
I composed one list that included the Euro Area and one that did not (only to make Sumner shut up) and ran a regression taking the change in the CAPB between 2009 and 2014 as the x variable and the growth of real GDP between 2009 and 2014 as the y variable. The first list included Austria, Belgium, Canada, the Czech Republic, Germany, Denmark, Spain, Finland, France, Greece, Ireland, Italy, Japan, the Netherlands, Portugal, the United States, and the United Kingdom while the second list only had the Czech Republic, Japan, the United States, and the United Kingdom. Note that all of the countries were either at the zero lower bound or otherwise in a liquidity trap (defined as a case in which lowering the nominal interest rate to zero could not have resulted in full employment).
The second list was too small to yield any useful data, but the first list suggested that the coefficient on the change in the CAPB is about $-1.27$, with a p value of $9.2274 \cdot 10^{-4}$ (t-stat is $-4.112416418$). The second list did have a negative coefficient of about $-0.35$, but 5 samples is really way too few to actually conclude anything (maybe this is one of the reasons Sumner wants to remove the Euro Area).
One interesting fact that I noticed, that anyone could check in about 5 minutes if they cared to download the publicly available spreadsheet of CAPB from the IMF, is that, at least between 2009 and 2014, contrary to what Sumner claims, the US did less austerity than Europe. How much less? This much: $-0.3359236583554$. That is, the CAPB increased by about a third of a percentage point less in the US than in the Euro Area. If you included the UK, it would be even worse.
Why don't I include countries like Iceland in my calculations? Simple: they did not go to the zero lower bound in 2009 and they were never close to the zero lower bound, thereby making their addition to a regression about the effect of austerity in a liquidity trap a complete waste of time.
I'm still not perfectly happy with this assessment because I wanted to use the real GDP to working age population ratio instead of simply real GDP, as it would have made the intercept for the regression a little bit less uncertain -- while steady state real GDP growth is indeed quite variable, adjusting for demographics seems to solve the problem most of the time. Unfortunately FRED has annoying limitations with graphs and data lists, so this is the best I could do without spending more than a day on this.
The main point here is that, excluding countries obviously not at the zero lower bound, austerity is highly negatively correlated with real GDP growth and this correlation persists, even if it is not significant, when you exclude the Euro Area.
I composed one list that included the Euro Area and one that did not (only to make Sumner shut up) and ran a regression taking the change in the CAPB between 2009 and 2014 as the x variable and the growth of real GDP between 2009 and 2014 as the y variable. The first list included Austria, Belgium, Canada, the Czech Republic, Germany, Denmark, Spain, Finland, France, Greece, Ireland, Italy, Japan, the Netherlands, Portugal, the United States, and the United Kingdom while the second list only had the Czech Republic, Japan, the United States, and the United Kingdom. Note that all of the countries were either at the zero lower bound or otherwise in a liquidity trap (defined as a case in which lowering the nominal interest rate to zero could not have resulted in full employment).
The second list was too small to yield any useful data, but the first list suggested that the coefficient on the change in the CAPB is about $-1.27$, with a p value of $9.2274 \cdot 10^{-4}$ (t-stat is $-4.112416418$). The second list did have a negative coefficient of about $-0.35$, but 5 samples is really way too few to actually conclude anything (maybe this is one of the reasons Sumner wants to remove the Euro Area).
Why don't I include countries like Iceland in my calculations? Simple: they did not go to the zero lower bound in 2009 and they were never close to the zero lower bound, thereby making their addition to a regression about the effect of austerity in a liquidity trap a complete waste of time.
I'm still not perfectly happy with this assessment because I wanted to use the real GDP to working age population ratio instead of simply real GDP, as it would have made the intercept for the regression a little bit less uncertain -- while steady state real GDP growth is indeed quite variable, adjusting for demographics seems to solve the problem most of the time. Unfortunately FRED has annoying limitations with graphs and data lists, so this is the best I could do without spending more than a day on this.
The main point here is that, excluding countries obviously not at the zero lower bound, austerity is highly negatively correlated with real GDP growth and this correlation persists, even if it is not significant, when you exclude the Euro Area.
No comments:
Post a Comment