Saturday, 10 November 2012

The War of the Multipliers

One month ago, the IMF presented evidence that structural macro econometrics models used by international organisations are underestimating current fiscal multipliers. Last week the European Commission presented evidence that focusing on the euro area the IMF results need to be interpreted with caution. To EC study acknowledges that there are good reasons why the fiscal multipliers can be larger in the current environment but also says that she finds no evidence of underestimation of those, al least when focusing on the euro area countries. 

More precisely the EC study argues that (I quote):

1."the forecast errors over the 2010-11 period have been predominantly underestimations of higher-than-expected growth in 2010, which were in fact associated with stimulus measures in that year";

2. "when controlling for increases in sovereign-bond yields the correlation between forecast errors and changes in the fiscal stance breaks down."

Reason 1. says that the fiscal multipliers may be underestimated, but this is due to countries that have adopted a stimulative fiscal stance (I would add: good for them!). Furthermore these stimuli are temporary in nature and imply higher multipliers than those due to permanent fiscal consolidations.

The empirical exercise is performed using structural fiscal balance, a measure that is constructed to filter out temporary factors from the actual fiscal balance. The reason why negative forecast errors in the structural fiscal stance are associated with temporary fiscal changes while positive forecast errors in the structural fiscal stance are associated with permanent fiscal changes eludes me. If this was true I would look at structural fiscal balance measures suspiciously.  

Reason 2 is harder. In the EC study, adding as an explanatory variable the change in government yields, the underestimation result disappears.
The explanation is as follows:
"…the negative coefficient for the fiscal stance in the first regression should not be interpreted as an underestimation of the fiscal multiplier but rather as capturing a negative response of investors to possibly insufficient fiscal effort in countries with severe debt problems." 
I would argue that the case for reverse causality is strong: were the yields reflecting the preoccupation of the markets for a not strong enough fiscal consolidation or for a not strong enough economic growth? In the latter case the regression suffers from an endogeneity problem (see IMF footnote 3) and the coefficients estimates would be biased. 
Here I report the benchmark results using the IMF and the EC data for the euro area countries. (GFE is cumulated growth forecast errors and FCFE is cumulated fiscal consolidation forecast errors). 

Notice how the estimates are close. The main differences in the two datasets appear to be: 1. the presence of Luxembourg in the EC data, and 2. a significant difference (sign and size) in the forecast error of the fiscal consolidation of Ireland.