Last year, Olivier Blanchard and Daniel Leigh at the IMF produced a paper that claimed that the IMF and other international organizations had underestimated the size of fiscal policy multipliers. The paper argued that the assumed multiplier of about 0.5 was too low and that the right number was about 1.5 (the way you think about this number is the $ impact on GDP of a $1 fiscal policy contraction).
While that number is already large, it is possible that the true costs of fiscal consolidations are much larger. In a recent research project (draft coming soon) I have been looking at the effects that fiscal consolidations have on potential GDP. Why is this an interesting topic? Because it happens to be that during the last 5 years we have been seriously revising the level (and in some cases the growth rate) of potential GDP in these economies. And while there might be good reasons to do so, the extent to which we have done this is dramatic and one gets the sense by analyzing the data that what is really happening is that the cyclical contraction is just leading to a permanent revision of long-term GDP forecasts (I wrote about this in my latest post).
To prove my point I decided to replicate the analysis of Blanchard and Leigh but instead of using the forecast error on output growth, I used the forecast error on potential output changes. Here are the details, which follow Blanchard and Leigh as close as I can: I take the April 2010 issue of the IMF World Economic Outlook and calculate:
- The predicted fiscal consolidation over the next two years (2010-11)
- The expected change in potential output over the next two years (2010-11).
I then look at the actual change in potential output during those years (2010-11) as presented in the April 2014 IMF World Economic Outlook. Comparing this figure to the forecast done back in April 2010, we can calculate for each country the forecast error associated to potential output growth.
Most models assume that there should be no correlation between these two numbers. Fiscal consolidations affect output in the short run but not in the long run. But under some assumptions (hysteresis or growth effects of business cycles) cyclical conditions can have a permanent effect on potential output (I have written about it here). So what is the evidence?
If we include all European countries that are part of the Advanced Economics group as defined by the IMF we get the relationship depicted in the graph above. There is a strong correlation between the two variables: fiscal consolidations have led to a large change in our views on potential output. The coefficient (strongly significant from a statistical point of view) is around -0.75.
Just for comparison, and going back to the original work of Blanchard and Leigh, the coefficient using output growth (not potential) is around -1.1. Because the forecast for output growth already included a multiplier of about 0.5, Blanchard and Leigh's interpretation was that the IMF had been underestimating multipliers and instead of 0.5 the true number was 1.6. In my regression, the theoretical multiplier built into the IMF model must be zero, which means that the true long-term multiplier is just the coefficient on the regression, about 0.7. But this number is very large and it provides supporting evidence of the arguments made by DeLong and Summers regarding the possibility of fiscal contractions leading to increases in debt via the permanent effects they have on potential output.
There are many interesting questions triggered by the correlation above: What are the mechanisms through which potential output is affected? Is potential output really changing or is just our perception about long-term growth that is changing? These are all interesting questions that I hope to address as I translate the analysis into a proper draft for a paper. To be continued.
Antonio Fatás