The Economic Report of the President and Fiscal Stimulus: Romer Does it Again

Chapter 3 in this year’s ERP makes the case for the positive benefits of fiscal stimulus.  The chapter brings much evidence to bear in support of the case that fiscal multipliers are large and positive.  But perhaps the most convincing is chart 3-13.  In that graph, the ERP compares forecast errors against fiscal stimulus, measured as a percent of GDP. 

Here I have replicated the chart.  The vertical axis is the deviation in 2009Q2 growth (annual rate) from private sector forecasts in the third quarter of 2008.  The horizontal axis is the level of fiscal stimulus in 2009 as a percent of GDP.  I use data from table 3-1 in the chapter.  The dates are chosen carefully:  the first date is pre-stimulus for most economies; the second date maximizes the slope of the regression line. 

The results are impressive.  The slope is positive and highly statistically significant.  Indeed, since the time of the ERP was put to bed, data revisions—especially for the Czech Republic—have improved the relationship.  For every additional percentage point additional fiscal stimulus, the economies grew 2 percentage points faster.  Notice, because we are using annual rates that is not a multiplier of 2 but rather a multiplier of 0.25. 

But the chart is deceptive.  First, the result in the chart is specific to Q2.  If we chose any other quarter in 2009 (or the year as a whole), the result is no longer statically significant.  Second, the results as shown rely entirely on the data points for Japan and South Korea.  In the absence of these two countries, the regression line has a slope of zero and the R-squared is 0.002.  Of course, one might make the argument that it would be deceptive to exclude those two data points. 

I agree, which brings me to my third and final point.  Why didn’t they include the rest of the countries included in table 3-1.  The hard part of this exercise is finding comparable measures of fiscal stimulus but they already took the trouble to include these numbers in the chapter.  With twenty minutes of effort, I tracked down both the comparable private-sector forecast for these countries and the realization of GDP growth in the third quarter. 

The results including the omitted countries are shown in the chart below.  By coincidence, the excluded economies yield a regression with almost exactly the opposite result to the included countries.  These countries yield a line with a slope of negative 2.2—every extra percentage point of stimulus lowers GDP growth by a little more than 2.  Combining all of the data into a single line yields an insignificant regression. 

Once again, the case for fiscal stimulus does not survive the microscope.  I wish, however, that the ones responsible for deciding on the level, type, and timing of stimulus were a little more honest in their presentation of the facts.