Backtesting the Modified Limo

In between our slew of ongoing projects, we've also taken time to work with our modified method of applying the Climbing Limo approach to the task of forecasting inflation-adjusted GDP for the U.S. in future quarters. The chart below shows some of the preliminary backtesting we've done to see how well our Modified Limo would have fared against where real GDP really came in at:

The main difference between the two methods is how they're applied. The Climbing Limo method is used to forecast real GDP some three quarters ahead in time, while the Modified Limo is only used to forecast where the final revision of real GDP will be in the next quarter. In the chart above, we find that for the next quarter for which GDP data will be released, 2007-Q3, the Modified Limo forecasts a real GDP level of 11,583.3 billion (2000) USD, while the original Climbing Limo method forecast a real GDP level of 11,530.0 billion (2000) USD. Simply projecting the most recent quarter's growth into the next quarter would put this figure at 11,628.6 billion (2000) USD.

Just for fun, we backtested the Modified Limo method using quarterly real GDP data going back to 1947, comparing the forecast results against what simply projecting the previous quarter's rate of growth into the next quarter. In looking at the absolute error (the absolute difference between forecast and actual results), the Modified Limo method produced the following outcomes:

• Beginning with 1948-Q2, the Modified Limo was better than a 1-quarter projection in 59.5% of the quarters since.
  
• Beginning with 1950-Q1, the Modified Limo was better than a 1-quarter projection in 60.0% of the quarters since.
  
• Beginning with 1960-Q1, the Modified Limo was better than a 1-quarter projection in 62.1% of the quarters since.
  
• Beginning with 1970-Q1, the Modified Limo was better than a 1-quarter projection in 64.0% of the quarters since.
  
• Beginning with 1980-Q1, the Modified Limo was better than a 1-quarter projection in 65.5% of the quarters since.
  
• Beginning with 1990-Q1, the Modified Limo was better than a 1-quarter projection in 67.1% of the quarters since.
  
• Beginning with 2000-Q1, the Modified Limo was better than a 1-quarter projection in 76.7% of the quarters since.
  

We'll know how close we were with our Modified Limo forecast when the final revision for 2007-Q3 is released on December 20, 2007. And yes, when we get around to it, there will be a tool!