One of the most closely watched economic measures is US GDP (gross domestic product). Sales forecasters would love to have an early warning variable in their models. Is GDP the missing link for anticipating major economic shifts? Unfortunately, the short answer is probably no. US GDP forecasting is quite inaccurate.
The Federal Reserve Board sponsors a quarterly survey of professional forecasters, called the SPF. The survey is returned to the Fed mid-way through the quarter, and includes forecasts for the current quarter and the next four quarters. As shown in the chart below, the best MASE is only 61%, and that is when forecasting the current quarter -- which was half-complete when the forecast was submitted! The four-quarters-out forecasts are barely better than a random walk, or the simple projection that the current GDP growth rate will continue through the rest of the year.
One bit of good news: US GDP forecasting is improving. The forecast error trend channel is narrowing. In particular, the SPF has been resisting the temptation to be overly optimistic during economic downturns.
Trend channels are an easy and effective way to visualize forecast error patterns. Let's compare two Germany GDP forecasters: the Joint Diagnosis and the GAER (German Annual Economic Report).
As shown in the chart below, forecast error (Actual - Forecast) for the joint Diagnosis is not improving. The trend channel lines are parallel. They are quite vulnerable to excessive optimism during economic downturns. Note the negative errors in 1975 and 2009 (i.e., these forecast were much higher than actuals).
In contrast, forecast error is improving for the GAER. Most notably, they did not fall prey to excessive optimism during the recent economic crisis. Perhaps they learned from prior mistakes.
Let's say you sell a product in Germany, and have observed that there seems to be a correlation between GDP and your product sales. Should you include it in your forecasting model? Well, it depends on the accuracy of the Germany GDP forecast.
The best way to measure forecast accuracy is MASE (Mean Absolute Scaled Error). It was proposed by Rob Hyndman in 2006 and compares forecasts to the naive or the "no change" prediction. The chart below shows MASEs for four different Germany GDP forecasters.
As we can see, the best forecaster had a MASE of 55%. This is substantially better than a random walk, or simply predicting this month will repeat next month. Ironically, the mot publicized GDP forecaster, the Joint Diagnosis, has the worst record.
Many sales forecasts have MASEs well below 55%. If this is true for your Germany sales forecast, then there is a low probability that your forecasts will improve by adding Germany GDP.