[UPDATE: Click here for a brief, explicit summary of the math.]
Last week, the BEA (Bureau of Economic Analysis) said this:
Real gross domestic product -- the output of goods and services produced by labor and property located in the United States -- increased at an annual rate of 3.1 percent in the first quarter of 2005, according to advance estimates.
That sent the market (DJI) into a 128-point tumble. Ouch.
"The market digested the economic data and wasn't at all pleased that the GDP numbers came in at a lower-than-expected rate," said Gordon Fowler, Jr., chief investment officer at The Glenmede Trust Co.
Good news: It isn't really that bad
I know how the BEA calculates the annual growth rate, and I’ve tested it against some simple alternatives. (I have some private-sector experience in forecasting, and I've done this type of analysis many times.) Based on those tests, I’ve concluded that there’s a better way to estimate growth—and it says the GDP growth rate is 3.4%, not 3.1%. Three-tenths of a growth point isn't chump change. It's 10% more than we'd thought, and a ten percent difference can move markets. To borrow a cogent sentence from President Bush: That's a lot.
The BEA's method for calculating growth uses the change in GDP over a single quarter. (Divide “GDP now” by “GDP last qtr,” raise the result to the fourth power, subtract one, multiply by 100, and voila: that's the growth rate.)
The diagram below illustrates why a two-quarter approach seems to work better for GDP growth estimates than the BEA's one-quarter method. To illustrate the concept, I’ve used the analogy of a stretch limo climbing a mountain.
Click to enlarge.
The next chart shows the specific GDP numbers from last week’s BEA release, and shows how I arrived at the 3.4% estimate.
Why do I think 3.4% is a better estimate? Because I went back to 1948’s data and tested the BEA’s current one-quarter method against the two-quarter method, quarter-by-quarter, all the way to the present. I awarded one “matchpoint” to whichever method yielded a better forecast for GDP looking 3 quarters into the future.
Result: In the 226 quarters since Q4 of 1948, the two-quarter method beat the BEA’s one-quarter method by a score of 124-102, or 55% to 45%. Better yet, since Q1 of 1993, it won by a score of 31-18, or 63% to 37%. Below is a table showing how the backtesting turned out.
That’s it. That’s why I think the real economy is growing at a rate of 3.4%, instead of the “official” number of 3.1%.
Final footnote: I minimized the image below on purpose. If you are not interested in the math behind this analysis, you don’t need to proceed any further. However, if you’d like to see the math or check the calculations for yourself, click on the image.
----------------------------
UPDATE, 5-10-2005: Ironman (Craig Eyermann) has provided a web-based calculator for anyone who'd like to calculate the GDP growth rate using the "Climbing Limo" method described above. Click here to go to Craig's calculator.
