How does Atlanta Fed GDPNow track actual GDP?
How did the Bloomberg estimate compare?
FXL reader Saaby questioned the GDPNow estimate for the US GDP. He commented in the post:
"Ive never seen this report hit accuracy with the actual number. "
I felt the same thing.
So I thought I would take a look back as we are getting closer to the 1Q release at the end of this month and see how well the Atlanta Fed does do and compare them to the Bloomber estimates.
The results below are how the Atlanta Fed last estimate matched up with the 1st cut of the GDP released by the BEA. I also include the estimates from the Bloomberg survey of economists. This is how it all stacks up for the 7 quarters of data.
- 4Q 2015: Atlanta Fed estimate was 1.0%. The actual 1st cut of the GDP was 0.7%.. Atlanta overstated by 0.3% The Bloomberg estimate was +0.8% (overstated by +0.1%)
- 3Q 2015: Atlanta Fed estimate was 1.1%. The actual 1st cut of the GDP was 1.5%.. Atlanta understated by -0.4%. The Bloomberg estimate was 1.6%. (overstated by +0.1%)
- 2Q 2015: Atlanta Fed estimate was 2.4%. The actual 1st cut of the GDP was 2.3%. Atlanta overstated by 0.1%. The Bloomberg estimate was 2.5% (overstated by +0.2%)
- 1Q 2015: Atlanta Fed estimate was 0.1%. The actual 1st cut of GDP was +0.2%.. Atlanta understated by -0.1%. The Bloomberg estimate was 1.0% (overstated by +0.8%)
- 4Q 2014: Atlanta estimate was 3.5%. The actual 1st cut of GDP came in at 2.6%. Atlanta overstated by 0.9%. The Bloomberg, estimate was 3.0% (overstated by 0.4%)
- 3Q 2014: Atlanta, estimate was 2.7%. The actual 1st cut of GDP came in at 3.5% .Atlanta understated by -0.8%. The Bloomberg estimate came in at 3.0% (understated by -0.5%)
- 2Q 2014; Atlanta , estimate was 2.7%. The actual 1st cut of GDP came in at 4.0% . Atlanta understated by -1.3%. The Bloomberg estimates was 3.0% (understated by -1.0%).
That is the all the data I could get from the website (or so it seems).
So from the limited data (7 quarters), the Atlanta Fed:
- Got within +/-0.1% on two separate occasions. Both were in 2015.
- The largest miss was a 1.3% miss in the 2Q of 2014
- In 2015 they did much better than with an average miss of 0.225% for the 4 sample points.
- In 2014, the average miss was 1.0% for the 3 sample points
How did the Bloomberg estimates do vs. the actual?
- Economists got within +/- 0.1% on two separate occasions. This was the same as the Atlanta Fed.
- The largest miss was a 1.0% miss in the 2Q of 2014. This was the same quarter that the Atlanta Fed missed by the most (1.3% miss). The Bloomberg economists were closer.
- The 2015 average miss for the Bloomberg economists estimate was 0.3%. The Atlanta Fed missed by 0.0225%. Advantage Atlanta Fed in 2015.
- The 2014 average miss for the Bloomberg economists estimate was 0.633%. This was better than the Atlanta Fed which was off by 1.0% in 2014 on average. Advantage economist in 2014.
I don't know if 2015 was easier to predict, or GDP models got better, but for both the Atlanta Fed and the Bloomberg economists they did a better job in 2015 vs 2014 in their GDP estimates.
If the Atlanta Fed remains as good in 2016, with their estimate of 0.4%, the GDP should be in a range from 0.175% to 0.625%. There will be other contributing data before then of course, but right now, that is what we can expect (all things equal). That is not the quarter that was expected when the quarter started.
1Q GDP will be released on April 28, There are no estimates on Bloomberg yet.