Computer models outperform experts in predicting US GDP

Computer models outperform experts in predicting US GDP

Judging by projections of US second-quarter gross domestic product, which were prepared ahead of the actual earnings release last Thursday, the predictions made by automated computer models are much more accurate than the calculations made by economists.

    Image source: Bloomberg

Image source: Bloomberg

The US Department of Commerce’s preliminary estimate says the country’s economy contracted for the second straight quarter, with second-quarter GDP contracting 0.9% year-on-year. According to a Bloomberg poll, the median forecast was for the indicator to rise by 0.4%, and out of 74 economists polled by the resource, 23 were counting on a fall.

The forecast of the so-called “Short-Term Forecast” models (nowcast) proved to be more realistic. According to the GDPNow forecast model of the Federal Reserve Bank of Atlanta, there should have been a 1.2% decline in GDP.

The forecast of the S&P Global Market Intelligence computer model developed by Monetary Policy Analytics fully agreed with the actual result – 0.9%. This model is used by the US government, banks and the Federal Reserve System (FRS) itself.

Short-term predictive models are gaining adherents as their accuracy increases, and the estimates they produce tend to approximate actual results as the amount of data increases. The GDPNow index, for example, forecast a 1.8% contraction in GDP earlier in the week, but revised the forecast to 1.2% by Wednesday.

The S&P model provides forecasts that are within about 1.2 percentage points of actual GDP about three months before the US Bureau of Economic Analysis release, with the gap narrowing to about half a percentage point closer to the release date, said Ben Herzon ( Ben Herzon, CEO of S&P.


About the author

Robbie Elmers

Robbie Elmers is a staff writer for Tech News Space, covering software, applications and services.

Add Comment

Click here to post a comment