Computers outperform humans in accuracy on US GDP projections

    Need to understand where the US economic growth rate is now and likely to be in the future?

    Need to understand where the US economic growth rate is now and likely to be in the future?

    It might be better to look at calls from largely automated computer models rather than economists, judging by second-quarter gross domestic product estimates before actual results are released. on Thursday.

    The US economy contracted for a second straight quarter, with GDP shrinking at a 0.9% annualized rate from the first three months of the year, according to preliminary estimates from the US Department of Commerce. The median estimate in a Bloomberg survey was up 0.4%, and out of 74 economists’ forecasts, 23 were bearish.

    However, the forecasts in the so-called “forecast” model are closer to the results. For example, the Federal Reserve Bank of Atlanta’s GDPNow index fell 1.2%.

    Another similar computer model has had direct success: Estimates from S&P Global Market Intelligence, originally conceived by Monetary Policy Analytics Inc. — co-founded by former Fed Governor Larry Meyer — predicts a 0.9 percent decline in the second quarter. Its clients include governments, banks, and the Fed itself, which uses the data to glean insights into how the economy is going.

    The latest GDP report has been closely watched, given that two consecutive quarters of contraction is a general rule many people use to gauge whether an economy is in recession. The formal determination of the end and the beginning of the business cycle was made by a group of scholars at the National Bureau of Economic Research.

    Nowcast meters have gained more followers as their accuracy has improved, and the estimates they generate are closer to results as they accumulate data. GDPNow, for example, had forecast a 1.8% drop at the start of the week, but adjusted it to reflect a 1.2% drop on Wednesday.

    Ben Herzon, chief executive officer at the firm.

    Model activity

    We use the “sold-bean method,” says Herzon. “What we do is look at the BEA source data and replicate their methodology.” The trick is to estimate the value for the most recent month of data that has not been made publicly available, he said.

    The Atlanta Fed’s GDPNow model is largely publicly available and also mimics the methods the BEA uses to estimate real GDP growth. One key difference is that it is not adjusted for judgment.

    “The average absolute error of GDPNow’s final projections is 0.84 percentage points,” according to the Atlanta Fed, which stressed that the results were not the official forecasts of the bank, its president. , the Fed system or the Federal Open Market Commission.

    The most historically accurate forecaster of GDP in a Bloomberg survey, Brett Ryan at Deutsche Bank AG, called for a 0.6% drop. Deutsche Bank also has a current GDP tracker that aggregates other models and weighs them “in a historically optimal way,” Matt Luzzetti, senior US economist at the firm, said in a note. an interview at Bloomberg’s Washington office.

    Bloomberg Economics now forecasts a US GDP of minus 1.5% and compares with the official forecast of Bloomberg economists for a growth of 0.7%.

    Other Fed banks also have a variety of models such as the Chicago Fed National Activity Index, the Philadelphia Fed’s Monthly Coincidence Index, and the Aruoba-Diebold-Scotti Business Conditions Index to help gauge activity. economy.

    For the third quarter, as everything goes today, JPMorgan Chase & Co. shows that US GDP increased 1%, while the S&P model increased 0.9%. The Atlanta Fed will release its first-quarter third-quarter estimates on Friday and will update the model more than 30 times until it accumulates with its final third-quarter GDP release on December 22.

    Recent Articles


    Featured Article

    Leave A Reply

    Please enter your comment!
    Please enter your name here

    Stay on op - Ge the daily news in your inbox