Blog

Authored by Michael Lebowitz via RealInvestmentAdvice.com,

The Atlanta Fed’s GDPNow economic forecasting tool predicts an imminent recession, which is fueling investor angst. However, the New York and St. Louis Feds’ Nowcast economic forecasts predict continued economic growth in the first quarter.

Confused?

This article explores the GDPNow and Nowcast models to understand the recent forecast divergences. A better understanding of the two models helps us appreciate the current state of the economy and, therefore, better estimate the first quarter GDP. Importantly, it shows that investor angst over an imminent recession may be unwarranted.

As of March 24, 2025, these are the current forecasts for the three models:

• Atlanta Fed GDPNow: -1.80%

• St. Louis Fed Nowcast: +2.25%

• New York Fed Nowcast: +2.72%

Forecasting Accuracy

Before reviewing the differences in the models, it’s worth viewing historical data to see how well GDPNow and Nowcast forecast GDP. GDPNow and Nowcast update weekly, although, at times, GDPNow has two updates per week. For the analysis below, we only use the final, not interim, estimates for a comparison to GDP

We do not chart the New York Fed Nowcast as there is insufficient data. The y-axis is truncated as the wild economic swings in 2020 made it hard to appreciate the other data. The bar chart beneath the line graph shows the quarterly differences between the forecasts and actual GDP.

Both models have underestimated the GDP for the better part of the last four years. Since 2022, the St. Louis Nowcast has been short by 1.02% on average, while the Atlanta Fed GDPNow has been .44% below, on average. During the pre-pandemic years, the Nowcast ran almost half a percent above GDP on average, while GDPNow was about a quarter of a percent below average.

The bottom line is that the models have flaws. While they are not statistically great estimators of GDP, they are still extremely valuable. It’s important to appreciate that the data calculation underlying GDP is incredibly complex and often revised numerous times after the models’ estimates are final.

Lastly, encapsulating a nation’s economic activity in one number is impossible. Thus, any model trying to forecast GDP is bound to have problems.  

With that, let’s look at the two types of models and assess their strengths and weaknesses.

GDPNow

The Atlanta Fed GDPNow uses a bridge equation approach using the same data the Bureau of Economic Analysis (BEA) uses to calculate its 13 GDP subcomponents. The term bridge equation refers to regression analysis and other statistical tools that link data being released with different time frequencies to predict the quarterly GDP.

The data in the model mimics the methodology used by the BEA. It can be thought of as a running GDP calculation.

The data feeding the model is not predicted; it’s actual data. Therefore, GDPNow estimates at the beginning of a quarter are based on limited data and can result in volatile forecasts. For example, GDPNow fell from +2.5% to -1.8%, solely on the net trade balance data in late February. While it is volatile early in periods, it does tend to produce more accurate forecasts than Nowcast after the quarter ends.

The graph below shows that GDPNow has been in a nearly 7% range just since the beginning of February.

Nowcast

The St. Louis and New York Fed Nowcasts are dynamic factor models. This means they use a lot of economic data and statistical models to determine the relationship of said data to economic GDP. Unlike GDPNow, Nowcast uses data that is not included in the BEA GDP calculations. The large amount of data helps smooth out its estimate. However, while it may be more accurate and less volatile during the quarter, its final estimates are more error-prone than GDPNow’s.

Pros/Cons Table

The following table summarizes the pros and cons of each model.

Which Is Better?

That is a tricky question to answer. They both serve a purpose for those assessing current economic activity.

As we currently see with GDPNow, its forecast can be highly volatile at the beginning of a quarter. However, it is based on the same data that will feed the BEA. In volatile quarters like the current one, it’s best to take the GDPNow forecast with a grain of salt until two months of data are in the books.

The Nowcast forecasts use a much wider array of data; therefore, they’re often more stable. However, their ending forecasts tend to be worse than GDPNow. That said, changes in the Nowcast models throughout the quarter are more telling of trend changes.

Summary

We rely on both models as they provide insight and a more balanced view. Relying on only one model can produce a flawed opinion. For instance, GDPNow warns that GDP could be below zero if specific figures, like the trade balance, remain skewed for the remainder of the month. However, the Nowcast models inform us that the broader economy is generally in good shape. If Nowcast starts rapidly catching down to GDPNow, our economic concern will rise appreciably.