What macro data does (and does not) tell us about earnings

Suppose I’m a time traveler and I tell you what the S&P 500’s earnings will be next year. Just to show off, I give you the number to the third decimal. Would you use this number to bet all your money on a stock market return forecast?

I hope not. You’d also need to predict what will happen to investor sentiment, and I didn’t give you any information about the change in the price-to-earnings ratio.

This thought experiment illustrates why forecasting stock returns is one of the hardest tasks in finance. Remarkably, it’s almost as hard to explain realized returns. Even after the fact, investors do not always know why stocks went up or down, despite the confident narratives you hear daily in the financial media.

Explaining earnings is easier because investor behavior isn’t involved. Let us focus on that first. Then, we’ll forecast earnings and, last, discuss investor sentiment.

We’ve built a factor model that explains earnings based on three macroeconomic variables:

The long-term correlation between model-implied and realized earnings is an impressive historical 85%. It’s a simple model, but it’s powerful:

S&P 500 Earnings = 7% + 2.2x Inflation – 3.0x Unemployment + 1.0x Growth

Where:

Each of the variables is expressed as a year-over-year (YoY) difference, and earnings are expressed as a YoY percentage change. The model was estimated on monthly data from 12/31/1997 to 10/31/2024. To use this model, we insert forecasts for inflation, unemployment, and growth, and obtain a macro-implied forecast for S&P 500 earnings. It’s a plug-and-play model. You’re welcome.

We chose 12-month forward earnings because they’re smoother and better behaved than trailing earnings. Also, forward earnings approximate what’s priced in—hence, year‑over‑year changes should explain stock returns better than other earnings measures. You might be surprised to see that inflation is positively correlated. Inflation is good for earnings. U.S. firms have pricing power and can grow revenues faster than input costs.

Figure 1 compares realized¹ and macro-implied (or “fitted”) earnings. Despite its simplicity, our three-factor model explains changes in forward earnings surprisingly well. In contrast, when we ran this model to explain stock returns, including price-to-earnings ratio changes, the correlation dropped significantly (not reported here).

As the chart shows, our model doesn’t capture all the factors impacting earnings. Realized earnings underperformed macro-implied earnings around 2014-2016, when global growth was weak and the U.S. dollar was strong, diminishing the value of U.S. companies’ revenues earned abroad. Also, during that period, oil prices collapsed, hurting earnings for the energy sector.

In contrast, realized earnings significantly outperformed macro-implied earnings following cuts in the corporate tax rate in 2018. Again, our model missed a factor, in this case changes to the corporate tax rate.

When faced with such deviations, quantitative analysts often seek to add missing factors to improve their models. Done carefully, this can produce better forecasts.

But there’s another way to look at the problem. Adding factors isn’t always beneficial.

Simpler models have two advantages:

1. They help avoid the pesky technical problem of collinearity—in other words, when two factors tend to move in sync—which makes it challenging to isolate the effect of each.

2. They allow fundamental investors to interpret deviations (we call these “residuals”) based on factors that undeniably change over time.

This leads us to today. Figure 2 zooms in on the data over the last two years. S&P 500 earnings have significantly outperformed the broader economy. Based on changes in inflation, unemployment, and PMIs during the previous 12 months, our model implies that earnings should have grown at a rate of 5.5—yet realized earnings grew at a rate of 11%.

Why are earnings outperforming macro-implied earnings? You might have guessed it: the Magnificent Seven (Mag 7) (Alphabet, Amazon.com, Apple, Meta Platforms, Microsoft, NVIDIA, and Tesla). Our model doesn’t capture these companies’ highly scaled technology‑driven earnings power and the boom in artificial intelligence (AI) capital spending. Figure 3 shows that the rest of the S&P’s earnings growth has been better aligned with our macro factors.

Figure four displays our model assumptions, where the inputs are in the gray cells. These are the most recent 12-month changes in each macro factor.

To produce macro-implied earnings growth forecasts, we have estimated four scenarios for the possible direction of these three factors over the next 12 months (Figure 5).

FactSet reports that analysts are expecting 15% earnings growth in 2025. Technically, our model forecasts growth in forward earnings, but this measure is highly correlated with realized earnings growth potential.

Hence, the consensus is that corporate earnings will continue outperforming macro, even under our optimistic macro scenario. Industry analysts are expressing an upbeat view on Mag 7 earnings, tax cuts, and deregulation. Overall, earnings are accelerating, and the Fed is cutting rates. And the new administration seems to want a strong stock market. Taken together, it’s not surprising that most investors are bullish on stocks.

But are they too optimistic? Here are the top four risks I see on the horizon:

1. Valuations are elevated. There isn’t much room for the P/E ratio to continue to rise.

2. Microsoft, Google, Amazon, and Meta’s annual growth in capital spending on AI is expected to decelerate from 45% to 19%.²

3. Government spending cuts, tariffs, and immigration reforms could negatively affect the economy and investor sentiment.

4. Given deficits and the potential for higher-than-expected inflation, a significant rise in the U.S. 10-year Treasury yield remains a possibility, which could hurt stock valuations.

Predicting market returns remains a formidable challenge, even with precise earnings forecasts, because investor sentiment plays a pivotal role in market dynamics. This note focuses on earnings. Our three-factor model—anchored in inflation, employment, and growth—has achieved an impressive 85% historical correlation between macro-implied and realized changes in earnings expectations.

Recent earnings have outperformed the macro data due to AI investments, highlighting an area our model doesn’t capture. As we look ahead, potential risks loom while the market appears poised for growth. Our slightly overweight stance on stocks reflects a balanced view, acknowledging both the opportunities and the inherent risks of a market priced for perfection.

Glossary Terms:

Macro-implied earnings (also known as fitted earnings) are earnings predicted by an economic model.

Realized earnings are earnings changes for the 12 months ahead. Changes in forward earnings are correlated with changes in trailing earnings.

Please see vendor indices disclaimers for more information about the sourcing information: www.troweprice.com/marketdata

For additional definitions of terms, please see:
http://www.troweprice.com/glossary